# Polaris — full site content > The all-in-one workspace where AI workers are teammates. Generated 2026-08-24. 230 pages. --- --- title: "AI worker use cases by department | Polaris" description: "Sixty recurring jobs across twelve company functions, each with the brief an AI worker is given, the connections it needs, and what lands as a delivery comment." url: https://www.polarishq.co/use-cases section: Use cases updated: 2026-08-21 --- # What an AI worker actually does, department by department Twelve functions, sixty recurring jobs, and the exact tool connection each one needs. ## The short answer Polaris use cases are organised by company function: product, engineering, design, data, marketing, sales, support, finance, HR, legal, operations and the executive team. Each page describes one recurring job, the brief an AI worker is given for it, the tool connections that job requires, and what arrives as a delivery comment when the work is done. A human closes every task. - **Departments:** 12 - **Jobs covered:** 60 - **Software:** $0 - **Delivered work:** $2 / human-hour ## Why these pages are sorted by function, not by feature Nobody wakes up wanting an agent. They wake up with forty new bug reports, an interview backlog they have not read, or a metrics email due at nine. The unit that matters is the job, so that is how these pages are cut. Each one names a job that recurs, describes the brief a worker is given for it, lists the connections that job actually requires from the Polaris catalog, and shows what comes back. Where the job needs human judgement, the page says so instead of pretending otherwise. ## What every use-case page tells you The same four things, because these are the four things that decide whether the work gets done. - **The brief** — What the worker is told to do, in the words you would use with a new hire. Capabilities are stored as a SKILL.md file you can open and edit, not a hidden prompt. - **The connections** — Which tools the worker needs from the catalog: Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram or open web search. - **The delivery** — What lands on the task when the machine finishes. Deliveries are comments, and comments can carry files: a PDF, a written spec, a drafted document. - **The human call** — The part of the job that is judgement rather than assembly. The machine never marks a task done. You close it and you rate it. ## Where teams usually start One job per function, picked because it is the one people describe when asked what is eating their week. | Function | The job handed over first | Connections it needs | | --- | --- | --- | | Product | Weekly roadmap drift check | Linear, Notion, Slack | | Engineering | Monday bug triage | GitHub, Linear, Slack | | Design | Design system drift audit | Figma, GitHub, Google Drive | | Data | Daily data quality checks | Supabase, Slack | | Marketing | Competitor monitoring | Web search, Notion, Slack | | Sales | Pre-call account research | HubSpot, web search, Gmail | | Support | Ticket triage and tagging | Gmail, Slack, Notion | | Finance | Invoice and payment tracking | Stripe, Gmail, Google Drive | | HR | First-pass candidate screening | Gmail, Google Drive, Google Calendar | | Legal | Contract review tracking | Google Drive, Gmail, Notion | | Operations | SOP maintenance | Notion, Google Drive, Slack | | Executive | Weekly business review pack | Supabase, Stripe, Notion, Slack | ## Every use case - [Product management with an AI worker on the roster](https://www.polarishq.co/use-cases/product) — Five jobs a product team can hand to an AI teammate, and the ones it should never hand over. - [The engineering work that is not writing code](https://www.polarishq.co/use-cases/engineering) — Five jobs that sit between an engineer and the code, and what an AI teammate does with each. - [Design work an AI teammate can take, and the part it cannot](https://www.polarishq.co/use-cases/design) — Five jobs around the work, none of them the work itself. Taste stays where it belongs. - [One analyst, thirty requests, and an AI worker in between](https://www.polarishq.co/use-cases/data) — Five jobs that stand between a data team and the analysis they were hired to do. - [Your roadmap document and your tracker disagree](https://www.polarishq.co/use-cases/product/roadmap-planning) — A weekly diff between what the roadmap claims and what the tracker actually says. - [Sixty requests, and no idea how many people asked](https://www.polarishq.co/use-cases/product/feature-prioritization) — Clustered requests with a real count of who asked, so the argument is about value rather than recall. - [Twelve interviews recorded, two of them read](https://www.polarishq.co/use-cases/product/user-research-synthesis) — Themes with participant counts and verbatim quotes, plus a flag wherever the evidence is thin. - [You shipped twenty-three things and announced four](https://www.polarishq.co/use-cases/product/release-notes) — A draft built from what actually merged, plus a list of the changes nobody described. - [A competitor changed their pricing and you found out in a sales call](https://www.polarishq.co/use-cases/product/competitive-tracking) — A weekly diff of public competitor pages, reporting only what changed since last time. - [The two hours before planning that nobody schedules](https://www.polarishq.co/use-cases/engineering/sprint-planning) — Carry-over, gaps and blockers assembled the day before, so the meeting is about commitment. - [Forty new issues on Monday, half of them the same bug](https://www.polarishq.co/use-cases/engineering/bug-triage) — Every new issue checked for duplicates, version and repro before an engineer opens it. - [Six open pull requests and nobody knows whose turn it is](https://www.polarishq.co/use-cases/engineering/code-review-workflow) — The review queue, ordered by age and named by who is blocking it, posted every morning. - [The incident ended and the writeup never started](https://www.polarishq.co/use-cases/engineering/incident-postmortems) — The timeline assembled from the channel and the deploy history, with the gaps left honest. - [The README describes a version of the code that no longer exists](https://www.polarishq.co/use-cases/engineering/technical-documentation) — A monthly list of statements in your docs that the code no longer supports. - [Twenty minutes of every design review goes on finding the file](https://www.polarishq.co/use-cases/design/design-review) — The agenda written the day before, with last review's unresolved threads at the top. - [The component was renamed and the documentation was not](https://www.polarishq.co/use-cases/design/design-system-maintenance) — A weekly drift report between the component library and the documentation that describes it. - [Eight sessions recorded, and the same hesitation in six of them](https://www.polarishq.co/use-cases/design/user-testing-synthesis) — Session notes turned into a task-by-task table of where people stalled and what they said. - [Is this final? Asked for the fourth time this month](https://www.polarishq.co/use-cases/design/asset-handoff) — A completeness check before build starts, so the question is answered before it is asked. - [Four versions of the logo are live and nobody signed off on three](https://www.polarishq.co/use-cases/design/brand-consistency-audits) — A quarterly pass over your public surfaces, checked against the rules your brand doc states. - [Can you pull the numbers, sent as a direct message at 6pm](https://www.polarishq.co/use-cases/data/dashboard-requests) — Every ask restated as a question, with the clarifying questions asked before anyone writes SQL. - [The null rate tripled in March and the board slide was already wrong](https://www.polarishq.co/use-cases/data/data-quality-monitoring) — Row counts, null rates, orphans and freshness, checked every morning. Silence means clean. - [The Monday metrics email that somebody writes on Sunday night](https://www.polarishq.co/use-cases/data/reporting-automation) — The numbers pulled and the paragraph written, delivered as a file rather than a link. - [Three teams, three definitions of active user, one meeting](https://www.polarishq.co/use-cases/data/metric-definitions) — Every definition of a metric found and listed side by side, with the source of each. - [Thirty requests in the backlog and one analyst](https://www.polarishq.co/use-cases/data/analysis-backlog) — A scoping note per request, so the backlog is triaged on value rather than on arrival order. - [Marketing work, with AI teammates on the same board](https://www.polarishq.co/use-cases/marketing) — One board for the calendar, the briefs and the campaigns, with workers who draft while nobody is at a desk. - [A content calendar that maintains itself between meetings](https://www.polarishq.co/use-cases/marketing/content-calendar) — The calendar stops rotting the moment somebody other than you is responsible for updating it every week. - [SEO content production, briefed once and delivered as a file](https://www.polarishq.co/use-cases/marketing/seo-content-production) — The difference between usable drafts and filler is the brief, and the brief belongs in a file the worker actually reads. - [Campaign planning where the plan becomes owned tasks](https://www.polarishq.co/use-cases/marketing/campaign-planning) — Most campaigns do not fail at the idea. They fail at the fourteen small things nobody agreed to own. - [Competitor monitoring that actually happens every week](https://www.polarishq.co/use-cases/marketing/competitor-monitoring) — Everyone agrees competitor tracking matters and nobody has done it since the last time a deal was lost over it. - [A week of social posts drafted before you open the app](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) — Drafting is the part that eats the week. Approving is the part that needs you. - [Sales work, with the admin handed to an AI teammate](https://www.polarishq.co/use-cases/sales) — Selling is a conversation. Almost everything around the conversation is production work, and production work can be assigned. - [A pipeline review that is assembled before the meeting](https://www.polarishq.co/use-cases/sales/pipeline-management) — The forecast meeting is worth having. Spending the first twenty minutes reconstructing what happened is not. - [Proposals drafted from your call notes, not from a template](https://www.polarishq.co/use-cases/sales/proposal-writing) — Every proposal is 70% the same document and 30% the reason this customer is different. The 30% is the part worth your evening. - [Account briefs waiting for you before the first call](https://www.polarishq.co/use-cases/sales/lead-research) — Ten minutes of research changes a first call. Nobody has ten minutes before a first call. - [CRM cleanup as a standing job instead of a quarterly panic](https://www.polarishq.co/use-cases/sales/crm-hygiene) — Nobody sets out to let the CRM rot. It rots because cleaning it is a four-hour job with no owner and no deadline. - [Battle cards that are still true this quarter](https://www.polarishq.co/use-cases/sales/sales-enablement-content) — Enablement content is written once, used constantly, and updated never. The updating is the assignable part. - [Support work, with an AI teammate on the queue](https://www.polarishq.co/use-cases/customer-support) — The queue is a conveyor belt of small decisions. Sorting them is mechanical; making them is not. - [Ticket triage that hands you a sorted queue and drafted replies](https://www.polarishq.co/use-cases/customer-support/ticket-triage) — Reading forty messages to find the six that matter is the most expensive hour in a support team's day. - [Help articles that keep up with the product](https://www.polarishq.co/use-cases/customer-support/help-center-maintenance) — Every shipped change quietly makes a help article wrong, and the customer finds out before you do. - [Escalations that do not go quiet after the handoff](https://www.polarishq.co/use-cases/customer-support/escalation-tracking) — The customer's question is not what the bug is. It is whether anyone is still looking at it. - [Customer feedback that reaches product as evidence](https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops) — Support already knows what is wrong with the product. The problem is the format the knowledge arrives in. - [A template library that stays in your team's voice](https://www.polarishq.co/use-cases/customer-support/response-templates) — Templates go stale the same way documentation does, except a stale template gets sent to a customer. - [Running the company with an AI teammate on the reporting](https://www.polarishq.co/use-cases/executive) — The information you need to run the company exists. Assembling it every week is what nobody has time for. - [The weekly review, written before the meeting starts](https://www.polarishq.co/use-cases/executive/weekly-business-review) — A meeting that begins with everyone reading the same document is a different meeting. - [Board packs where the assembly is not your weekend](https://www.polarishq.co/use-cases/executive/board-reporting) — A board pack is 80% data you already have and 20% the story only you can tell. The 80% is what eats the week. - [OKRs that stay visible after the offsite](https://www.polarishq.co/use-cases/executive/okr-tracking) — OKRs do not fail at the writing. They fail in week three, when nobody has looked at them since the offsite. - [Research memos waiting for you in the morning](https://www.polarishq.co/use-cases/executive/strategic-research) — The questions worth researching are the ones you never have a free afternoon for. - [Every commitment from the meeting, owned and dated](https://www.polarishq.co/use-cases/executive/meeting-follow-ups) — The decisions were good. The problem is the eleven commitments that existed only in a document nobody reopened. - [Finance work with an AI worker on the roster](https://www.polarishq.co/use-cases/finance) — The chasing, sorting and assembling that fills a finance week, prepared by a worker you brief once and review every time. - [Invoice tracking that produces a chase list, not a dashboard](https://www.polarishq.co/use-cases/finance/invoice-tracking) — A worker matches Stripe payments to what you invoiced, ages the gap, and drafts the reminder for each account. You decide who actually gets chased. - [Expense categorization that escalates instead of guessing](https://www.polarishq.co/use-cases/finance/expense-categorization) — A worker pulls receipts out of the mailbox, codes them against your own chart of accounts, and puts anything ambiguous in a pile for you rather than picking a category to look finished. - [A monthly close that stops drifting into week two](https://www.polarishq.co/use-cases/finance/monthly-close-checklist) — The close is a chase, not a calculation. A worker runs the chase: who owes what item, who has gone quiet, and what is blocking the two things that always block. - [Budget reports written for the people who did the spending](https://www.polarishq.co/use-cases/finance/budget-reporting) — Plan against actuals is easy to produce and hard to read. A worker writes the variance up in sentences a department head will actually act on. - [Vendor management that catches the notice window](https://www.polarishq.co/use-cases/finance/vendor-management) — Most money lost on vendors is lost by missing a cancellation deadline nobody had written down. A worker keeps the register and dates the deadlines. - [HR work an AI worker can prepare, and where it must stop](https://www.polarishq.co/use-cases/hr) — Scheduling, onboarding logistics, policy drafts and review-cycle admin, prepared by a worker. Every decision about a person stays with a person. - [Application summaries, not candidate scores](https://www.polarishq.co/use-cases/hr/candidate-screening) — A worker reads every application in the same shape against the criteria you published. It does not rank anyone, and it does not reject anyone. - [Onboarding that is finished before the first Monday](https://www.polarishq.co/use-cases/hr/employee-onboarding) — Nobody's first day should start with an apology about accounts. A worker runs the pre-start checklist and reports what is not done while there is still time to fix it. - [Interview scheduling, including the reschedules](https://www.polarishq.co/use-cases/hr/interview-scheduling) — Four calendars, two time zones, a candidate who can only do early mornings, and a panellist who declines twice. This is the job. - [Policy documents that stay current and show their changes](https://www.polarishq.co/use-cases/hr/policy-documentation) — A policy nobody has updated in three years is worse than no policy. A worker keeps the drafts moving and the versions visible. A qualified adviser signs them off. - [Review cycles that finish, without a worker forming an opinion](https://www.polarishq.co/use-cases/hr/performance-review-cycles) — The administration of a review cycle is enormous and the judgment inside it is entirely human. A worker takes the first part and touches none of the second. - [Legal operations work an AI worker can carry](https://www.polarishq.co/use-cases/legal) — Tracking, chasing, assembling and watching. Everything a legal team spends time on that is not actually practicing law. - [Knowing where every contract is, without asking three people](https://www.polarishq.co/use-cases/legal/contract-review-tracking) — The question that eats a legal team's week is "where is that one now?". A worker keeps the answer current and ages every stalled agreement. - [Compliance checklists with evidence attached and gaps named](https://www.polarishq.co/use-cases/legal/compliance-checklists) — A checklist where every line is ticked and nothing is evidenced is not a control. It is a document that will fail an audit slowly. - [Knowing that a source changed, on the week it changed](https://www.polarishq.co/use-cases/legal/policy-updates) — A worker watches the public pages you nominate, quotes what changed, and dates it. What the change means for you is a question for your adviser. - [A register of what you actually signed](https://www.polarishq.co/use-cases/legal/vendor-agreement-management) — Finance tracks what a vendor costs. Legal needs to know what you promised them, what they promised you, and which of those promises has a date on it. - [Watching the open web for uses of your mark](https://www.polarishq.co/use-cases/legal/trademark-monitoring) — A worker sweeps public sources on a schedule, captures dated evidence of each use it finds, and hands you a file. Whether to act is a legal decision. - [Operations work with an AI worker doing the chasing](https://www.polarishq.co/use-cases/operations) — Ops is the function that holds the seams together. Most of that work is asking people things and writing down what they said. - [Getting a process out of one person's head](https://www.polarishq.co/use-cases/operations/process-documentation) — The process exists. It is in someone's habits and in six months of Slack threads. A worker interviews and reads until it is on a page. - [Procurement comparisons built from quotes, not from vendor websites](https://www.polarishq.co/use-cases/operations/vendor-procurement) — A worker researches the field, collects what the vendors actually told you, and lists the questions you have not asked yet. You choose. - [Stock alerts that arrive before you are out](https://www.polarishq.co/use-cases/operations/inventory-tracking) — Polaris does not hold your inventory. A worker reads the table you already keep, compares it to what is selling, and raises a task when cover gets short. - [Finding where the SOP and the real procedure came apart](https://www.polarishq.co/use-cases/operations/sop-maintenance) — Every written procedure starts accurate and drifts. A worker compares the document to how the work actually ran and reports the difference. - [Dependencies as dated items, not as things people said in a meeting](https://www.polarishq.co/use-cases/operations/cross-team-coordination) — Two teams agree something in a call. Neither writes it down. Three weeks later each is waiting for the other. A worker makes the promise a tracked item. > **Nothing delivered, nothing billed** > > The Polaris workspace costs nothing: unlimited people, tasks, workstreams and docs. You pay about $2 per human-hour of delivered work, estimated by an open formula and itemised job by job on the worker's work log. If a line looks wrong, you can challenge it from the log itself. ## Read next - [ai-workers](https://www.polarishq.co/ai-workers) - [integrations](https://www.polarishq.co/integrations) - [cloud-claude-code](https://www.polarishq.co/cloud-claude-code) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) ## Questions people ask **Do I need a different AI worker for every use case?** No. One worker can hold several related jobs, because its capabilities are a SKILL.md file that can describe more than one routine. Teams usually split workers by function rather than by task, so a data worker handles quality checks and the Monday report rather than hiring two. **Can an AI worker close its own tasks?** No, and this is deliberate. A worker posts its output as a comment on the task and ticks the acceptance criteria it was given, but the task stays open until a person closes and rates it. That rating is the review. **What happens to the tools we already use?** They keep working. Polaris workers connect to Slack, Linear, Notion, GitHub, Figma, HubSpot, Stripe, Supabase, Gmail, Google Calendar, Google Drive, WhatsApp and Instagram from a fixed catalog. Teams that want to consolidate can, and teams that want to keep Linear can do that too. **How long does it take to get a worker doing one of these jobs?** Hiring runs as a short chat interview and takes about sixty seconds: a name, a role, what the worker should be great at, and which tools it needs. Editing the SKILL.md afterwards to encode your team's specific rules is what actually makes the output good. ## Related - https://www.polarishq.co/ai-workers - https://www.polarishq.co/integrations - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/for/software-teams --- --- title: "AI workers for product teams | Polaris use cases" description: "Five product jobs an AI teammate can take: roadmap drift, request triage, research synthesis, release notes and competitor tracking. And the five to keep." url: https://www.polarishq.co/use-cases/product section: Use cases updated: 2026-08-21 --- # Product management with an AI worker on the roster Five jobs a product team can hand to an AI teammate, and the ones it should never hand over. ## The short answer A product team in Polaris hires an AI worker for the parts of the job that recur: roadmap drift, request triage, interview synthesis, release notes and competitor tracking. The worker reads Linear, Notion, Slack, Google Drive and the open web, then posts what it found as a comment on the task with files attached. Prioritisation calls, customer conversations and the decision to cut scope stay with people. - **Jobs covered:** 5 - **Usual connections:** Linear, Notion, Slack - **Still yours:** What ships ## The week a product manager actually has The roadmap doc was accurate on the day it was written. Since then two initiatives slipped, one lost its owner when someone changed teams, and the tracker has the real dates while the doc has the ones you showed the board. Meanwhile there are sixty feature requests scattered across Slack threads, support tickets and sales calls, twelve interview recordings nobody has read, a release that shipped without notes, and a competitor who quietly changed their pricing page. None of this is hard. All of it is assembly, and assembly is what eats the week that was supposed to go on the decision. ## The five jobs Each one is a page. Each page names the brief, the connections and what comes back. - **Roadmap planning** — A weekly diff between what the roadmap document claims and what the tracker actually says, with the initiatives that have no owner listed separately. - **Feature prioritisation** — Every open request clustered by theme, counted by how many distinct accounts asked, with the source link kept against each one. - **User research synthesis** — Interview transcripts read end to end and turned into themes with verbatim quotes, participant counts, and a flag where the evidence is thin. - **Release notes** — Merged pull requests since the last tag, grouped into user-visible changes and internal work, with a question list for the ones whose descriptions say nothing. - **Competitive tracking** — A weekly read of named competitors' public pricing and changelog pages, reported as a diff against last week rather than a fresh summary. ## Brief, connections, delivery What you write in the worker's SKILL.md, what you authorise it to reach, and what lands on the task. | Job | The brief | Connections | What arrives | | --- | --- | --- | --- | | Roadmap planning | Compare the roadmap doc to tracker state every Monday | Linear, Notion, Slack | A comment listing moved dates and unowned initiatives | | Feature prioritisation | Cluster open requests, count distinct askers | Slack, Linear, HubSpot | A ranked table with source links per cluster | | Research synthesis | Read the transcripts in this folder, theme them | Google Drive, Notion | A themes document attached to the task | | Release notes | Draft notes from merged PRs since the last tag | GitHub, Linear, Notion | A draft plus a list of PRs it could not interpret | | Competitive tracking | Diff these public pages against last week | Web search, Notion, Slack | Only what changed, with dates and URLs | ## The split that works Product judgement does not survive being handed to a machine. Assembly does. **The product manager keeps** - Talking to customers - Deciding what gets cut - The trade-off argument with engineering - Saying no to the loudest request - Owning the outcome when it was the wrong call **The AI worker takes** - Reconciling the doc against the tracker - Counting who asked for what, and when - Reading twelve transcripts nobody has time for - Turning merged pull requests into a notes draft - Watching public competitor pages every week ## Putting a product worker on the roster About a minute of chat, then the part that matters: writing down how your team works. 1. **Say what is falling through** — Describe the gap to the Chief of Staff in chat. It runs a short interview: a name, the role, what the worker should be great at, which tools it needs. Answers are one click each. 2. **Edit the SKILL.md** — Capabilities land as a readable skill file. This is where you write the rules that make the output yours: what counts as a real request, which competitors matter, how you like release notes grouped. 3. **Authorise the connections** — Linear and Notion for most product work, Slack when the requests live in threads, Google Drive when the transcripts do. Credentials are verified once and stored server-side. 4. **Assign the first task** — Give it acceptance criteria the way you would give a contractor a definition of done. A cloud machine claims the job and starts working, whether or not your laptop is open. 5. **Close it yourself** — Read the delivery comment, keep what is right, rate it. The rating and your comments are what the worker's next run is briefed against. ## The five product use cases - [use-cases/product/roadmap-planning](https://www.polarishq.co/use-cases/product/roadmap-planning) - [use-cases/product/feature-prioritization](https://www.polarishq.co/use-cases/product/feature-prioritization) - [use-cases/product/user-research-synthesis](https://www.polarishq.co/use-cases/product/user-research-synthesis) - [use-cases/product/release-notes](https://www.polarishq.co/use-cases/product/release-notes) - [use-cases/product/competitive-tracking](https://www.polarishq.co/use-cases/product/competitive-tracking) > **What this costs** > > The workspace is free, including unlimited docs, tasks and people. You are billed roughly $2 per human-hour of delivered work, and every hour is itemised on the worker's log with the searches, prose and files that produced the estimate. ## Questions people ask **Can an AI worker decide what goes on the roadmap?** It should not, and Polaris does not let it try. A worker delivers findings as a comment and ticks the acceptance criteria it was given, but the task stays open until a person closes it. Roadmap decisions depend on context that never makes it into a tracker. **Does this replace Linear or Jira for a product team?** It can, because Polaris has workstreams, lanes, a board and a list view over the same data. It does not have to. Many teams keep their tracker and give the worker the Linear connection so it reads issue state without anyone changing tools. **How does a worker know what our team means by a feature request?** Because you write it down in the SKILL.md file. A rule like discount asks from a sales call do not count as product requests is one line, and it changes every subsequent run. The skill file is plain text you can read and edit at any time. **What if the worker gets a synthesis wrong?** You reject the delivery and say why in a comment. Nothing was published, because the worker's output arrives as a comment on the task rather than as a change to a document. The rating and the correction inform how the next run is briefed. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/for/product-teams --- --- title: "AI workers for engineering teams | Polaris use cases" description: "Triage, sprint prep, review queues, postmortems and docs rot. Five engineering jobs an AI teammate can take, with the GitHub and Linear connections each needs." url: https://www.polarishq.co/use-cases/engineering section: Use cases updated: 2026-08-21 --- # The engineering work that is not writing code Five jobs that sit between an engineer and the code, and what an AI teammate does with each. ## The short answer Engineering teams use Polaris AI workers for the coordination layer around the code: bug triage, sprint preparation, review queue management, postmortem assembly and documentation drift. A worker reads GitHub, Linear, Slack and Google Calendar, then posts proposed labels, queues, timelines or draft pages as comments on tasks. Merging, closing issues and deciding what a contributing factor was remain human actions. - **Jobs covered:** 5 - **Usual connections:** GitHub, Linear, Slack - **Never automated:** The merge ## Where engineering time actually goes The interesting problem is not the ticket. It is the forty minutes before the ticket, spent working out whether the bug is a duplicate, whether the reporter is on a version from March, and which of the six open pull requests is blocking the release. None of that is engineering. It is reading, matching and listing, done by the person with the most expensive hour in the room because they are the only one who knows the codebase well enough to do it quickly. ## The five jobs - **Sprint planning** — A pre-meeting brief: carry-over issues with their age, issues with no estimate, work assigned to people who are on leave that week, and what is blocked on another team. - **Bug triage** — Every new issue checked for near-duplicates, version currency and repro completeness, with a proposed label set posted as a comment. Nothing is closed by the machine. - **Code review workflow** — A morning queue of open pull requests sorted by age, showing who is blocking each one, which have failing checks, and which touch code with no test changes. - **Incident postmortems** — A timeline assembled from the incident channel and deploy history, with the gaps marked unknown rather than guessed at. - **Technical documentation** — A monthly pass comparing documented behaviour against the current code, listing statements that are no longer true with file references. ## What a machine can and cannot tell The line is not intelligence. It is whether the answer exists in a system the worker can read. **Readable, so the worker does it** - Which issues look like duplicates of each other - How long a pull request has been open, and who has not reviewed it - Which deploys happened during the incident window - Which config keys the docs mention that no longer exist - Who is on leave during next sprint **Judgement, so a person does it** - Whether a bug is severity one or severity three for your users - Whether the pull request is a good idea - What the contributing factor really was - What the team can commit to this sprint - Whether the fix is worth the regression risk ## Brief, connections, delivery | Job | The brief | Connections | What arrives | | --- | --- | --- | --- | | Sprint planning | Assemble the pre-planning brief every Thursday | Linear, GitHub, Google Calendar | A comment with carry-over, unestimated and blocked work | | Bug triage | Triage new issues against the rubric in your skill file | GitHub, Linear, Slack | One comment per issue with proposed labels and duplicates | | Review workflow | Post the review queue each morning | GitHub, Slack | An ordered queue naming who each PR is waiting on | | Postmortems | Build the timeline when an incident task is assigned | Slack, GitHub, Notion | A timeline draft with gaps marked unknown | | Documentation | Check documented behaviour against the code monthly | GitHub, Notion, Google Drive | A list of false statements with file references | ## The economics of handing over the assembly Numbers from how Polaris meters work, not from a case study. Polaris is in free public beta and has no customer results to quote. - **$0** — For the workspace. Unlimited people, tasks, workstreams and docs - **$2** — Per human-hour delivered. Estimated by an open formula, itemised per job - **8h** — Cap on a single session. Sessions are clamped between five minutes and eight hours ## The five engineering use cases - [use-cases/engineering/sprint-planning](https://www.polarishq.co/use-cases/engineering/sprint-planning) - [use-cases/engineering/bug-triage](https://www.polarishq.co/use-cases/engineering/bug-triage) - [use-cases/engineering/code-review-workflow](https://www.polarishq.co/use-cases/engineering/code-review-workflow) - [use-cases/engineering/incident-postmortems](https://www.polarishq.co/use-cases/engineering/incident-postmortems) - [use-cases/engineering/technical-documentation](https://www.polarishq.co/use-cases/engineering/technical-documentation) > **Polaris does not merge your code** > > A worker given the GitHub connection reads repository state and reports on it. It does not merge, deploy, or close issues on its own. Delivery is a comment on a Polaris task, and the task stays open until an engineer closes it. ## Questions people ask **Is this a coding agent?** No. Polaris workers handle the coordination work around a codebase: triage, queues, timelines and documentation drift. Teams that want a coding agent keep the one they use and give Polaris the jobs that currently sit in nobody's calendar. **How is this different from running Claude Code locally?** A local agent stops when the laptop closes, cannot be assigned work by a teammate, and leaves no shared record. In Polaris a cloud machine claims each job from a queue, works while you are away, and posts the result where the whole team can read it. **Does the worker need write access to our repository?** Not for any of these five jobs. Each one is a read of repository state that gets reported back as a comment on a Polaris task. Connections are authorised once, org-wide, and stored server-side where browsers cannot read them back. **What happens if the worker misreads an issue?** Its output is a proposal on a comment, so a wrong label suggestion costs a glance rather than a cleanup. Rejecting the delivery and saying why in a reply is how the correction gets carried into the next run. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/ai-workers/qa-engineer - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams --- --- title: "AI workers for design teams | Polaris use cases" description: "Review agendas, design system drift, testing synthesis, handoff specs and brand audits: five design jobs for an AI worker with the Figma connection." url: https://www.polarishq.co/use-cases/design section: Use cases updated: 2026-08-21 --- # Design work an AI teammate can take, and the part it cannot Five jobs around the work, none of them the work itself. Taste stays where it belongs. ## The short answer Design teams give Polaris AI workers the bookkeeping around design: assembling review agendas, auditing design system drift, synthesising usability sessions, checking handoff completeness, and running brand consistency audits. A worker with the Figma, Google Drive and Slack connections reads files and comments, then posts a written list with links as a comment on the task. It does not critique, and it does not design. - **Jobs covered:** 5 - **Usual connections:** Figma, Google Drive, Slack - **Never automated:** The critique ## The tax on a design team is not the designing It is the twenty minutes at the start of every review working out which file is current. It is the component that got renamed and now appears under two names in the documentation. It is the engineer asking, for the fourth time this month, whether the frames in that page are final. A designer can do all of this. A designer doing all of this is a designer not designing, and it is the kind of work that quietly expands to fill whatever time it is given. ## The five jobs - **Design review** — The agenda built a day before the meeting: every file up for review, its linked task, the comment threads still unresolved since last time, and what changed in between. - **Design system maintenance** — A weekly drift report: library components with no documentation page, documentation describing components that were renamed, and tokens referenced in docs that no longer exist. - **User testing synthesis** — Session notes turned into a table of task, participants who completed it, where they hesitated, and the verbatim line that showed it. - **Asset handoff** — A completeness check before handoff: which frames are marked ready, which exports exist in the shared drive, and what the ticket references but nobody produced. - **Brand consistency audits** — A quarterly pass over public pages and the social grid against the brand document, listing each departure with a link to the instance. ## Brief, connections, delivery | Job | The brief | Connections | What arrives | | --- | --- | --- | --- | | Design review | Assemble the agenda the day before review | Figma, Linear, Google Calendar | An agenda comment with unresolved threads listed first | | System maintenance | Diff the library against the documentation weekly | Figma, Notion, GitHub | A drift list, grouped by undocumented and stale | | Testing synthesis | Read this session folder, build the task table | Google Drive, Notion | A table of hesitations and quotes, attached as a file | | Asset handoff | Check handoff completeness when the task is assigned | Figma, Google Drive, Linear | A ready or not-ready list with the missing items named | | Brand audits | Audit public surfaces against the brand doc quarterly | Web search, Instagram, Google Drive | A list of departures, each with a link and a rule cited | > **An audit is only as good as the document it audits against** > > A worker checking brand consistency compares what it finds to the rules written in your brand document. If the document says the logo needs breathing room, the audit will be vague in exactly that way. Specific rules produce specific findings, and writing them down is the work no machine does for you. ## What the worker never does It does not say whether a design is good. It does not resolve a comment thread, choose between two directions, or decide that a component should be deprecated. Those are the decisions the job exists for. What it does is arrive at the review with the agenda already written, so the hour goes on the argument that matters instead of on finding the file. ## The five design use cases - [use-cases/design/design-review](https://www.polarishq.co/use-cases/design/design-review) - [use-cases/design/design-system-maintenance](https://www.polarishq.co/use-cases/design/design-system-maintenance) - [use-cases/design/user-testing-synthesis](https://www.polarishq.co/use-cases/design/user-testing-synthesis) - [use-cases/design/asset-handoff](https://www.polarishq.co/use-cases/design/asset-handoff) - [use-cases/design/brand-consistency-audits](https://www.polarishq.co/use-cases/design/brand-consistency-audits) ## Questions people ask **Can a Polaris worker produce design files?** No. Workers deliver written output as comments on tasks, and those comments can carry generated files such as documents and PDFs. Figma files stay authored by designers; the connection is used to read file state, comments and structure. **Does this replace our design documentation in Notion?** Only if you want it to. Polaris has a nested document tree with a block editor, versioning, file review and comments, so design system documentation can live there. Teams that keep Notion give the worker the Notion connection instead. **How does the worker know which frames are ready for handoff?** By the convention your team already uses, written into the worker's SKILL.md file. If ready means a frame in a page called Ready for dev, that is one line in the skill file and the check becomes reliable from the next run onward. **Is the brand audit checking our live website?** It checks the public pages you name, using the open web search connection, plus the Instagram connection for the social grid. It cannot see anything behind a login, and pages it could not reach are listed rather than skipped silently. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/integrations/figma - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/for/design-studios - https://www.polarishq.co/glossary/skill-file --- --- title: "AI workers for data teams | Polaris use cases" description: "Dashboard requests, quality checks, reports, metric definitions and the analysis backlog. Five data jobs for an AI worker with the Supabase connection." url: https://www.polarishq.co/use-cases/data section: Use cases updated: 2026-08-21 --- # One analyst, thirty requests, and an AI worker in between Five jobs that stand between a data team and the analysis they were hired to do. ## The short answer Data teams use Polaris AI workers for the queue in front of the analysis: scoping incoming dashboard requests, running daily quality checks, assembling recurring reports, reconciling conflicting metric definitions, and writing scoping notes for the backlog. A worker with the Supabase, Stripe and Slack connections posts findings as a comment on the task. Production changes and query approval stay with a person. - **Jobs covered:** 5 - **Usual connections:** Supabase, Stripe, Slack - **Normal output:** Silence ## The request that arrives as a direct message Can you pull the numbers for last month, broken out by plan. No decision named, no deadline, no indication of whether the answer changes anything. It arrives in a direct message, it does not become a ticket, and three weeks later somebody asks why the backlog is not moving. Underneath that, a null rate quietly tripled in one column in March and nobody noticed until a board slide was already wrong. Data teams do not lose time to hard analysis. They lose it to intake, reconciliation and checks that nobody has automated because writing the check has never been the urgent thing. ## The five jobs - **Dashboard requests** — Every incoming ask restated as a question, with the tables it would need named and the two clarifying questions asked before anyone writes SQL. - **Data quality monitoring** — Daily checks defined in the worker's skill file: row counts against yesterday, null rates on critical columns, orphaned foreign keys, freshness of the newest row. It posts only when something trips. - **Reporting automation** — The recurring report assembled and written, saying what moved and by how much, delivered as a file on the task rather than a link to a dashboard. - **Metric definitions** — Every place a metric is defined, found and listed side by side, so the three incompatible versions of active user are visible in one comment. - **Analysis backlog** — A scoping note per request before it enters the queue: the decision it informs, the data required, whether that data exists, and the cheapest version of the answer. ## Setting up the daily quality check The one most data teams put on the roster first, because it runs every day and costs almost nothing when nothing is wrong. 1. **Hire the worker in chat** — Tell the Chief of Staff you need daily checks on the warehouse. The interview asks for a name, the role and the tools. It takes about a minute. 2. **Write the checks into the SKILL.md** — List the tables that matter, the columns that must never be null, and the thresholds that count as a problem. This file is plain text and you edit it directly. 3. **Authorise Supabase and Slack** — One click each. Credentials are verified at connect time and stored server-side, so the worker uses them and browsers never read them back. 4. **Create the recurring task** — Put it in the data workstream with acceptance criteria stating that a clean run reports nothing. A cloud machine claims the job each morning. 5. **Read only the exceptions** — When a threshold trips, the delivery comment names the table, the column, yesterday's value and today's. When nothing trips, you get a closed task and a short line on the work log. ## Brief, connections, delivery | Job | The brief | Connections | What arrives | | --- | --- | --- | --- | | Dashboard requests | Scope every incoming ask before it is queued | Slack, Supabase, Notion | A restated question plus the clarifying questions to ask | | Quality monitoring | Run the checks daily, report only exceptions | Supabase, Slack | An exception comment naming table, column and delta | | Reporting | Assemble and write the recurring report | Supabase, Stripe, Google Drive | A written report attached as a file on the task | | Metric definitions | Find every definition of these metrics | Supabase, Notion, Slack | A side-by-side list of definitions with their sources | | Analysis backlog | Write a scoping note per request | Supabase, Notion, Linear | A note naming the decision, the data and the cheap version | ## Where the line sits **The worker does** - Read schema and row-level state through the Supabase connection - Compare today's counts against yesterday's - Find every document and thread where a metric is defined - Write the paragraph that says what moved - Ask the clarifying question the requester skipped **A person does** - Approve anything that writes to production - Ratify which definition of a metric wins - Decide which analysis is worth doing - Say what caused the movement - Close and rate the delivery ## The five data use cases - [use-cases/data/dashboard-requests](https://www.polarishq.co/use-cases/data/dashboard-requests) - [use-cases/data/data-quality-monitoring](https://www.polarishq.co/use-cases/data/data-quality-monitoring) - [use-cases/data/reporting-automation](https://www.polarishq.co/use-cases/data/reporting-automation) - [use-cases/data/metric-definitions](https://www.polarishq.co/use-cases/data/metric-definitions) - [use-cases/data/analysis-backlog](https://www.polarishq.co/use-cases/data/analysis-backlog) ## Questions people ask **Does the worker run queries against our production database?** It uses the Supabase connection you authorise, with whatever access that connection carries. Teams that care about this give the worker a read path and keep anything that writes behind human approval, which is also how the five jobs on this page are briefed. **What does a quality check cost if nothing is wrong?** Very little. Hours are estimated from observable effort: pickup time, searches performed, prose written and files produced, clamped to a minimum of five minutes per session. A clean run produces almost no prose and no files, so it lands at the floor. **Can it decide which definition of a metric is correct?** No. It finds and lists the definitions that exist, with the source of each, which is usually the part nobody has done. Ratifying one definition is a decision with consequences for every dashboard, so a person makes it and the document that records it is versioned. **Where does the report end up?** As a comment on the task, with the generated file attached. Deliveries in Polaris are comments rather than silent updates, so there is a record of what was produced, when, and by which run of which worker. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "AI roadmap planning: catch the drift weekly | Polaris" description: "Brief an AI worker to diff the roadmap doc against tracker state every Monday: moved dates, unowned initiatives, and the slips nobody announced." url: https://www.polarishq.co/use-cases/product/roadmap-planning section: Use cases updated: 2026-08-21 --- # Your roadmap document and your tracker disagree A weekly diff between what the roadmap claims and what the tracker actually says. ## The short answer Roadmap planning with an AI worker means a weekly reconciliation rather than a quarterly rewrite. The worker reads every initiative in the roadmap workstream, compares tracker state to the dates written in the roadmap document, and posts a comment listing what moved, by how long, and which initiatives now have no owner. The product manager decides what to cut, resequence or tell the board. - **Runs:** Weekly - **Connections:** Linear, Notion, Slack - **Output:** A drift comment ## Roadmaps do not go wrong all at once They go wrong two days at a time. An initiative slips a sprint and nobody updates the document, because updating the document is a twenty-minute job with no deadline attached. Three of those in a quarter and the roadmap you present is a work of fiction that everyone in the room politely agrees with. The information needed to fix this already exists. It is in the tracker, it is accurate, and reconciling it against the document is pure comparison work. ## How to brief a roadmap worker Five minutes of setup, then it runs without being asked. 1. **Point it at the roadmap document** — Name the document in the worker's SKILL.md: where it lives, what each column means, and which field holds the committed date. If the roadmap lives in Polaris Docs it is versioned, so every weekly state is recoverable. 2. **Define what counts as drift** — Two weeks of slip may be noise for a six-month initiative and a crisis for a two-week one. Write your threshold down rather than leaving the worker to guess at it. 3. **Give it Linear and Notion** — Linear for issue and project state, Notion if the roadmap document lives there. Both are one click from the connections catalog and are authorised once for the whole organisation. 4. **Set the acceptance criteria** — A useful set: every initiative in the document accounted for, every date discrepancy stated with both dates, every unowned initiative listed separately. The worker ticks these as it goes. 5. **Read it before planning** — Schedule the task for Monday morning so the delivery comment is waiting before your planning meeting rather than being written during it. ## What the delivery comment contains One table, three lists, no narrative. | Section | What it holds | Why it is separate | | --- | --- | --- | | Moved dates | Initiative, document date, tracker date, days of slip | This is the part you take to the board | | Unowned | Initiatives with no assignee in the tracker | Usually the result of someone changing teams | | Not in the tracker | Roadmap items with no matching issues at all | Often the ones that were never actually started | | Not in the roadmap | Active projects the document does not mention | Scope that arrived without a decision | > **The worker never edits the roadmap** > > It proposes. The delivery arrives as a comment on the task, with the revised document attached as a file if you asked for a draft. Applying it is your call, because a date that slipped is often a conversation rather than a correction. ## What changes after a month of this - **Slips get discussed while they are small** — A two-day slip surfaced on Monday is a scheduling note. The same slip found in week nine is a missed commitment. - **Ownership gaps stop hiding** — Initiatives lose their owner quietly when people change teams. A weekly list of unowned work makes that a five-second fix. - **The document becomes worth reading** — A roadmap that is reconciled weekly is one people check before asking. One that is rewritten quarterly is one people ignore. ## Questions people ask **What if our roadmap is a spreadsheet rather than a document?** Put it in Google Drive and give the worker the Google Drive connection. The brief is the same: name where the committed date lives and what each column means. The connection catalog covers Google Drive, Notion and Linear, which between them hold most roadmaps. **Can the worker update the tracker to match the roadmap?** That would be backwards. The tracker holds what is true and the document holds what was promised, so the reconciliation runs in one direction and the fix is a human decision about the promise. The worker delivers the comparison, not a correction. **How much does a weekly roadmap check cost?** Hours are estimated from observable effort and billed at about $2 per human-hour delivered, with a five-minute floor per session. A reconciliation over a few dozen initiatives is a short session, and every line of the estimate is visible on the worker's work log. **Does this work if half the roadmap lives in someone's head?** Partly, and the report will say so. Initiatives with no matching tracker issues appear in their own list, which is usually the fastest way to discover which commitments were never actually started. ## Related - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/product/feature-prioritization - https://www.polarishq.co/use-cases/engineering/sprint-planning - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/glossary/workstream --- --- title: "Feature prioritization with an AI worker | Polaris" description: "An AI worker clusters every feature request across Slack, tickets and the CRM, counts distinct accounts per theme, and keeps the source link. You still pick." url: https://www.polarishq.co/use-cases/product/feature-prioritization section: Use cases updated: 2026-08-21 --- # Sixty requests, and no idea how many people asked Clustered requests with a real count of who asked, so the argument is about value rather than recall. ## The short answer Feature prioritisation with an AI worker starts with counting. The worker collects open requests from Slack threads, tracker issues and CRM notes, clusters them by theme, counts how many distinct accounts asked for each, and keeps a source link against every mention. The delivery is a ranked table posted as a comment. Weighting the themes and choosing what to build remains a human decision. - **Runs:** Before each planning cycle - **Connections:** Slack, Linear, HubSpot - **Output:** A counted, sourced table ## Prioritisation arguments are usually memory arguments Someone says customers keep asking for this. What they mean is that three customers asked, one of them twice, and the last time was in a call four months ago that stuck in their head. Someone else remembers a different set. Neither person is lying and neither can produce the list. The requests exist. They are in Slack threads, in support tickets, in tracker issues and in the notes attached to deals. They have simply never been counted, because counting them by hand takes a full day and expires the moment a new request arrives. ## What the worker is briefed to do - **Collect, with the source kept** — Every mention gets a link back to the thread, ticket or issue it came from. A cluster with no traceable sources is a cluster you should not trust. - **Cluster by what was asked for, not by wording** — Export to CSV and can I get this as a spreadsheet are the same request. The rules for what merges live in the worker's SKILL.md, so they are yours to tighten. - **Count distinct accounts, not mentions** — One vocal customer asking eight times is not eight customers. This distinction is the entire value of the exercise and it is where hand-counting usually fails. - **Flag the singletons** — Requests with exactly one asker are listed separately rather than ranked. Some of them are the most important thing on the page, and burying them in a ranking hides that. ## What each connection contributes | Connection | What it holds | What the worker takes from it | | --- | --- | --- | | Slack | Requests made in channels and threads | The ask, who made it, the thread link, the date | | Linear | Requests already filed as issues | Existing issue text and any linked duplicates | | HubSpot | Requests raised in deals and calls | Notes attached to accounts, and which account asked | | Notion | Requests captured in meeting notes | Anything written down outside the tracker | ## Counting is not deciding **The worker supplies** - How many distinct accounts asked - When each one asked, and where - Which requests are already filed - Which requests only one person has ever raised **You supply** - Which accounts you actually want to serve - What the request costs to build - Whether the request is the real problem - The decision, and the reason you will give for it > **A count is not a mandate** > > The most-requested feature is frequently the most-requested workaround for a problem the customer has already given up describing. The table tells you what people asked for. Working out what they needed is still the job. ## Questions people ask **How does the worker avoid double-counting the same request?** By counting distinct accounts rather than mentions, and by keeping every source link so a cluster can be audited. Where two mentions look like the same account under different names, they appear in the delivery comment as an ambiguity for you to resolve. **Can it pull requests from our support inbox?** Yes, through the Gmail connection if support runs on email. The connection catalog is fixed: Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web search. **Will it invent a request that nobody made?** Every clustered item carries the links it was built from, so a claim with no source is visible immediately. Requiring a source link per mention is worth writing into the acceptance criteria of the task, and the worker ticks those criteria as it works. **How often should this run?** Before each planning cycle rather than daily. Request volume does not change much week to week, and a table that is regenerated constantly stops being read. Monthly is enough for most teams under fifty people. ## Related - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/product/roadmap-planning - https://www.polarishq.co/use-cases/product/user-research-synthesis - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/glossary/ai-worker --- --- title: "AI user research synthesis from transcripts | Polaris" description: "An AI worker reads every interview transcript in a folder and returns themes with verbatim quotes, participant counts, and a flag where evidence is thin." url: https://www.polarishq.co/use-cases/product/user-research-synthesis section: Use cases updated: 2026-08-21 --- # Twelve interviews recorded, two of them read Themes with participant counts and verbatim quotes, plus a flag wherever the evidence is thin. ## The short answer User research synthesis with an AI worker means the transcripts get read. The worker is given a folder of interview transcripts through the Google Drive connection, reads all of them, and returns themes with the exact quote that supports each one, the number of participants who raised it, and an explicit note where a theme rests on one or two people. The synthesis document arrives attached to the task. - **Input:** A folder of transcripts - **Connections:** Google Drive, Notion - **Output:** Themes, quotes, counts ## The research that got done and never got used Twelve conversations, roughly fourteen hours of recording, and a transcript folder that fills up faster than anyone opens it. The person who ran the interviews remembers the two that were vivid. The rest inform nothing, which means the research budget bought a feeling rather than a finding. Reading twelve transcripts properly takes most of a day. It is the kind of task that never wins against anything urgent, and it stays undone until the quarter ends and the folder is stale. ## What good synthesis output looks like The rules worth writing into the worker's skill file, because they are what separate a synthesis from a summary. - **Every theme carries a verbatim quote** — Not a paraphrase. The participant's own words, with the participant identifier, so anyone can go back to the transcript and check. - **Counts are stated, not implied** — Four of twelve participants, not many participants. The difference decides whether a theme is a finding or a coincidence. - **Thin evidence is labelled thin** — A theme built on one person is worth recording and worth flagging. Instruct the worker to mark it rather than to promote it into the same list as everything else. - **Disagreement survives** — Where participants contradicted each other, both sides appear. Synthesis that resolves every tension has smoothed away the interesting part. ## The shape of the delivered document | Section | Contents | | --- | --- | | Method | How many participants, which segment, dates of the sessions | | Themes | Each with a participant count, verbatim quotes and an evidence strength note | | Contradictions | Where participants disagreed, with both quotes | | Unanswered | Questions the interviews raise but do not answer | | Source map | Which transcript each quote came from | > **The worker cannot run the interview** > > It reads what was recorded. If the questions were leading, the synthesis will faithfully report a leading result. Talking to customers stays a human job, and it is the half of research where the value is actually created. ## Where the synthesis lands The delivery is a comment on the task with the synthesis document attached. In Polaris, documents are versioned and support review comments, so the synthesis can be argued with in place rather than forwarded around as an attachment that forks into four versions. The task stays open until you close it. Closing is where you decide the synthesis is fair, and the rating you leave is the review that informs the next run. ## Questions people ask **Does the worker transcribe recordings?** No. It reads transcripts and notes you already have in Google Drive. Recording and transcription stay with whatever tool you use today, and the worker picks up from the text. **How is this different from usability testing synthesis?** Discovery interviews produce themes about problems, needs and context. Usability sessions produce task-level failures: where someone hesitated, what they clicked, what they could not find. The output shapes differ enough that they are briefed as separate jobs. **Can I trust the quotes?** They are copied from the transcripts and each one names its source file, so any quote can be checked in under a minute. Requiring a source reference per quote is worth writing into the task acceptance criteria, which the worker ticks as it works. **What if the transcripts are in different formats?** Mixed notes and transcripts in one folder are normal and the worker handles them, but the method section of the delivery will say what it was working from. A synthesis built partly on rough notes is weaker evidence, and the document should say so. ## Related - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/design/user-testing-synthesis - https://www.polarishq.co/use-cases/product/feature-prioritization - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/glossary/delivery-comment --- --- title: "AI release notes from merged pull requests | Polaris" description: "An AI worker reads merged pull requests since the last tag, drafts release notes split user-visible from internal, and asks about the ones it cannot interpret." url: https://www.polarishq.co/use-cases/product/release-notes section: Use cases updated: 2026-08-21 --- # You shipped twenty-three things and announced four A draft built from what actually merged, plus a list of the changes nobody described. ## The short answer Release notes from an AI worker are assembled from merged pull requests since the last tag. The worker reads each pull request and its linked issue through the GitHub and Linear connections, separates user-visible changes from internal work, drafts notes in your house format, and returns a question list for the changes whose descriptions explain nothing. A person edits and publishes; the worker never publishes anything. - **Trigger:** Each release cut - **Connections:** GitHub, Linear, Notion - **Output:** Draft plus a question list ## Release notes lose to everything They are written after the release, by the person most relieved that the release is over, for an audience whose reaction nobody measures. So they get written for the three changes somebody remembered and skipped for the rest, and customers learn about improvements six weeks later by accident. The raw material is complete and sitting in the repository. Every merged pull request has a title, a description of some quality, and usually a linked issue explaining why the work happened. ## How the draft gets built 1. **Find the boundary** — Everything merged since the previous tag. The worker states the range it used at the top of the delivery, so you can tell immediately if it started in the wrong place. 2. **Split user-visible from internal** — A refactor and a new export button both merged. Only one belongs in the notes. Where the split is ambiguous, the change appears in a third list rather than being silently sorted. 3. **Follow the link to the issue** — Pull request descriptions explain what changed. Linked issues explain why it mattered, which is what a customer-facing note needs. The worker reads both. 4. **Write in your format** — Grouping, heading style, whether you name contributors, whether you link the issue. These live in the worker's SKILL.md as rules, so the draft arrives in your shape rather than a generic one. 5. **Ask about the rest** — Pull requests titled fix stuff get listed as questions with links, not guessed at. That list is usually short and takes two minutes to answer. ## What the worker can tell, and what it has to ask | From the repository | Only from you | | --- | --- | | What files changed and when it merged | Whether the change is worth announcing | | The pull request title and description | How to describe it to a non-technical customer | | The linked issue and its labels | Whether it closes a promise made to a specific account | | Who wrote it and who reviewed it | Whether this release deserves a headline at all | > **The draft is a comment, not a publication** > > The worker delivers into the task. Nothing reaches a changelog, a customer email or a Slack announcement until a person moves it there. That is the same rule as everywhere else in Polaris: agents deliver, humans close. ## The side effect worth having - **Bad pull request descriptions become visible** — The question list is a weekly report on which changes shipped without an explanation. Teams that read it tend to write better descriptions within a month, which is a bigger win than the notes. - **The gap between shipped and announced closes** — Once a draft exists on every release cut, publishing becomes an edit rather than a writing task, and edits actually get done. ## Questions people ask **Does the worker need write access to the repository?** No. It reads merged pull requests and linked issues through the GitHub connection and delivers the draft as a comment on a Polaris task. Publishing is a human action taken wherever your changelog actually lives. **What if our pull request descriptions are useless?** Then the question list will be long, which is the honest result. The worker is briefed to ask rather than to invent, so a change it cannot interpret appears as a link and a question instead of a confident sentence that turns out to be wrong. **Can it write notes for a mobile release too?** Yes, if the changes are in a repository the GitHub connection can reach. Store-listing text has its own constraints, so most teams write that rule into the skill file rather than reusing the web format. **How long does a release notes job take?** Long enough to read every merged pull request in the range, which the work log records job by job. Billing is about $2 per human-hour delivered, estimated from observable effort including the prose produced, and every line can be challenged from the log. ## Related - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/use-cases/engineering/code-review-workflow - https://www.polarishq.co/integrations/github - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work --- --- title: "AI competitor tracking, weekly and diffed | Polaris" description: "An AI worker rereads named competitors' public pricing and changelog pages every week and reports only what changed, with the date and the URL for each change." url: https://www.polarishq.co/use-cases/product/competitive-tracking section: Use cases updated: 2026-08-21 --- # A competitor changed their pricing and you found out in a sales call A weekly diff of public competitor pages, reporting only what changed since last time. ## The short answer Competitive tracking with an AI worker is a weekly diff rather than a research report. The worker rereads the public pricing, changelog and product pages of the competitors you name, compares them against the snapshot it stored last week, and posts only what changed, with the date and the URL for each change. Everything behind a login or a sales call stays invisible to it. - **Runs:** Weekly - **Connections:** Web search, Notion, Slack - **Output:** Changes only ## The problem with competitor research is that it expires Somebody builds a thorough comparison deck in January. By March two of the six competitors have changed their pricing, one shipped the feature the deck said they lacked, and nobody has reopened the file. The deck is now worse than nothing, because people quote it. Tracking is a different job from researching. It needs to be boring, it needs to happen every week, and it needs to report deltas rather than restate the whole landscape. ## How the weekly run is briefed - **A named list, not a category** — Six named competitors and the specific pages that matter for each: pricing, changelog, the product page for the overlapping feature. Vague briefs produce vague scanning. - **Snapshot, then diff** — The worker stores what it read as a document, so next week's run compares against text rather than against memory. Documents in Polaris are versioned, so the history of a competitor's pricing page accumulates on its own. - **Report changes, not the landscape** — A week with no changes should produce a short comment saying nothing changed. Reports that restate everything every week stop being read by week three. - **Date and link every claim** — Each reported change carries the URL and the date it was observed. That is what makes the report usable in a sales conversation without someone having to re-verify it. ## What the worker can see, and what it cannot Worth being blunt about, because competitive intelligence tools tend to imply otherwise. | Visible | Not visible | | --- | --- | | Public pricing pages and plan tables | Discounts given in negotiation | | Public changelogs and release posts | Anything shipped behind a feature flag | | Documentation and help centres | Roadmaps discussed with customers under NDA | | Job listings and public announcements | Headcount, revenue or churn | > **What a research job like this actually costs** > > In the recorded Polaris demo, a competitor pricing brief with live web research was delivered in about seven minutes for roughly $2.80, against an estimated 1 hour 24 minutes and about $70 from a $50-per-hour analyst. That recording is a real machine session, sped up rather than staged. ## Where the report goes As a comment on the task, and optionally into a Slack channel through the Slack connection, so the people who need it in a sales call see it without opening the workspace. The Polaris Inbox works in the other direction too. A message in Slack pointing at something a competitor announced arrives as a prefilled task suggestion, with the bucket, lane, labels and owner already set, waiting for one click to become a real task. ## Questions people ask **How does the worker access competitor websites?** Through the open web search connection, which is part of the Polaris connection catalog. It reads public pages the same way anyone with a browser would, and it lists any page it could not reach rather than passing over it quietly. **Can it monitor a competitor's social accounts?** The Instagram connection covers Instagram. For other networks, the worker is limited to what is publicly reachable through web search. Anything requiring a logged-in session is outside what the connection catalog provides. **Will the report be different from what our sales team already knows?** Often it confirms what one person suspected and nobody had confirmed. The value is the date and the URL attached to each change, which turns a hallway claim into something quotable on a call. **How many competitors is too many?** Past about eight, the weekly report stops being read. Most teams track the three they lose deals to and check the rest quarterly, which is also cheaper because billing follows the work actually delivered. ## Related - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/product/feature-prioritization - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "AI sprint planning prep, done before the meeting | Polaris" description: "An AI worker assembles the sprint planning brief: carry-over age, unestimated issues, owners on leave, and cross-team blockers. The team decides scope." url: https://www.polarishq.co/use-cases/engineering/sprint-planning section: Use cases updated: 2026-08-21 --- # The two hours before planning that nobody schedules Carry-over, gaps and blockers assembled the day before, so the meeting is about commitment. ## The short answer Sprint planning preparation is assembly work an AI worker can do overnight. The worker reads the tracker, the repository and the team calendar, then posts a brief listing carry-over issues with their age, issues with no estimate, work assigned to people on leave during the sprint, and items blocked on another team. The team still decides what it can commit to. - **Runs:** The day before planning - **Connections:** Linear, GitHub, Google Calendar - **Output:** A pre-planning brief ## Planning meetings mostly reconstruct state The first half hour goes on remembering why an issue carried over, discovering that three items have no estimate, and realising that the person who owns the migration is away for the first week of the sprint. By the time everyone has the same picture, the energy for the actual decision is gone. All of that state is retrievable before anyone sits down. It just requires someone to spend an hour in four systems the evening before, and nobody volunteers for that. ## What the brief contains Four lists, each answering a question that otherwise gets asked out loud. | List | The question it answers | Where it comes from | | --- | --- | --- | | Carry-over, by age | How long has this been rolling? | Tracker issue history | | No estimate | What are we about to commit to blind? | Tracker fields | | Owner unavailable | Who is away while their work is scheduled? | Google Calendar | | Blocked externally | What is waiting on another team? | Issue links and labels | | Open pull requests | What is nearly done but not merged? | GitHub | ## The parts that matter more than they look - **Age, not count** — Ten carry-over issues is a number. One issue that has carried over five sprints is a problem, and it only shows up if the brief reports age. - **Leave against ownership** — Cross-referencing the calendar against assignments catches the single most common planning failure, which is committing to work whose owner is on a plane. - **Nearly-merged work** — Open pull requests are the cheapest capacity in the sprint. Listing them first puts that capacity in front of the team before anyone reaches for new work. ## Assembly against decision **The worker brings** - Every issue's current state and age - The gaps in estimates and ownership - Calendar conflicts against assignments - The list of external blockers with links **The team decides** - What the sprint goal is - How much the team can honestly commit to - Which carry-over should be dropped rather than rolled - Whether to estimate or to split > **It runs while you sleep** > > Assign the task on Thursday evening and a cloud machine claims it from the job queue and works whether or not anyone's laptop is open. The brief is a comment on the task by the time the team logs in, which is the difference between a cloud worker and an agent running on a local machine. ## Questions people ask **Does this work if we use Jira rather than Linear?** The connection catalog covers Linear, and Polaris itself has workstreams, lanes and a board that can hold the sprint directly. Teams on other trackers usually run the sprint in Polaris and give the worker GitHub and Google Calendar for the rest of the picture. **Can the worker assign issues for the next sprint?** No. It delivers the brief as a comment on a task and the task stays open until a person closes it. Assignment during planning is a negotiation about capacity, and it happens with people in the room. **How does it know who is on leave?** Through the Google Calendar connection, using whatever convention your team already uses to mark time off. Naming that convention in the worker's SKILL.md is what makes the check reliable rather than approximate. **Is it worth running for a two-week sprint?** Most teams find the leave and estimate checks pay for themselves in the first cycle, and hours are metered per job with a five-minute floor rather than charged per seat. If a run stops being useful, you stop assigning the task and there is nothing to cancel. ## Related - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/engineering/bug-triage - https://www.polarishq.co/use-cases/product/roadmap-planning - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed - https://www.polarishq.co/ai-workers/project-coordinator --- --- title: "AI bug triage: duplicates, repro and severity | Polaris" description: "An AI worker checks each new issue for near-duplicates, version currency and missing repro steps, then proposes labels. No machine closes an issue." url: https://www.polarishq.co/use-cases/engineering/bug-triage section: Use cases updated: 2026-08-21 --- # Forty new issues on Monday, half of them the same bug Every new issue checked for duplicates, version and repro before an engineer opens it. ## The short answer Bug triage with an AI worker means every new issue arrives pre-checked. The worker searches existing issues for near-duplicates, verifies whether the reported version is current, extracts repro steps into a consistent shape, and proposes a severity label using the rubric written in its skill file. Each proposal is posted as a comment. Closing, merging duplicates and setting real severity stay with an engineer. - **Runs:** Daily, or on arrival - **Connections:** GitHub, Linear, Slack - **Never automated:** Closing an issue ## Triage is expensive because of who has to do it Deciding whether a report is a duplicate requires knowing the codebase. So the person doing triage is an engineer, and the forty minutes they spend before lunch reading reports is forty minutes of the most expensive time available, spent on matching text against text. Worse, it is interrupt-shaped. Reports arrive all day, triage happens in batches, and the reports that arrive after the batch sit untouched until the next one. ## What the worker does with each new report In this order, because each step can make the next unnecessary. 1. **Search for near-duplicates** — Against open and recently closed issues, matching on symptom rather than wording. Candidates are listed with links and a note on why they look similar, so the engineer confirms in seconds instead of searching. 2. **Check the reported version** — If the reporter is three releases behind and the symptom matches something already fixed, that fact belongs at the top of the comment. This one check resolves a meaningful share of reports on its own. 3. **Normalise the repro** — Steps, expected, actual, environment. Reports arrive as paragraphs; engineers need a shape. What is missing is listed as missing rather than filled in with a guess. 4. **Propose severity** — Using the rubric in the worker's SKILL.md, which you wrote. Data loss is severity one, cosmetic on a settings page is severity three, and the worker cites which rule it applied. 5. **Post, and stop** — The proposal lands as a comment on the task. The machine does not label, close, merge or assign in your tracker. An engineer spends ten seconds accepting or overriding. ## Signal in the report, and what the worker does with it | What the report contains | What the worker does | | --- | --- | | A stack trace | Searches issues for the same top frames and links matches | | A version number | Compares against the current release and flags staleness | | A screenshot only | Records that there are no written repro steps and says so | | Wording close to an existing issue | Proposes a duplicate link with both issue references | | No environment details | Lists the missing fields for the reporter to fill in | > **Severity is a business decision wearing a technical costume** > > Whether a broken export is severity one depends on who exports and why, which is not in the issue text. The worker applies the rubric it was given and names the rule it used, so an override is a judgement call rather than an argument about what the machine was thinking. ## What changes for the engineer - **Triage becomes review** — Instead of investigating forty reports, an engineer reads forty proposals and disagrees with a handful. The expensive part of the hour disappears. - **Reports stop aging** — A cloud machine claims triage jobs from a queue, so a report filed at 11pm is checked before anyone reads it, rather than waiting for the next batch. - **The rubric gets written down** — Briefing the worker forces the team to state what severity one means. Most teams discover during this exercise that they did not agree. ## Questions people ask **Can the worker close duplicates automatically?** No. It proposes a duplicate link with both issue references and the reason it thinks they match. A wrong duplicate closure costs a user's report, which is more expensive than the ten seconds an engineer spends confirming. **Does it need write access to GitHub?** Not for triage as described here. It reads issues through the GitHub connection and delivers proposals as comments on Polaris tasks. Connections are authorised once for the organisation and stored server-side, where browsers cannot read them back. **How does it learn our severity rubric?** You write it into the worker's SKILL.md file, which is plain text you can open and edit. Changing a rule there changes every subsequent run, and the worker cites the rule it applied so you can see when a rule is producing bad calls. **What about reports that arrive by email or Slack?** The Polaris Inbox catches Slack messages and turns them into prefilled task suggestions with bucket, lane, labels and owner already set. You approve the suggestion with one click and the triage worker picks it up like any other task. ## Related - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/engineering/incident-postmortems - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/ai-workers/qa-engineer - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "AI-managed code review queue | Polaris use cases" description: "An AI worker posts the review queue each morning: age, who is blocking, failing checks, diffs with no test changes. Polaris does not review or merge your code." url: https://www.polarishq.co/use-cases/engineering/code-review-workflow section: Use cases updated: 2026-08-21 --- # Six open pull requests and nobody knows whose turn it is The review queue, ordered by age and named by who is blocking it, posted every morning. ## The short answer An AI worker can manage the queue around code review without touching the code. Each morning it reads open pull requests through the GitHub connection and posts an ordered queue showing how long each has been open, which reviewer has not responded, which have failing checks, and which change code without changing tests. Reviewing and merging remain entirely human actions. - **Runs:** Every weekday morning - **Connections:** GitHub, Slack, Linear - **Never automated:** The review itself ## Pull requests do not stall for interesting reasons They stall because the requested reviewer was on call, then went on holiday, and the author has moved on to something else and stopped chasing. Nobody is blocking on purpose. There is simply no place where the queue is visible, so the oldest pull request is invisible precisely because it is old. Every fact needed to fix this is in the repository, and none of it requires understanding the diff. ## What the morning queue shows | Column | Why it is there | | --- | --- | | Age in days | The oldest item is the one everyone has stopped seeing | | Waiting on | Names the specific person, not the team | | Check status | A red build means the author is blocking, not the reviewer | | Tests touched | Flags diffs that change code with no test changes at all | | Linked issue | Shows whether the pull request closes committed work | | Size | Files and lines changed, because a large diff needs a scheduled slot | > **Polaris does not review or merge code** > > A worker given the GitHub connection reads repository state and reports on it. It does not approve, comment on diffs, or merge. If you want an agent reading the diff itself, keep the coding tool you already use and give Polaris the queue, the chasing and the record. ## The rules worth putting in the skill file The queue is only useful if it reflects how your team has agreed to work. - **What counts as stale** — Two days for a small team shipping daily, a week for a team on a fortnightly cadence. The worker escalates against your number, not a default. - **Who gets named** — Some teams want the individual reviewer named, others want the team. Naming an individual is more effective and more uncomfortable, and that is a decision for the team to make deliberately. - **What to do with drafts** — Draft pull requests either belong in the queue as visible work in progress or are noise. Say which, or the queue fills with things nobody expects to review. - **Where it posts** — The delivery is always a comment on the task. Add the Slack connection if the queue should also land in the channel where people actually look in the morning. ## The queue against the review **The worker handles** - Sorting by age and naming who is blocking - Reporting failing checks and unmerged approvals - Flagging code changes with no test changes - Posting the queue where the team will see it **Engineers handle** - Whether the change is correct - Whether the approach is right - Whether the missing test matters here - Approving and merging ## Questions people ask **Does the worker comment on our pull requests?** No. It delivers the queue as a comment on a Polaris task, and optionally posts to a Slack channel if you have given it the Slack connection. The repository is read, not written to. **Is flagging untested diffs useful, or just noise?** It is a signal rather than a verdict. Plenty of legitimate changes touch no tests, including copy changes and configuration. The flag exists so a reviewer notices before merging, not so a rule can block the merge. **How is this different from a bot that posts reminders?** A reminder bot repeats a fixed message. A Polaris worker is briefed in a skill file you write, so it applies your definition of stale, your escalation rules and your position on drafts, and each run is logged with what it did. **What if reviewers ignore the queue?** Then you have a prioritisation problem rather than a visibility one, and the queue at least makes it measurable. Teams often use the age column in a retrospective, which is a conversation the data makes possible. ## Related - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/engineering/sprint-planning - https://www.polarishq.co/use-cases/product/release-notes - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/ai-workers/qa-engineer --- --- title: "AI incident postmortem timelines | Polaris use cases" description: "An AI worker builds the incident timeline from the Slack channel and deploy history, marks gaps unknown, and drafts action items. Causes stay human." url: https://www.polarishq.co/use-cases/engineering/incident-postmortems section: Use cases updated: 2026-08-21 --- # The incident ended and the writeup never started The timeline assembled from the channel and the deploy history, with the gaps left honest. ## The short answer An AI worker turns an incident into a timeline draft while the details are still recoverable. It reads the incident channel and deploy history through the Slack and GitHub connections, orders the events by timestamp, and marks every gap as unknown rather than filling it in. Action items become tasks in the workstream. Naming the contributing factors is written by the people who were there. - **Trigger:** When the incident closes - **Connections:** Slack, GitHub, Notion - **Written by a person:** Contributing factors ## The writeup competes with recovery The incident ends at 2am. The people who understand it best spend the next day catching up on everything the incident displaced, and the writeup slides to Friday, then to next week. By the time somebody opens the channel, the details have blurred and the reconstruction takes three times as long as it would have on the morning after. Assembling a timeline is not the hard part of a postmortem. It is the part that costs the most time and produces the least insight, and it is the part that decays fastest. ## How the timeline gets built 1. **Read the incident channel** — Every message in the channel across the incident window, with timestamps and authors preserved. The channel is usually the only complete record of what people believed at each moment. 2. **Overlay the deploys** — Merge and release timestamps from the repository, placed on the same timeline. The relationship between a deploy and the first alert is often the whole story and is tedious to reconstruct by hand. 3. **Mark the gaps** — Twenty minutes with no messages is reported as twenty minutes with no messages. The worker is briefed to write unknown rather than to produce a plausible sentence, because a plausible sentence in a postmortem is worse than a hole. 4. **Separate observation from explanation** — The draft holds what happened. The section on why it happened arrives empty, with the questions the timeline raises listed underneath it. 5. **Turn follow-ups into tasks** — Every action item mentioned in the channel becomes a task in the workstream, with owner and date left blank for the review meeting to fill in. ## What is reconstructed, and what is not | Reconstructed from records | Only from the people involved | | --- | --- | | When the first alert fired | Why nobody acted on it for eleven minutes | | What was deployed, and when | Why the change was believed to be safe | | Who joined the channel and when | Who thought they were leading | | What was tried, in order | What was ruled out silently and why | > **A machine-written cause is worse than no cause** > > The worker is briefed to leave the contributing factors section empty. Postmortem value comes from people saying what they believed at the time and why it was reasonable, which does not exist in any log and cannot be inferred from one. ## Why it runs immediately Assign the task when the incident closes and a cloud machine picks it up straight away, including at 2am when everyone involved has gone to bed. The draft is waiting in the morning, built from a channel nobody has had time to scroll away from yet. The document lives in Polaris Docs with versioning and review comments, so the review meeting edits one artefact instead of circulating four copies. ## Questions people ask **Can it write the whole postmortem?** It writes the timeline, the deploy overlay, the open questions and the follow-up tasks. The analysis of contributing factors is left empty on purpose, because that section is the reason the document exists and it depends on what people believed rather than what systems recorded. **What if the incident was handled in a call, not in Slack?** Then the timeline will be thin and the draft will say so, with the gaps marked explicitly. Teams that read one sparse draft tend to start narrating in the channel during the next incident, which is a useful side effect. **Does this need access to our monitoring tools?** The Polaris connection catalog covers Slack, GitHub, Notion, Google Drive, Supabase and the other listed tools, not observability platforms. In practice the alert messages posted into the incident channel carry most of the monitoring timeline anyway. **Who owns the follow-up tasks it creates?** Nobody until a person assigns them. The worker creates them in the workstream with owner and date blank, because assigning follow-up work to someone who was up all night is a decision for the review meeting. ## Related - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/engineering/bug-triage - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/github - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "AI checks for documentation that has gone stale | Polaris" description: "A monthly AI worker pass comparing docs against the current code, listing statements that are no longer true with file references. It drafts missing pages too." url: https://www.polarishq.co/use-cases/engineering/technical-documentation section: Use cases updated: 2026-08-21 --- # The README describes a version of the code that no longer exists A monthly list of statements in your docs that the code no longer supports. ## The short answer An AI worker keeps technical documentation honest by comparing it against the code on a schedule. Each month it reads the documented setup steps, configuration keys and entry points through the GitHub connection, checks them against the current repository, and posts a list of statements that are no longer true with a file reference for each. Rewriting is proposed as a draft, never applied silently. - **Runs:** Monthly - **Connections:** GitHub, Notion, Google Drive - **Output:** A list of false statements ## Documentation rots quietly and then all at once A config key gets renamed. The setup section still names the old one, and it will keep naming it until a new engineer follows the instructions, fails, and asks in Slack. That question is the only detection mechanism most teams have, and it fires once per new hire. Nobody rereads documentation against the code, because there is no moment when that becomes urgent and the work is dull in a way that resists volunteering. ## The four checks that catch most of the rot - **Named configuration keys** — Every environment variable and config key mentioned in the docs, checked for existence in the repository. This one check finds the majority of failed onboarding. - **Setup commands** — The commands in the getting-started section, compared against the scripts and manifests that actually exist. A script renamed six months ago is still in three documents. - **Referenced paths and entry points** — File paths quoted in the docs that no longer resolve. Common after any restructuring, and invisible until someone follows the link. - **Pages with no recent edits** — Documents untouched for a long time while the code they describe changed often. Not proof of rot, but the right place to look first. ## The two jobs, briefed separately Auditing and drafting need different acceptance criteria, so they are usually two tasks. | Job | What the worker delivers | What you do with it | | --- | --- | --- | | The monthly audit | A list of statements that no longer match the code, each with a file reference | Decide which are worth fixing | | The missing page | A draft written from the pull request description and the code itself | Edit it, then publish it as a versioned doc | > **Docs in Polaris are versioned and reviewable** > > The documentation tree supports nested pages, a block editor with markdown shortcuts, to-do items and sub-pages, with file versioning and review comments. A drafted page can be argued with in place, and the version history records what changed when the correction landed. ## What this does not solve It catches statements that contradict the code. It does not catch documentation that is accurate and useless, which is the more common failure: a page that describes every parameter and never says what the thing is for. That problem needs a person who understands both the system and the reader, and no schedule fixes it. ## Questions people ask **Where should the documentation live for this to work?** Anywhere the worker can read: Polaris Docs, Notion, Google Drive, or markdown files in the repository through the GitHub connection. Docs kept inside Polaris get versioning and review comments, which makes the correction cycle shorter. **Can the worker fix the documentation itself?** It delivers a proposed rewrite as a comment on the task, with the draft attached. Applying it is a human action, which matters because a statement that looks stale is sometimes documenting an intentional legacy path. **How is this different from release notes?** Release notes describe what changed for the user in a specific release. This job checks whether long-lived reference documentation still matches the code, which is a different question and runs on a different schedule. **What does a monthly audit cost?** It depends on the size of the documentation set, and the work log records the estimate line by line: pickup, searches, prose written and files produced. Billing is about $2 per human-hour delivered, and any line can be challenged from the log. ## Related - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/product/release-notes - https://www.polarishq.co/use-cases/operations/process-documentation - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/glossary/skill-file --- --- title: "AI-prepared design review agendas | Polaris use cases" description: "An AI worker builds the design review agenda the day before: files up for review, unresolved threads, what changed, and the decisions still open." url: https://www.polarishq.co/use-cases/design/design-review section: Use cases updated: 2026-08-21 --- # Twenty minutes of every design review goes on finding the file The agenda written the day before, with last review's unresolved threads at the top. ## The short answer An AI worker prepares design review by assembling the agenda before the meeting. Using the Figma, Linear and Google Calendar connections, it lists each file up for review, the task it belongs to, the comment threads still unresolved from last time, and what changed in the file since the previous session. The critique itself, and every decision that comes out of it, belongs to the people in the room. - **Runs:** The day before review - **Connections:** Figma, Linear, Google Calendar - **Never automated:** The critique ## Reviews start cold Everyone arrives without having looked. The first twenty minutes go on establishing which file is current, what was decided last time, and whether the change in front of everyone is a response to that decision or an unrelated iteration. Then the good conversation gets thirty minutes instead of fifty. None of that setup requires design judgement. It requires somebody to open four files and read the comment history the afternoon before. ## What the agenda contains, in order | Item | Why it is in that position | | --- | --- | | Unresolved from last review | Decisions that were deferred are the ones that get deferred again | | Files up this session | Each with its linked task and the designer presenting | | Changed since last time | So the room knows whether it is looking at an iteration or a new direction | | Open comment threads | Grouped by file, with the question each thread is stuck on | | Decisions needed | The explicit list, so the meeting can be judged on whether it made them | ## The details that make the agenda usable - **Threads that were never answered** — A comment nobody replied to for three weeks is a decision that quietly did not happen. Surfacing those first is most of the value of the agenda. - **Named presenters** — Pulled from the linked task owner, so nobody discovers in the meeting that the person who was going to walk through the work is not there. - **Calendar reality** — Google Calendar tells the worker who is actually in the session, which changes what is worth putting on the list. > **The worker has no opinion about the design** > > It reports what is up for review and what is unresolved. It does not summarise whether a direction is working, and it does not resolve threads. A machine assessment of a design would be confident, fast and worth nothing, so the brief excludes it. ## After the review The decisions get written into the task as comments by the people who made them. Assigning a second task to the worker to reconcile the agenda against what was decided is possible, but most teams find that the writing-down is the part that has to stay human, because a decision nobody typed is a decision nobody made. ## Questions people ask **Can the worker see comments inside Figma files?** It reads file state and comment threads through the Figma connection, which is authorised once for the organisation and stored server-side. What it reports back is a list with links, delivered as a comment on the Polaris task. **What if our design work lives in another tool?** The connection catalog is fixed: Figma is the design tool in it. Teams using something else usually keep the review task and its linked files in Polaris and give the worker Google Drive, which covers exported files and written specs. **Does this replace a design ops person?** No. It removes one recurring chore from whoever currently does it, which in most teams under fifty people is a designer doing it at 6pm the night before. Facilitating the review and holding the standard is a different job. **How far ahead should the agenda run?** A day is usually right. Any earlier and files change after the agenda is written; any later and nobody reads it before the meeting. The task can be scheduled so a cloud machine builds it overnight. ## Related - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/use-cases/design/asset-handoff - https://www.polarishq.co/use-cases/design/design-system-maintenance - https://www.polarishq.co/integrations/figma - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/glossary/workstream --- --- title: "AI design system drift audits, weekly | Polaris" description: "A weekly AI worker pass over the component library and its documentation: undocumented components, stale pages, and design tokens that no longer exist." url: https://www.polarishq.co/use-cases/design/design-system-maintenance section: Use cases updated: 2026-08-21 --- # The component was renamed and the documentation was not A weekly drift report between the component library and the documentation that describes it. ## The short answer Design system maintenance with an AI worker is a weekly drift report. The worker compares the component library against its documentation through the Figma, Notion and GitHub connections, then lists components with no documentation page, documentation describing components that were renamed or removed, and design tokens referenced in docs that no longer exist in the library. Deprecation decisions stay with the design system owner. - **Runs:** Weekly - **Connections:** Figma, Notion, GitHub - **Output:** Three drift lists ## Design systems fail at the documentation, not the components The components are fine. Somebody renamed a variant, somebody else added a spacing token, a third person deprecated a card and left it in the library so nothing would break. Each move was reasonable. Together they mean the documentation now describes a system that partially exists, and engineers have started guessing. The moment a designer cannot trust the docs, they open an old file and copy from it, and the system starts forking in a way that takes a quarter to notice. ## The three lists | List | What it means | Usual cause | | --- | --- | --- | | Undocumented components | In the library, described nowhere | Added in a hurry during a project | | Stale documentation | Describes something renamed or removed | A rename that touched the library and not the docs | | Dangling references | Docs cite tokens or variants that do not exist | Token cleanup done without a documentation pass | ## Two things worth adding to the brief - **Check the code side too** — If the implemented components live in a repository, the GitHub connection lets the worker report components that exist in Figma and not in code, or the reverse. That gap is usually where engineers start improvising. - **Report the delta, not the inventory** — A weekly report that lists every component is a report nobody reads. Brief the worker to report what changed since last week, with a full inventory only when you ask for one. > **Deprecation is a decision, not a finding** > > The worker can tell you a component has no documentation and has not been used in six months. Whether it should be removed depends on what still ships with it, and that answer is not in the file. The report ends where the judgement begins. ## Who does what **The worker** - Reads the library and the docs weekly - Names every mismatch with a link to both sides - Reports the change since last week - Drafts the missing documentation page when asked **The system owner** - Decides what gets deprecated - Sets the naming convention the audit checks against - Approves and publishes the drafted pages - Decides when drift is worth a migration ## Questions people ask **Does the worker edit our Figma library?** No. It reads the library through the Figma connection and delivers findings as a comment on a Polaris task. Files stay authored by designers, which is also why the report links to each component rather than proposing an edit. **Can it write the missing documentation pages?** It can draft them from the component structure and any existing usage, delivered as a file on the task. Publishing is yours, and pages published into Polaris Docs are versioned with review comments so the draft can be argued with in place. **How does it know what our naming convention is?** From the worker's SKILL.md file, which you write and can edit at any time. A convention stated there, such as every component page carries a Usage and a Do not use section, turns a vague audit into a checkable one. **Is weekly too often?** For a stable system, monthly is enough and the report is shorter. During a migration, weekly catches divergence while it is still one component rather than a pattern. Billing follows delivered work, so a quiet week is a cheap week. ## Related - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/use-cases/design/brand-consistency-audits - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/integrations/figma - https://www.polarishq.co/integrations/github - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/glossary/skill-file --- --- title: "AI usability testing synthesis by task | Polaris" description: "An AI worker turns usability session notes into a table of task, completion, hesitation point and verbatim quote, with severity scored against your rubric." url: https://www.polarishq.co/use-cases/design/user-testing-synthesis section: Use cases updated: 2026-08-21 --- # Eight sessions recorded, and the same hesitation in six of them Session notes turned into a task-by-task table of where people stalled and what they said. ## The short answer Usability testing synthesis with an AI worker produces a table rather than a narrative. The worker reads session notes from Google Drive and returns one row per task attempted: how many participants completed it, where they hesitated, what they said at that moment, and a severity score against the rubric in its skill file. Deciding what to change in response stays with the design team. - **Input:** Session notes and transcripts - **Connections:** Google Drive, Notion, Linear - **Output:** A task-by-task table ## The findings that survive are the ones somebody remembered After eight sessions, the team remembers the participant who got angry and the one who could not find the settings. Those two moments shape the redesign. The quieter pattern, where five people hesitated for three seconds at the same step and then recovered, does not get discussed at all, because recovering looks like success in a note. A task-level table catches that pattern, because it counts hesitations rather than failures. ## The shape of the delivered table One row per task, one column per thing you will argue about. | Column | What goes in it | | --- | --- | | Task | The task as the participant was asked to perform it | | Completed | Count of participants who finished unaided | | Assisted | Count who finished after a prompt from the moderator | | Hesitation point | The specific step where the pause happened | | Verbatim | What the participant said at that moment, in their words | | Severity | Scored against your rubric, with the rule cited | ## How this differs from research synthesis Both read transcripts. They answer different questions and are briefed differently. - **Usability synthesis is task-shaped** — The unit is a task attempt and the outcome is a completion rate with an observed failure point. It tells you where the interface is wrong. - **Discovery synthesis is theme-shaped** — The unit is a topic raised across participants, with quotes and counts. It tells you what problem people have, which is a different question. - **Mixing them buries the interface finding** — A hesitation at step four is specific and actionable. Folded into a theme about confidence in the product, it stops being either. > **Small samples stay small** > > Six of eight participants is six of eight, and the delivered table says so in every row. Brief the worker to report counts rather than percentages on samples under about twenty, because a percentage on eight people invites a confidence nobody has earned. ## Turning rows into work Each severity-one row can become a task in the design workstream with the verbatim quote in the description, which is the single most effective way to keep the finding alive through three weeks of implementation. The Focus lane holds the ones being fixed this week across three horizons, so a usability finding does not slide behind whatever arrived on Monday. ## Questions people ask **Does the worker watch session recordings?** No. It reads notes and transcripts you have already stored in Google Drive. Moderating sessions and taking observational notes stay with the researcher, and the quality of those notes sets the ceiling for the synthesis. **How does it score severity?** Against the rubric written in its SKILL.md file, and it cites which rule it applied for each row. A rubric like blocked without help is severity one, hesitated and recovered is severity three makes the scoring checkable rather than a matter of tone. **Can it compare two rounds of testing?** Yes, if both rounds are in folders it can read and the tasks are named consistently. The comparison is usually delivered as a second table showing completion counts side by side, which is the fastest way to see whether a fix worked. **What if the sessions were unmoderated?** Unmoderated sessions produce fewer verbatims and no assisted-completion column, so the table will be sparser. The worker reports the columns it could fill and names the ones it could not rather than leaving the gap ambiguous. ## Related - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/use-cases/product/user-research-synthesis - https://www.polarishq.co/use-cases/design/design-review - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "AI handoff checks before design goes to build | Polaris" description: "An AI worker runs the handoff completeness check: frames marked ready, exports that exist, what the ticket references but nobody produced, plus spec values." url: https://www.polarishq.co/use-cases/design/asset-handoff section: Use cases updated: 2026-08-21 --- # Is this final? Asked for the fourth time this month A completeness check before build starts, so the question is answered before it is asked. ## The short answer An AI worker runs the completeness check that sits between design and build. Using the Figma, Google Drive and Linear connections, it lists which frames are marked ready by your team's convention, which exported assets are present in the shared folder, what the ticket references but nobody produced, and the spacing and colour values pulled into a written spec. An engineer still asks whatever the check did not cover. - **Trigger:** When handoff is assigned - **Connections:** Figma, Google Drive, Linear - **Output:** Ready or not-ready, itemised ## Handoff fails on the small things The design is finished. The icon set was never exported at two times, the empty state exists only in a comment, and the hover colour is in the file but not in the ticket. Each of these is a two-minute fix and each one costs a half-day round trip, because the engineer discovers it at the point of needing it and the designer is in a workshop. The check that would catch all three is mechanical. Nobody runs it because running it is boring and the failure is somebody else's afternoon. ## The handoff check, in order 1. **Confirm what is marked ready** — By your team's convention, whatever it is: a page named Ready for dev, a status field, a frame prefix. Write the convention into the SKILL.md so the check is exact rather than interpretive. 2. **Match frames against the ticket** — Every state the ticket references, checked for a corresponding frame. Empty states, error states and loading states are the three that go missing, in that order. 3. **Check the exports** — Assets referenced by the design, checked against what is actually in the shared Google Drive folder, at the sizes the ticket asks for. 4. **Write the spec values** — Spacing, type sizes and colour values pulled from the file into written text in the comment, so an engineer can read them without opening the design tool. 5. **Deliver a verdict with a list** — Ready, or not ready with the specific missing items named. A verdict without the list is a status; the list is what makes it fixable in five minutes. ## The three states that go missing Worth checking explicitly, because they are complete in the designer's head and absent from the file. | State | How it usually goes missing | What the check asks | | --- | --- | --- | | Empty | Designed once in a different file, never linked | Is there a frame for zero items? | | Error | Discussed in a comment thread, never drawn | Is there a frame for the failure case? | | Loading | Assumed to be the standard spinner | Does the ticket say which loading treatment? | > **The check does not judge the design** > > It reports presence and absence against the ticket and your ready convention. Whether the empty state is any good is a design review question, and it belongs in the review rather than in a handoff checklist. ## Questions people ask **Can the worker export assets itself?** No. It reports which referenced assets are missing from the shared folder, and exporting stays with the designer. Delivery in Polaris is a comment on the task, which can carry generated documents such as the written spec. **What does it do about responsive breakpoints?** It checks for whatever your ticket template asks for. If the template names three breakpoints and the file has one, that appears in the not-ready list. If the template says nothing, neither will the check, which is a template problem worth fixing once. **Who runs the check, the designer or the engineer?** Usually the designer assigns it before marking the ticket ready, which is the point where the fix is cheapest. Some teams assign it on the engineering side as a gate, which works but produces more back-and-forth. **Does this need a Figma seat for the worker?** It uses the Figma connection from the Polaris catalog, authorised once for the organisation. Credentials are verified at connect time and stored server-side, so workers use them and browsers cannot read them back. ## Related - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/use-cases/design/design-review - https://www.polarishq.co/use-cases/engineering/code-review-workflow - https://www.polarishq.co/integrations/figma - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/glossary/delivery-comment --- --- title: "AI brand consistency audits across public surfaces | Polaris" description: "A quarterly AI worker pass over public pages and the social grid against your brand document, listing each departure with a link and the rule it breaks." url: https://www.polarishq.co/use-cases/design/brand-consistency-audits section: Use cases updated: 2026-08-21 --- # Four versions of the logo are live and nobody signed off on three A quarterly pass over your public surfaces, checked against the rules your brand doc states. ## The short answer A brand consistency audit by an AI worker compares your public surfaces against your brand document. Using open web search and the Instagram connection, the worker reads the pages and posts you name, checks logo usage, typography and colour against the rules the document states, and lists every departure with a link to the instance and the rule it breaks. Vague rules produce vague findings, which the report says plainly. - **Runs:** Quarterly - **Connections:** Web search, Instagram, Google Drive - **Output:** Departures, each with a link ## Brand drift happens where nobody is looking The website is fine because the design team touches it. The drift is on the careers page a recruiter built, the partner microsite from a campaign two years ago, the deck template circulating in sales, and the twelve most recent posts on the social grid where the type sizes have crept. Each of those was made by somebody doing their job with the assets they could find. The result is four logo lockups in the wild, and the only way to know is to look at all of it at once. ## What gets checked, and against what | Surface | Connection | What the rule usually is | | --- | --- | --- | | Public web pages you name | Web search | Logo lockup, clear space, approved type | | The Instagram grid | Instagram | Colour palette, type treatment, template use | | Shared decks and templates | Google Drive | Cover slide, type scale, approved colours | | The brand document itself | Google Drive or Notion | The source of every rule the audit applies | > **The audit inherits the quality of your brand document** > > A rule saying use the logo with adequate breathing room produces a finding saying breathing room may be inadequate, which helps nobody. A rule saying clear space equals the height of the mark on all sides produces a finding with a measurement. Writing the specific rule is the part that has to happen first. ## How to make the report actionable - **One row per instance, not per problem** — Three pages using the old logo is three rows with three links. A summary saying the old logo is still in use gets read and not fixed. - **Cite the rule** — Each finding names the section of the brand document it comes from, which stops the audit becoming an argument about taste. - **Separate what is behind a login** — The worker only sees public surfaces. Anything it could not reach is listed explicitly, so nobody assumes the product interface was covered. - **Turn severity-one rows into tasks** — A wrong logo on a live pricing page is a task with an owner and a date. The rest can sit in the report until the next quarterly pass. ## The boundary **Checkable** - Which logo file is in use on which page - Whether the colours match the stated palette - Whether the typeface is one of the approved ones - Whether templates were used or bypassed **Not checkable** - Whether the photography feels like the brand - Whether the tone of voice is right - Whether an exception was a good exception - Whether the brand itself still fits the company ## Questions people ask **Can it audit pages behind a login?** No. The worker uses the open web search connection for public pages and the Instagram connection for the social grid. Anything requiring authentication is listed as unreachable rather than skipped quietly, so the coverage of the audit is always explicit. **How does it know our palette?** From the brand document you point it at, through the Google Drive or Notion connection. The rules in that document are the entire basis of the audit, and the worker cites the section it applied for each finding. **Is quarterly the right cadence?** For most teams, yes, because fixing findings takes longer than finding them and a monthly report would repeat itself. Teams mid-rebrand often run it monthly for two quarters and then drop back. **What about the product interface?** That is better served by the design system drift audit, which compares the component library against its documentation and against the implemented code. Brand audits cover public marketing surfaces, where the drift is caused by people outside the design team. ## Related - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/use-cases/design/design-system-maintenance - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/integrations/instagram - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/ai-workers/social-media-manager - https://www.polarishq.co/for/design-studios --- --- title: "AI intake for dashboard and data requests | Polaris" description: "An AI worker scopes every incoming data request before it reaches the queue: the decision it informs, the tables it needs, and the two questions nobody asked." url: https://www.polarishq.co/use-cases/data/dashboard-requests section: Use cases updated: 2026-08-21 --- # Can you pull the numbers, sent as a direct message at 6pm Every ask restated as a question, with the clarifying questions asked before anyone writes SQL. ## The short answer An AI worker handles the intake step for data requests. When a request arrives in Slack, the Polaris Inbox turns it into a prefilled task suggestion, and the worker restates the ask as an answerable question, names the tables it would need through the Supabase connection, and posts the clarifying questions that were skipped. The analyst decides whether the request is worth doing at all. - **Trigger:** Every incoming request - **Connections:** Slack, Supabase, Notion - **Output:** A scoped, answerable question ## The request arrives without any of the information needed to answer it Can you pull last month by plan. Which last month, calendar or trailing thirty days. Which plan field, the one on the subscription or the one on the account. Does churned count. Is this for the board deck on Thursday, in which case the definition has to match the last three decks, or is it curiosity. The analyst asks two of those questions, gets an answer to one, and builds something that turns out to be the wrong shape. Then it gets rebuilt. The rebuild is the actual cost of bad intake. ## What the worker adds before the request is queued - **The decision it informs** — If nobody can name a decision that changes based on the answer, that is worth knowing before an afternoon is spent. The worker asks, and records the answer or its absence. - **The restated question** — The ask rewritten so it has exactly one interpretation, with the date range, the grain and the filters stated explicitly. - **The tables it would touch** — Read from the schema through the Supabase connection, so the analyst knows whether this is a ten-minute query or a joins problem before committing. - **The existing answer, if there is one** — Half of all requests have been answered before. The worker checks the documented metrics and previous requests, and links the answer instead of queueing a repeat. ## From Slack message to scoped task 1. **The message arrives** — Someone asks in a channel or a direct message. The Polaris Inbox catches it and turns it into a task suggestion with the bucket, lane, labels and owner already filled in. 2. **One click accepts it** — Suggestions never become tasks silently. Accepting one creates a fully specified task in the data workstream, which is the point where it becomes real work. 3. **The worker scopes it** — A cloud machine picks the task up, reads the schema, checks whether the question has been answered before, and posts the restated question with its open clarifications. 4. **The requester answers, or does not** — Requests whose clarifications go unanswered for a week are usually requests nobody needed, and now that is visible instead of being an analyst's private suspicion. 5. **The analyst decides** — Do it now, schedule it, or reply with the existing answer. The scoping note is attached, so the decision takes a minute rather than a meeting. > **Nothing runs against production without a person** > > The worker reads schema and asks questions. Approving anything that writes, or anything expensive enough to matter, stays with the data team. The tasks on this page are briefed as read-and-report work for exactly that reason. ## Questions people ask **Does this make it harder for people to ask for data?** Slightly, and that is the point. A request that survives being restated as a specific question with a named decision is a request worth doing. Teams typically find a meaningful share of asks stop at the clarifying question stage. **What if the request comes by email instead?** Give the worker the Gmail connection and the same intake applies. The connection catalog covers Slack, Gmail, WhatsApp and the rest of the fixed list, so intake can run wherever your requests actually arrive. **Can the worker answer simple requests itself?** It can look up documented metrics and previous answers and reply with those. Producing a new number from the warehouse is a different job, briefed separately, and most teams keep query execution behind human approval. **Where does the scoping note live?** As a comment on the task, which stays with the request permanently. That record is what lets you answer next quarter's identical request with a link rather than a rebuild. ## Related - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/use-cases/data/analysis-backlog - https://www.polarishq.co/use-cases/data/metric-definitions - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/glossary/workstream --- --- title: "AI daily data quality checks | Polaris use cases" description: "An AI worker runs your data quality checks every morning against Supabase and posts only when a threshold trips. A clean run reports nothing." url: https://www.polarishq.co/use-cases/data/data-quality-monitoring section: Use cases updated: 2026-08-21 --- # The null rate tripled in March and the board slide was already wrong Row counts, null rates, orphans and freshness, checked every morning. Silence means clean. ## The short answer Daily data quality monitoring by an AI worker means the checks actually run. Each morning the worker executes the checks written in its skill file against the Supabase connection: row counts against yesterday, null rates on critical columns, orphaned foreign keys, and the age of the newest row in each table. It posts a delivery comment only when a threshold trips. A clean run reports nothing. - **Runs:** Every morning - **Connections:** Supabase, Slack - **Clean run output:** Nothing ## Data quality problems are found by accident Somebody notices a number looks small. They check, and the ingestion for one source has been failing since a schema change three weeks ago. Every dashboard built on that table has been quietly wrong, including the one used in the board meeting, and nobody can say for how long without going back through the data. The check that would have caught it on day one takes about four lines to describe. It does not exist because writing monitoring is never this week's priority, and because a check that nobody has to maintain is not a thing that exists. ## The four checks worth having first Each one catches a different failure mode, and all four fit in a skill file. | Check | What it catches | What trips it | | --- | --- | --- | | Row count against yesterday | Ingestion that stopped or doubled | A change outside the band you set | | Null rate per critical column | A field that silently stopped populating | Null share above your threshold | | Orphaned foreign keys | Rows referencing records that do not exist | Any count above zero, usually | | Freshness of the newest row | A pipeline that runs but writes nothing | Newest row older than the expected interval | ## Setting it up This is the job most data teams put on the roster first. 1. **List the tables that matter** — Not every table. The ten that dashboards and billing depend on, because a monitor that alerts on unimportant tables gets muted within a fortnight. 2. **Write the thresholds down** — In the worker's SKILL.md: the acceptable day-over-day band per table, the null ceiling per column, the expected freshness interval. These are yours and you edit them directly. 3. **Authorise Supabase** — One click from the connections catalog. Credentials are verified at connect time and stored server-side, so the worker uses them and browsers cannot read them back. 4. **State that silence is success** — Write it into the acceptance criteria: a clean run reports no exceptions and closes. Otherwise you get a daily comment saying everything is fine, which trains everyone to stop reading. 5. **Escalate through Slack** — Add the Slack connection so a tripped threshold reaches the channel, while the full detail stays as the delivery comment on the task. > **A quiet day is close to free** > > Hours are estimated from observable effort: pickup time, searches performed, prose written, files produced, clamped to a five-minute floor per session. A clean run produces no prose and no files, so it lands at the floor, and the work log shows exactly that. ## What this is not - **Not a replacement for tests in the pipeline** — Checks that run inside your transformation layer catch problems before the data is written. This catches what gets through, which is a different and complementary position. - **Not anomaly detection** — It compares against thresholds you wrote, not against a learned baseline. That makes it predictable and explainable, and it means it will miss things you did not think to check. ## Questions people ask **Does the worker need write access to the database?** No. All four checks are reads. Teams usually give the worker a read path through the Supabase connection and keep anything that writes behind human approval, which is how every data job on this site is briefed. **What happens when a check trips at 6am?** A cloud machine runs the job on schedule regardless of whether anyone is online, and the delivery comment lands on the task. With the Slack connection authorised, the exception also reaches the channel where the on-call analyst will see it. **How do I stop it alerting on known problems?** Write the exception into the skill file with the reason and, ideally, a date to revisit. Keeping known-issue exclusions in a file you can read is better than a mental list, because the file survives the person who made the exception. **Can it check tables outside Supabase?** The connection catalog is fixed, and Supabase is the database connection in it. Stripe is available separately, which covers a common source of billing-side reconciliation checks. ## Related - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/use-cases/data/reporting-automation - https://www.polarishq.co/use-cases/engineering/incident-postmortems - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/glossary/work-log --- --- title: "AI recurring reports, written not just generated | Polaris" description: "An AI worker assembles the recurring report, writes the paragraph saying what moved and by how much, and delivers it as a file rather than a link." url: https://www.polarishq.co/use-cases/data/reporting-automation section: Use cases updated: 2026-08-21 --- # The Monday metrics email that somebody writes on Sunday night The numbers pulled and the paragraph written, delivered as a file rather than a link. ## The short answer Reporting automation with an AI worker covers the writing, not only the pulling. The worker gathers the figures through the Supabase and Stripe connections, writes the paragraph stating what moved and by how much against the previous period, attaches the report as a file, and posts it as a delivery comment on the task. It states movements. Claiming what caused them is a human addition. - **Runs:** Weekly or monthly - **Connections:** Supabase, Stripe, Google Drive - **Output:** A written report file ## A dashboard link is not a report Everyone has a dashboard. Almost nobody opens it, because opening it requires knowing which of the eleven numbers changed this week, and the dashboard is equally happy to show you all eleven whether or not any of them moved. What people read is the paragraph. Two hundred words saying signups are up nine percent week over week, driven by one channel, and the conversion rate held. Writing that paragraph is the part that takes forty minutes and the part that gets skipped when the week is busy. ## What separates a written report from a generated one - **It says what moved, and by how much** — Both the absolute and the relative change, against the same period last time and against the same period last year where the data goes back that far. - **It says what did not move** — A metric that has been flat for six weeks is a finding. Reports that only mention movement make stability invisible until it becomes a problem. - **It stops before the causation** — The worker reports that a number moved. Saying why it moved requires knowing about the campaign that launched on Tuesday, and unless it was briefed to look, it should not guess. - **It arrives as a file** — Deliveries in Polaris are comments that can carry generated documents, so the report is an artefact with a date on it rather than a link whose contents change under you. ## Where the numbers come from | Connection | Typical contribution | | --- | --- | | Supabase | Product usage, signups, activation, retention cohorts | | Stripe | Revenue, refunds, failed payments, subscription movements | | Google Drive | Where the finished report file is filed for the record | | Gmail | Sending the report on, when it goes to people outside the workspace | | Slack | Posting the summary where the team reads on Monday morning | > **The definition problem does not go away** > > An automated report built on three incompatible definitions of active user produces three wrong numbers reliably, every week, forever. Reconciling the definitions is a separate job and it should happen before the report is scheduled rather than after somebody notices. ## Why the recurring shape is cheap The same report every week is the ideal shape for metered work. The searching is minimal because the sources are known, the prose is short, and the estimate lands near the bottom of the range. Every run is itemised on the worker's work log with the effort that produced it. If a week's report costs more than usual, the log shows which part took the work, which is a more useful conversation than a line item on an invoice. ## Questions people ask **Can it email the report to people outside our workspace?** With the Gmail connection authorised, yes. Most teams keep the delivery inside Polaris as a comment with the file attached, and send it on themselves, because that keeps a human between the numbers and the recipient. **Will it explain why a number changed?** Only if you brief it to look, and only from sources it can reach. A worker told to check the campaign calendar and the deploy history before writing can offer candidate explanations, and it should mark them as candidates rather than conclusions. **What if the numbers are wrong?** Reject the delivery and say why in a reply. Nothing was published, because the report arrives as a comment on an open task, and the task stays open until a person closes and rates it. **How is this different from a scheduled dashboard export?** An export ships the same table every week regardless of whether anything happened. This delivers prose that names what changed, what held and what is missing, which is what people actually read on a Monday morning. ## Related - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/use-cases/data/data-quality-monitoring - https://www.polarishq.co/use-cases/data/metric-definitions - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files --- --- title: "Conflicting Metric Definitions, Reconciled by AI | Polaris" description: "An AI worker finds every place a metric is defined across dashboards, docs and Slack threads, and lists the definitions side by side with sources." url: https://www.polarishq.co/use-cases/data/metric-definitions section: Use cases updated: 2026-08-21 --- # Three teams, three definitions of active user, one meeting Every definition of a metric found and listed side by side, with the source of each. ## The short answer An AI worker can find every definition of a metric that exists in a company. It reads dashboards, documents and past discussions through the Supabase, Notion and Slack connections, lists each distinct definition side by side with its source, and drafts a single definition page stating the rule in words. Which definition becomes canonical is ratified by people, and the resulting document is versioned. - **Runs:** Once, then on demand - **Connections:** Supabase, Notion, Slack - **Output:** Definitions, side by side ## Nobody decided to have three definitions Marketing counted anyone who opened the app. Product counted anyone who completed a core action. Finance counted anyone on a paying account, because that was the number in the model. Each definition was reasonable in the room where it was made and none of the three rooms knew about the others. The discovery usually happens in a meeting, with a slide up, when two numbers that should match do not. Then forty minutes go on archaeology instead of on the decision the meeting was called for. ## Where definitions hide - **In query logic** — The real definition is the WHERE clause, and it usually differs subtly between two dashboards that carry the same title. - **In documents** — Written definitions in Notion or Google Drive, often accurate on the day they were written and never revised after the logic changed. - **In Slack threads** — The most common location for the actual working definition, agreed in a thread by three people and never written anywhere permanent. - **In the deck** — A number in a board deck with a footnote nobody has re-read, which is frequently the definition the company is actually managed by. ## What the delivered comparison looks like | Column | Contents | | --- | --- | | Definition | The rule, stated in words rather than as SQL | | Source | Where it was found, with a link | | Last touched | When that source was last edited | | Population | Who it counts and who it excludes | | Used by | Which dashboards, docs or decks depend on it | > **The worker does not pick the winner** > > Choosing which definition becomes canonical changes every dashboard, every historical comparison and possibly a number that has already been shown to investors. That is a decision with consequences, so a person makes it and the document recording it carries a date and a version history. ## After the decision The ratified definition goes into a versioned document in Polaris Docs, with review comments, so a future change to it is visible rather than silent. Each dependent dashboard that needs updating becomes a task in the data workstream. Six months later, when somebody proposes changing the definition again, the version history shows what it was, when it changed and who agreed, which is the argument that the meeting would otherwise have to reconstruct from memory. ## Questions people ask **How does the worker find definitions in Slack?** Through the Slack connection, searching the channels you point it at for discussion of the metric by name. Threads where a definition was agreed are listed with links, so the archaeology is checkable rather than something you take on trust. **Can it rewrite our dashboards to match the ratified definition?** No. It lists the dependent dashboards it found, and each one becomes a task for a person to update. Changing query logic on live dashboards is exactly the kind of write action that stays with the data team. **How many metrics should we do at once?** Start with the three or four that appear in board reporting, because those are where a mismatch is most expensive. Doing twenty at once produces a document nobody ratifies, and an unratified definition page is no better than the confusion it replaced. **What if two definitions are both legitimate?** Then the answer is two named metrics rather than one contested one, and the definition page says which is used where. The worker will surface the conflict; naming both is the human decision that ends the argument. ## Related - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/use-cases/data/reporting-automation - https://www.polarishq.co/use-cases/data/dashboard-requests - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "AI scoping notes for the analysis backlog | Polaris" description: "Every analysis request gets a written scoping note first: the decision it informs, the data required, whether it exists, and the cheapest version of the answer." url: https://www.polarishq.co/use-cases/data/analysis-backlog section: Use cases updated: 2026-08-21 --- # Thirty requests in the backlog and one analyst A scoping note per request, so the backlog is triaged on value rather than on arrival order. ## The short answer An AI worker writes the scoping note that turns an analysis request into a decision about whether to do it. For each item in the backlog it records the decision the analysis informs, the data required, whether that data exists in the warehouse, and the cheapest version that would still answer the question. The analyst then triages on value instead of on arrival order. - **Runs:** Per request, then as a batch - **Connections:** Supabase, Notion, Linear - **Output:** A scoping note per item ## Backlogs are ordered by who asked most recently Not by value. The item at the top is there because somebody mentioned it in standup, and the item at the bottom has been there since February and might be the one that matters. Nobody reorders the list because reordering requires knowing what each item is worth, and that information was never captured. So the analyst works down from the top, the February item stays at the bottom, and every month somebody says the backlog is out of control without anyone being able to say which parts of it should simply be deleted. ## The four fields in a scoping note | Field | The question it answers | Why it kills requests | | --- | --- | --- | | Decision | What changes based on the answer? | A request with no decision behind it does not survive being written down | | Data required | Which tables and fields does this need? | Reveals when the answer requires data nobody collects | | Data exists | Do we actually have it, at the grain needed? | Half of ambitious requests fail here | | Cheapest version | What would a good-enough answer look like? | Turns a two-week project into a two-hour one, often | ## The cheapest-version field earns its place - **It reframes the request** — A cohort model built over two weeks and a count over two hours frequently point at the same decision. Naming both lets the requester choose, which they were never offered before. - **It exposes the requests that need the model** — Some genuinely do, and the scoping note makes that case explicitly instead of the analyst having to defend the time. - **It gives the analyst a starting point** — Even when the full version is approved, the cheap version usually runs first and sometimes ends the question. > **Focus holds the two that matter** > > Polaris pins a Focus lane first in every view, across three horizons: today, this week, and the next thirty days. A triaged backlog is only useful if the two items that matter this week stay visible, and the Focus lane is where they sit rather than at position seven of thirty. ## Scoping against analysing **The worker scopes** - Reads the request and restates the question - Checks the schema for the data required - Proposes the cheapest sufficient version - Links previous work on the same question **The analyst analyses** - Decides what is worth doing - Writes and runs the query - Judges whether the result is trustworthy - Delivers the answer and defends it ## Questions people ask **Does every request need a scoping note?** Anything larger than a quick lookup benefits, and the note takes the worker minutes rather than an analyst's afternoon. Requests answerable in five minutes are usually answered rather than scoped, which the intake step already sorts out. **Who decides what gets deleted from the backlog?** A person. The scoping note makes deletion defensible by recording that nobody could name a decision the analysis would inform, which is a much easier conversation than declining a request on instinct. **Can the worker estimate how long an analysis will take?** It can describe the data required and whether it exists, which is the part that actually determines effort. A time estimate from a machine that has not written the query would be a number with nothing behind it. **How does this interact with request intake?** Intake turns a message into a specified task and asks the clarifying questions. Scoping happens after that, on requests that survived, and produces the note the analyst triages against. Running both means the backlog contains only requests somebody could defend. ## Related - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/use-cases/data/dashboard-requests - https://www.polarishq.co/use-cases/data/metric-definitions - https://www.polarishq.co/use-cases/product/feature-prioritization - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/glossary/focus-lane --- --- title: "Polaris for Marketing Teams | AI Workers on One Board" description: "How a marketing team runs calendars, briefs, campaigns and competitor tracking in Polaris, with AI workers holding tasks next to the people who hired them." url: https://www.polarishq.co/use-cases/marketing section: Use cases updated: 2026-08-21 --- # Marketing work, with AI teammates on the same board One board for the calendar, the briefs and the campaigns, with workers who draft while nobody is at a desk. ## The short answer Polaris gives a marketing team one board where people and AI workers hold the same kind of task. A worker is hired in chat in about sixty seconds, given a skill file describing how your content should read, and connected to Slack, Notion, Google Drive or web search. It drafts calendars, briefs, competitor updates and campaign plans, then posts each one as a comment a human approves. - **Hiring time:** ~60 seconds, in chat - **Typical connections:** Notion · Google Drive · Slack · web search - **Software cost:** $0, unlimited people - **Work cost:** ~$2 per human-hour delivered ## The marketing week, honestly Most marketing teams do not lose time on the interesting decisions. They lose it on the twelve small productions that sit between decisions: the calendar nobody updated, the brief that was never written down, the competitor page nobody checked since March, the three posts that need drafting before Thursday. Those productions are the ones an AI worker can hold. Not the positioning call, not the brand judgement. The drafting, the checking, the assembling, the chasing. In Polaris a worker is a member of the team in the literal sense: humans and agents live in the same members table, with the same assignment flow. You drag a task to Mara the way you drag one to Dan. The difference is that a cloud machine wakes for Mara's task and keeps going after you close the laptop. ## What you brief a marketing worker on Hiring runs as a short interview in chat. Every answer is a click, and what you pick becomes an editable SKILL.md file you can read later. - **House style** — Sentence length, banned words, whether you use headings or run long. The skill file is where your style guide finally lives somewhere a machine reads. - **Sources it may use** — Your own docs in Polaris, files in Google Drive, a Notion space, or the open web. Connections are authorised once for the whole org and stored server-side. - **What finished looks like** — Acceptance criteria the worker ticks as it goes: word count band, sources linked, a named CTA, three headline options rather than one. - **Where the work lands** — A workstream and a lane. A blog draft goes to Content · In review. A competitor update goes to Research · This week. ## Five recurring marketing jobs, and what each one needs Each of these has its own page below with the actual brief. | Job | Connections | What the worker delivers | What the human decides | | --- | --- | --- | --- | | Content calendar | Google Calendar · Notion · Slack | A rolling calendar plus a task per slot, owned and dated | What is worth publishing at all | | SEO content production | web search · Google Drive | A drafted piece with sources linked, as a file on the task | The angle, the edit, whether it ships | | Campaign planning | Slack · Google Drive · HubSpot | A plan doc plus one task per channel with dates | Budget, offer, and the launch date | | Competitor monitoring | web search | A weekly diff of what changed on named competitor pages | Whether a change deserves a response | | Social media scheduling | Instagram · Slack | A week of drafted posts, per channel, in one comment | Which posts go out and when | ## The split that works Teams that get value here draw the line in the same place. **Humans keep** - Positioning and the argument the brand is making - The offer, the price and the launch date - Anything a customer will read with your name on it, before it goes out - Which competitor move is worth reacting to **The worker takes** - First drafts, at any hour, with sources attached - The weekly check of pages you would otherwise forget - Turning a plan into tasks with owners and dates - Assembling the same report every Monday morning > **You are billed for delivered work, not for marketers** > > Polaris charges nothing for the software and about $2 for each human-equivalent hour a worker delivers. The hours come from an open formula based on observable effort, and every line is itemised on the worker's work log, where you can challenge it. A marketing team of twenty pays the same $0 for the board as a team of two. ## The five marketing jobs in detail - [use-cases/marketing/content-calendar](https://www.polarishq.co/use-cases/marketing/content-calendar) - [use-cases/marketing/seo-content-production](https://www.polarishq.co/use-cases/marketing/seo-content-production) - [use-cases/marketing/campaign-planning](https://www.polarishq.co/use-cases/marketing/campaign-planning) - [use-cases/marketing/competitor-monitoring](https://www.polarishq.co/use-cases/marketing/competitor-monitoring) - [use-cases/marketing/social-media-scheduling](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) ## Questions people ask **Can an AI worker publish content on its own?** No. A Polaris worker delivers its work as a comment on the task, with any files attached, and stops there. A person reads it, closes the task and rates the delivery. The machine never marks its own work done. **How does a worker learn our brand voice?** During hiring you describe how the team writes, and that becomes a SKILL.md file attached to the worker. It is a real file you can open and edit, so correcting the voice means editing a document rather than re-prompting. **What does a marketing worker cost per month?** There is no monthly fee. Polaris bills about $2 per human-equivalent hour of delivered work, calculated from an open formula and logged job by job. A worker that delivers nothing in a month costs nothing that month. **Do we have to move off Notion to use this?** No. Notion is one of the connections in the catalog, so a worker can read from a Notion space while your team keeps writing there. Teams that later consolidate do it because they stopped wanting two places for the same document, not because Polaris forces it. **Which marketing roles can we hire as workers?** The roster includes a content writer, an SEO specialist, a social media manager and a competitive analyst, among others. Each is hired the same way and given a different skill file and different connections. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/use-cases/marketing/seo-content-production - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/ai-workers/social-media-manager - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/glossary/skill-file --- --- title: "AI Content Calendar Management | Polaris" description: "Brief an AI worker to keep your content calendar current: slots become owned tasks, gaps get flagged weekly, and every change waits for a human approval." url: https://www.polarishq.co/use-cases/marketing/content-calendar section: Use cases updated: 2026-08-21 --- # A content calendar that maintains itself between meetings The calendar stops rotting the moment somebody other than you is responsible for updating it every week. ## The short answer A content calendar in Polaris is a workstream with a lane per stage and one task per planned piece, each with an owner and a date. An AI worker with the Google Calendar and Notion connections keeps it current: it opens tasks for upcoming slots, flags weeks with nothing planned, and posts a gap list every Friday as a comment. A person approves what actually gets scheduled. - **Connections:** Google Calendar · Notion · Slack - **Lives in:** A workstream, list and board share lanes - **Cadence:** Weekly recurring task ## Why calendars die A content calendar is accurate on the day it is built and wrong two weeks later. Not because anyone is careless, but because updating it is nobody's actual job. The person who made the spreadsheet moved on to writing, and the spreadsheet quietly became a record of what you once intended to publish. The fix is not a better template. It is making the maintenance a task that somebody holds, with a date on it, every week. That somebody can be an AI worker. ## Setting it up About fifteen minutes, once. 1. **Create the workstream** — Call it Content. Add lanes for the stages you really use: Ideas, Briefed, Drafting, In review, Scheduled. List view and board view share these lanes, so you organise once. 2. **Hire the worker** — Tell your Chief of Staff the calendar keeps going stale. It runs a short interview: a name, the role, what the worker should be great at, which tools it needs. Pick Google Calendar and Notion. The answers become an editable SKILL.md. 3. **Write the maintenance brief into the skill file** — Your publishing cadence, the lead time each format needs, which lane a piece should be in by which day, and who owns each format by default. 4. **Assign the recurring task** — One task, due every Friday: review the next four weeks, open a task per unfilled slot, and report gaps. Put it in the Focus lane so it never scrolls away. 5. **Approve the first delivery** — The worker posts its review as a comment on the task, with the proposed new tasks listed. You read it, close the task and rate it. Nothing was created behind your back. ## What comes back every week - **The gap list** — Weeks in the next month with no piece in Briefed or later, named by format and audience rather than left as an empty cell. - **Slipped pieces** — Anything still in Drafting past its lead time, with how many days it has been there and who owns it. - **Proposed tasks** — A task per gap, prefilled with bucket, lane, format and a suggested owner, waiting for one click. - **Calendar collisions** — Two pieces landing the same morning, or a publish date on a company holiday it read off Google Calendar. > **Suggestions, never silent tasks** > > Polaris treats machine output as a proposal. The Inbox holds prefilled task suggestions with bucket, lane, labels and owner already set, and one click accepts them. Nothing enters your board because a worker thought it should. ## Related marketing work - [use-cases/marketing/seo-content-production](https://www.polarishq.co/use-cases/marketing/seo-content-production) - [use-cases/marketing/social-media-scheduling](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) - [use-cases/marketing/campaign-planning](https://www.polarishq.co/use-cases/marketing/campaign-planning) ## Questions people ask **Does the worker create tasks automatically?** It proposes them. Prefilled task suggestions arrive in the Inbox with bucket, lane, labels and owner already set, and a person accepts them with one click. Polaris never adds a task to your board without approval. **Can it read our existing calendar in Notion?** Yes, if you connect Notion. Connections are authorised once for the whole organisation and stored server-side, so the worker reads the space while your team keeps editing it as usual. **What happens to the calendar when the worker is not running?** It is a normal workstream of tasks, so people use it exactly as they would any board in Polaris. The worker maintains it; it does not own it. **How is a weekly calendar review billed?** By the human-equivalent hours the delivery represents, at about $2 an hour, itemised on the worker's work log. A short weekly review is a short line on the bill, and you can challenge any line from the log itself. ## Related - https://www.polarishq.co/use-cases/marketing - https://www.polarishq.co/use-cases/marketing/seo-content-production - https://www.polarishq.co/use-cases/marketing/campaign-planning - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/glossary/focus-lane --- --- title: "SEO Content Production with AI Workers | Polaris" description: "Give an AI worker a skill file with your outline rules and internal-link policy, assign the brief, and get a sourced draft as a comment with the file attached." url: https://www.polarishq.co/use-cases/marketing/seo-content-production section: Use cases updated: 2026-08-21 --- # SEO content production, briefed once and delivered as a file The difference between usable drafts and filler is the brief, and the brief belongs in a file the worker actually reads. ## The short answer SEO content production in Polaris means assigning a brief to an AI worker whose skill file holds your outline structure, internal-link rules and sourcing standard. A cloud machine wakes for the task, runs live web search, ticks its acceptance criteria as it writes, and posts the draft as a comment with the file attached. An editor decides whether it publishes. - **Connections:** web search · Google Drive · Notion - **Delivered as:** A comment on the task, file attached - **Checked by:** Acceptance criteria the worker ticks ## Bad AI content is a briefing failure Ask a model for a thousand words on a keyword and you get the average of everything ever written about that keyword. That is not a model problem. It is what happens when the only input is a topic. A worker in Polaris starts from a skill file you wrote: how your outlines are built, which claims need a source, how many internal links a piece carries and where they point, what your team refuses to say. Then the task itself carries the angle, the audience and the acceptance criteria. The output still needs an editor. What changes is where the editor starts. ## What goes in the skill file, not the prompt Prompts vanish. A SKILL.md is a document on the worker that survives every task. - **Outline shape** — Whether you open with the answer, how deep the headings go, whether you allow a conclusion section at all. - **Sourcing standard** — Which claims need a link, whether vendor blogs count, how to handle a number you cannot verify. - **Internal-link policy** — How many links per piece, that they point to pages that exist, and which hub each cluster reports to. - **The banned list** — The words your team will not publish. Editing that list means editing a file, not rewriting a prompt in six places. ## How one piece moves 1. **Write the task, not the prompt** — Target query, the reader, the angle, and the acceptance criteria: sources linked, a named example, an answer paragraph under 75 words, three title options. 2. **Assign it to the worker** — Same drag, same assignment flow as assigning it to a person. Humans and agents are the same kind of member. 3. **The machine runs** — A cloud machine claims the job, compiles the worker's instructions and skill files with the task context, and runs a live tool loop with real web search. It posts progress while it works. 4. **The checklist ticks** — Each acceptance criterion is marked as it is met, so you can see what the worker believes it has done before you read a word of the draft. 5. **Delivery arrives as a comment** — The draft, plus the file, plus what it could not verify. The task is still open. 6. **You edit and close** — The human closes the task and rates the delivery. That rating is the review. ## What one research-heavy piece looked like From the competitor-pricing brief recorded for the delivery demo, sped up but not staged. - **~7 min** — Machine time, end to end. Live web research included - **$2.80** — Billed for the delivery. At about $2 per human-equivalent hour - **~1h24m** — The human clock it replaced. Roughly $70 at $50 an hour > **The part that stays yours** > > A worker can find sources, structure an argument and write clean prose. It cannot know that a claim will embarrass you in front of a customer, or that the angle is technically true and strategically wrong. That is why the machine never closes its own task. ## Questions people ask **Does the worker have real web access?** Yes. Web search is one of the connections in the catalog, and the delivery runs a live tool loop rather than answering from memory. The recorded demo on the Polaris site shows a worker verifying competitor pricing on the open web during a task. **Can it write directly into our CMS?** No. Deliverables arrive as a comment on the task with files attached, and a person moves them onward. Publishing stays a human action. **How do we stop it producing generic pages at scale?** Give each page a task with its own angle and acceptance criteria, and review the output as content rather than as volume. A page that has nothing specific to say should not be commissioned, and the acceptance criteria are where you enforce that. **What is billed if we reject the draft?** The delivered work is still billed by the same open formula, and the line appears on the work log where you can challenge it. Nothing delivered means nothing billed, but a draft you did not like was still delivered. ## Related - https://www.polarishq.co/use-cases/marketing - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/delivery-comment --- --- title: "Campaign Planning with AI Workers | Polaris" description: "Turn a campaign plan into a workstream with one dated task per channel. An AI worker drafts it, flags what has no owner, and reports the week before launch." url: https://www.polarishq.co/use-cases/marketing/campaign-planning section: Use cases updated: 2026-08-21 --- # Campaign planning where the plan becomes owned tasks Most campaigns do not fail at the idea. They fail at the fourteen small things nobody agreed to own. ## The short answer Campaign planning in Polaris starts as a workstream with one dated task per channel and deliverable. An AI worker drafts the plan document from the brief, proposes a task for every asset the launch needs, names what has no owner, and posts a readiness check as launch week approaches. Marketing keeps the offer, the budget and the date. The worker keeps the list honest. - **Connections:** Slack · Google Drive · HubSpot - **Structure:** One workstream, one task per deliverable - **Readiness check:** Recurring task in the Focus lane ## The gap between the plan and the launch A campaign plan is usually a good document that nobody converts into work. It says what the campaign is, and it does not say who is writing the second email or when the landing page copy is due. Two weeks later somebody discovers the landing page copy was due yesterday. In Polaris the plan and the work are the same object. The document lives in Docs, versioned and commentable, and every line in it that requires somebody to do something becomes a task with an owner and a date. ## What a worker drafts, what it cannot decide | Piece of the campaign | Worker's part | Human's part | | --- | --- | --- | | The offer | Writes it up once decided, in the plan doc | Decides what the offer is | | Channel list | Proposes channels used in past campaigns in this workstream | Cuts the list to what the team can actually run | | Asset inventory | One task per asset, prefilled with format, owner and due date | Approves the owners | | Timeline | Back-dates each task from the launch date and flags collisions | Sets the launch date | | Readiness | Posts what is unowned, undated or still in draft | Decides whether to slip or ship | ## The readiness comment, seven days out One recurring task, assigned to the worker, pinned in Focus. - **Unowned work** — Every task in the campaign workstream with no assignee, named rather than counted. - **Undated work** — Tasks with no due date, which is how things quietly land after the launch. - **Still drafting** — Assets that should be in review by now, with how many days they are late. - **Dependencies** — Where one task blocks three others, so you know which slip is the expensive one. > **Everything in one place, including the chat about it** > > The plan doc, the tasks, the Slack thread that changed the offer and the worker's delivery comments all sit in the same product. There is no version of the campaign living in a fourth tool that only one person has open. ## Campaign work continues here - [use-cases/marketing/content-calendar](https://www.polarishq.co/use-cases/marketing/content-calendar) - [use-cases/marketing/social-media-scheduling](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) - [use-cases/marketing/competitor-monitoring](https://www.polarishq.co/use-cases/marketing/competitor-monitoring) ## Questions people ask **Can the worker send the campaign emails?** No. A worker drafts and delivers as a comment with files attached; sending happens in your email tool, by a person. HubSpot is in the connection catalog, so a worker can be given access to read from it, but the decision to send stays human. **How do we keep the plan and the tasks in sync?** They are the same workstream. The plan document lives in Docs alongside the board, and the worker's readiness comment reads the tasks rather than a copy of them, so there is nothing to reconcile. **What if the launch date moves?** Change the date on the launch task and re-assign the readiness check. The worker re-dates the dependent tasks in its next delivery and posts what now collides, as a comment you approve. **Is this useful for a two-person marketing team?** That is where the readiness check earns the most, because a small team has no producer role. The software costs nothing regardless of headcount, and you only pay for the hours the worker delivers. ## Related - https://www.polarishq.co/use-cases/marketing - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/for/startups --- --- title: "Automated Competitor Monitoring | Polaris AI Workers" description: "A recurring task, a worker with web search, and a weekly diff of what changed on your competitors' pricing, product and careers pages. Delivered as a comment." url: https://www.polarishq.co/use-cases/marketing/competitor-monitoring section: Use cases updated: 2026-08-21 --- # Competitor monitoring that actually happens every week Everyone agrees competitor tracking matters and nobody has done it since the last time a deal was lost over it. ## The short answer Competitor monitoring in Polaris is a recurring task assigned to an AI worker with the web search connection. Each week the worker visits the named pages you listed in its skill file, compares them to the last delivery, and posts a diff as a comment: what changed on pricing, positioning, product and hiring, with links. Marketing decides which change is worth answering. - **Connection:** web search - **Cadence:** Weekly recurring task - **Delivered as:** A dated diff comment with links ## Why it never gets done Competitor tracking is important and never urgent, which is the exact profile of work that gets postponed forever. It also has no natural owner: it is nobody's deliverable, so it becomes everybody's good intention. Giving it to an AI worker changes the economics. The job is boring, repetitive, and requires reading twelve pages carefully. A cloud machine wakes up on Monday morning and does it whether or not anyone remembered. ## What goes in the brief The skill file names the targets. The task names the week. - **The page list** — Exact URLs, not company names. Pricing, the product pages you care about, changelog, careers. A named page produces a checkable diff; a company name produces vague summary. - **What counts as a change** — A new pricing tier matters. A rotated testimonial does not. Say so in the file, and the noise drops immediately. - **How to report** — Old value, new value, the link, the date you last saw the old one. Not a paragraph of interpretation. - **What to escalate** — Which changes should arrive as a proposed task rather than a line in a report. ## The report you want versus the report you get by default **Vague monitoring** - Competitor X continues to focus on enterprise - Their messaging appears to have shifted - Pricing seems competitive - No sources, no dates **A diff** - Team tier moved from $12 to $15 per seat, checked 18 Aug, link - Homepage h1 changed, both versions quoted - Two backend roles posted in Berlin - Nothing changed on the other nine pages > **The escalation path** > > When something on the list crosses the threshold you set, the change arrives as a prefilled task suggestion with bucket, lane and owner set, waiting for one click. A competitor cutting their price becomes a decision on somebody's board rather than a paragraph in a report nobody finished reading. ## Questions people ask **Does it scrape competitor sites?** It uses the same open web research any person does, through the web search connection, and reports what it read with links. Nothing is stored on your behalf beyond the delivery comment and the files attached to it. **How much does a weekly competitor check cost?** It is billed by human-equivalent hours at about $2 an hour. Searches are one of the observable inputs in the open formula, so a week with twelve pages checked costs more than a week with three, and each session is itemised on the work log. **Can it monitor competitors' social accounts?** Instagram is in the connection catalog and open web research covers public pages. Keep the target list to pages that produce a checkable diff, because a feed of posts generates summary rather than evidence. **Who reads the report?** It is a comment on a task in your board, so anyone in the workstream sees it, and the activity feed shows when it landed. There is no separate report inbox to forget about. ## Related - https://www.polarishq.co/use-cases/marketing - https://www.polarishq.co/use-cases/marketing/campaign-planning - https://www.polarishq.co/use-cases/executive/strategic-research - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/delivery-comment --- --- title: "Social Media Drafting and Scheduling | Polaris" description: "An AI worker drafts a week of channel-specific social posts against your voice file, delivers them in one comment, and a person approves what goes out." url: https://www.polarishq.co/use-cases/marketing/social-media-scheduling section: Use cases updated: 2026-08-21 --- # A week of social posts drafted before you open the app Drafting is the part that eats the week. Approving is the part that needs you. ## The short answer Social media work in Polaris runs as a weekly task assigned to an AI worker. The worker drafts posts per channel against the voice rules in its skill file, pulls the week's material from your content workstream, and delivers all of them in one comment with the reasoning attached. A person picks what goes out, edits anything that is close, and schedules it. - **Connections:** Instagram · Slack · Google Drive - **Cadence:** One weekly task, one delivery - **Approval:** Human picks and schedules every post ## The real cost of social The expensive part of social media is not strategy and it is not scheduling. It is producing twelve pieces of short copy a week that sound like the same company, in the middle of everything else. That production has a shape: source material, a channel's constraints, your voice, a hook. A worker with the voice rules written down and access to what you shipped this week can produce the first version of all twelve. What it cannot do is know that the third one will read badly to your biggest customer. ## The weekly loop 1. **Material lands in the workstream** — A published post, a product change, a customer question worth answering. Anything in the Content workstream is source material. 2. **The task fires** — One recurring task: draft next week's posts, per channel, with the acceptance criteria you set. Number of posts, formats, which ones need an image brief. 3. **The worker drafts** — A cloud machine wakes, reads the material and the voice file, and writes. It ticks each criterion as it completes it. 4. **One delivery comment** — Every draft in one place, grouped by channel, each with the source it came from so you can check the claim. 5. **You approve and schedule** — Cut what does not work, edit what nearly does, and publish from your own tools. The worker does not post on your behalf. ## What separates a usable draft from filler - **It cites its source** — Every post points back to the doc, task or page it came from, so approving takes seconds instead of a fact-check. - **It respects the channel** — Length, format and convention differ per channel, and the skill file says how for each one you actually use. - **It shows the alternatives** — Two hooks for the same post is more useful than one polished hook you have to argue with. - **It admits the gaps** — Where the worker had no material for a slot, it says so rather than inventing something to fill it. > **Approval is not a formality here** > > Social copy is the fastest way to say something your company does not believe. Polaris is built so the machine cannot publish: work arrives as a comment, a human closes the task, and the rating is the review. Keep that step slow. ## Questions people ask **Can Polaris post to Instagram directly?** Instagram is one of the connections a worker can be given, but the delivery model does not change: work arrives as a comment for a person to approve. Treat scheduling and publishing as a human step. **How do we keep every post from sounding identical?** The skill file is where you set variety rules, and the task sets the acceptance criteria for the week. Asking for two hooks per post and rejecting the ones that repeat a structure fixes this faster than editing prompts. **What does a week of drafts cost?** It is billed on human-equivalent hours at about $2 each, computed from observable effort including the finished prose produced, and logged as a line you can challenge on the worker's work log. **Can we run different workers per channel?** Yes. Each worker is a member with its own skill file and connections, so a worker that writes long-form and a worker that writes short social copy can have different voice rules and different sources. ## Related - https://www.polarishq.co/use-cases/marketing - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/use-cases/marketing/campaign-planning - https://www.polarishq.co/ai-workers/social-media-manager - https://www.polarishq.co/integrations/instagram - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/delivery-comment --- --- title: "Polaris for Sales Teams | AI Workers on the Pipeline" description: "Pipeline reviews, proposals, account research and CRM hygiene handled by AI workers on the same board as your reps, with every deliverable approved by a human." url: https://www.polarishq.co/use-cases/sales section: Use cases updated: 2026-08-21 --- # Sales work, with the admin handed to an AI teammate Selling is a conversation. Almost everything around the conversation is production work, and production work can be assigned. ## The short answer Polaris puts AI workers on the same board as a sales team. A worker is hired in chat, given a skill file with your qualification criteria and proposal structure, and connected to HubSpot, Gmail or web search. It researches accounts, drafts proposals, lists the deals missing a next step, and posts each result as a comment. Reps keep the calls and the commitments. - **Typical connections:** HubSpot · Gmail · web search - **Software cost:** $0, unlimited reps - **Work cost:** ~$2 per human-hour delivered - **Hiring time:** ~60 seconds, in chat ## What reps do instead of selling Ask a rep where their week went and the answer is rarely calls. It is the account brief written the night before, the proposal rebuilt from an old one, the CRM fields updated on Friday from memory, the follow-up that needed a summary of a forty-minute conversation. None of that is judgement. All of it is production: gather, structure, write, check. It is exactly what a briefed worker does well, and doing it at two in the morning costs the same as doing it at noon. ## The five sales jobs, and what each needs | Job | Connections | Delivered as | Human keeps | | --- | --- | --- | --- | | Pipeline management | HubSpot · Slack | A stage-by-stage review plus one task per stalled deal | The forecast call | | Proposal writing | Google Drive · HubSpot | A drafted proposal file on the task | Price, scope, and hitting send | | Lead research | web search · HubSpot | An account brief with sources linked | Who is worth a call | | CRM hygiene | HubSpot | A correction list, record by record | Approving what gets changed | | Sales enablement content | Google Drive · web search | Battle cards and objection docs, versioned | What we claim in a deal | ## How a sales worker is briefed The interview takes about a minute; the skill file it produces is what makes the output usable. - **Your qualification bar** — What makes an account worth research time, and what disqualifies it in one line. - **The proposal skeleton** — Sections in order, what is negotiable, which numbers a worker may never state on its own. - **Stage definitions** — What has to be true for a deal to sit in each stage, so the review is checkable rather than a matter of opinion. - **Escalation rules** — Which findings become a task in Focus and which stay a line in a report. ## Where the line sits **The rep keeps** - Every conversation with a person - Discovery, and what the answers mean - Price, terms and concessions - The commitment to a customer **The worker takes** - Account briefs before the first call - The Monday pipeline review, assembled the same way every week - Proposal drafts from the call notes - Finding the records that will break the forecast > **No seats to buy for a growing team** > > Sales teams change size constantly, and per-seat tools punish that. Polaris charges nothing for the board, the docs or the chat, at any headcount, and bills about $2 per human-equivalent hour of work a worker actually delivered, itemised on its work log. ## The five sales jobs in detail - [use-cases/sales/pipeline-management](https://www.polarishq.co/use-cases/sales/pipeline-management) - [use-cases/sales/proposal-writing](https://www.polarishq.co/use-cases/sales/proposal-writing) - [use-cases/sales/lead-research](https://www.polarishq.co/use-cases/sales/lead-research) - [use-cases/sales/crm-hygiene](https://www.polarishq.co/use-cases/sales/crm-hygiene) - [use-cases/sales/sales-enablement-content](https://www.polarishq.co/use-cases/sales/sales-enablement-content) ## Questions people ask **Does Polaris replace our CRM?** No. HubSpot is one of the connections in the catalog, so a worker can be given access to it while the CRM stays your system of record. Polaris is where the work around the pipeline lives: the tasks, the docs, the deliverables and the review. **Can a worker email a prospect?** Gmail is in the connection catalog, but the delivery model is fixed: work arrives as a comment on a task and a person closes it. Treat any outbound message as something a human sends. **How is this different from an AI feature inside a CRM?** A CRM feature acts inside the CRM. A Polaris worker is a member of your team with its own tasks, its own editable skill file, a work log and an audit trail, and it can be assigned anything on the board rather than only records in one product. **What does it cost for a five-person sales team?** The software is free for any number of people. You pay about $2 for each human-equivalent hour of delivered work, so the bill tracks how much you hand over rather than how many reps you employ. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/use-cases/sales/lead-research - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/for/agencies --- --- title: "AI Pipeline Management and Deal Reviews | Polaris" description: "An AI worker reads the pipeline, writes the stage-by-stage review, and opens a dated task for every deal without a next step. Reps arrive to a prepared meeting." url: https://www.polarishq.co/use-cases/sales/pipeline-management section: Use cases updated: 2026-08-21 --- # A pipeline review that is assembled before the meeting The forecast meeting is worth having. Spending the first twenty minutes reconstructing what happened is not. ## The short answer Pipeline management in Polaris means a weekly task assigned to an AI worker with the HubSpot connection. The worker reads every open deal, checks it against the stage definitions in its skill file, and delivers a stage-by-stage review as a comment: what moved, what stalled, what has no next step and no date. Each stalled deal arrives as a proposed task for the rep who owns it. - **Connections:** HubSpot · Slack - **Cadence:** Weekly, before the forecast call - **Output:** Review comment plus one task per stalled deal ## Two different meetings In the first kind, the manager reads the pipeline aloud and each rep explains from memory what happened. It takes an hour and produces a shared feeling rather than a decision. In the second, everyone has already read the same written review: which deals moved a stage, which have not been touched in fourteen days, which are dated to close this month with no meeting booked. The meeting is about the five hard ones. It takes twenty minutes. The only difference is that somebody wrote the review. That somebody does not have to be a person. ## What the review contains - **Movement** — Deals that changed stage since the last delivery, in both directions. Backwards movement gets named, because that is the number that predicts the quarter. - **Silence** — Open deals with no activity for longer than the threshold in the skill file, with the number of days and the owner. - **Dates that cannot be true** — Deals dated to close inside a window shorter than your average cycle for that stage. - **Missing next steps** — Every open deal with no next action recorded, which is the single most reliable predictor of a deal that will not close. - **What the worker could not check** — Fields it could not read or deals with data it did not trust, stated rather than smoothed over. ## From review to work 1. **The worker delivers the review** — One comment on the weekly task, structured the same way each week so the comparison is easy. 2. **Stalled deals become suggestions** — Prefilled task suggestions land in the Inbox with the deal, the owner, the lane and a due date already set. 3. **The rep accepts what is real** — One click each. A deal the rep knows is dead gets rejected instead, and that judgement stays with the person who has the context. 4. **The important ones go to Focus** — Deals closing this week sit in the Focus lane, which is pinned first in every view and does not scroll away. > **The worker never edits the pipeline** > > It reads, reports and proposes. Changing a stage, a date or an amount is a decision with a forecast attached, and Polaris keeps decisions with the people who answer for them. Agents deliver; humans close. ## Questions people ask **Does it write back to HubSpot?** The deliverable is a review comment and a set of proposed tasks. Treat any change to a CRM record as a human action, approved by the rep who owns the deal. **How does it know what stalled means for us?** You define it during hiring, and it lives in the worker's SKILL.md as a real file. Changing the threshold from fourteen days to seven means editing a line in that document. **Can each rep get their own review?** Yes. Assign the worker one task per rep, or one per segment, and each delivery is a separate comment on a separate task with its own audit trail. **What does a weekly pipeline review cost?** It is billed by human-equivalent hours at about $2 each, computed from observable effort and clamped between five minutes and eight hours per session. Every session appears as a challengeable line on the worker's work log. ## Related - https://www.polarishq.co/use-cases/sales - https://www.polarishq.co/use-cases/sales/crm-hygiene - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/glossary/work-log --- --- title: "AI Proposal Writing for Sales Teams | Polaris" description: "Assign the proposal to an AI worker whose skill file holds your structure and pricing rules. It delivers the finished document as a file on the task." url: https://www.polarishq.co/use-cases/sales/proposal-writing section: Use cases updated: 2026-08-21 --- # Proposals drafted from your call notes, not from a template Every proposal is 70% the same document and 30% the reason this customer is different. The 30% is the part worth your evening. ## The short answer Proposal writing in Polaris means assigning a task to an AI worker whose skill file holds your section order, scope language and pricing rules. The worker reads the call notes and the deal context from your docs, drafts the proposal, ticks the acceptance criteria you set, and delivers it as a file attached to a comment. The rep edits the argument and sends it. - **Connections:** Google Drive · HubSpot - **Delivered as:** A generated file on the task - **Never automated:** Price, terms, and sending ## The Friday-night proposal A proposal usually gets written at the worst possible time: after the call, before the follow-up, by the person who most needs to be preparing for tomorrow. So it gets built by opening the last one and replacing names, which is how a paragraph about a different customer's warehouse ends up in a software contract. A worker starts from the same materials the rep would, and it does not get tired at the fourth section. ## Where each part of the document comes from | Section | Source | Who is responsible | | --- | --- | --- | | Understanding of the problem | Discovery notes in Docs | Worker drafts, rep corrects the emphasis | | Scope of work | The skill file plus the deal's notes | Worker drafts from your standard scopes | | Timeline | Your delivery norms in the skill file | Rep confirms it is achievable | | Pricing | Never generated | Rep, every time | | Terms | Your approved terms document | Worker inserts, legal owns the wording | | The reason us | Discovery notes | Rep writes or rewrites | ## Acceptance criteria worth setting The worker ticks each one as it works, so you see what it believes it did before you open the file. - **Every claim traceable** — No statement about the customer's situation that is not in the notes. This is the criterion that stops invented context. - **Pricing left blank** — A visible placeholder rather than a number, so nobody can send a proposal with a machine's guess in it. - **Named deliverables** — Each line of scope has an artifact and a date, not a verb. - **One page of summary** — A summary the buyer can forward to somebody who was not on the call. > **A proposal is a promise** > > Polaris workers produce files, and files get signed. Keep pricing and scope commitments human, keep the acceptance criteria strict, and read the delivery before it leaves the building. The machine cannot close its own task, which is the point. ## Related sales work - [use-cases/sales/lead-research](https://www.polarishq.co/use-cases/sales/lead-research) - [use-cases/sales/sales-enablement-content](https://www.polarishq.co/use-cases/sales/sales-enablement-content) - [use-cases/sales/pipeline-management](https://www.polarishq.co/use-cases/sales/pipeline-management) ## Questions people ask **Can it produce a PDF?** Yes. Deliveries arrive as comments on the task and can include generated files, including PDFs and documents, attached where the rest of the deal context lives. **Where do the call notes come from?** From Docs in Polaris, or from files in a connected Google Drive. The notes need to exist somewhere the worker can read; a proposal drafted from nothing reads exactly like a proposal drafted from nothing. **How do we stop it inventing customer details?** Make traceability an acceptance criterion: no statement about the customer that is not in the notes. The worker ticks the criteria as it goes, and anything it could not source it reports in the delivery comment. **Is a proposal draft worth about $2?** It is billed on the human-equivalent hours the formula computes from observable effort, including the finished prose produced, so a long proposal costs more than a short one. Each session is a line on the work log you can challenge. ## Related - https://www.polarishq.co/use-cases/sales - https://www.polarishq.co/use-cases/sales/sales-enablement-content - https://www.polarishq.co/use-cases/sales/lead-research - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files --- --- title: "AI Lead and Account Research | Polaris" description: "An AI worker with web search builds a sourced account brief for every meeting: what the company does, what changed recently, and why they might care now." url: https://www.polarishq.co/use-cases/sales/lead-research section: Use cases updated: 2026-08-21 --- # Account briefs waiting for you before the first call Ten minutes of research changes a first call. Nobody has ten minutes before a first call. ## The short answer Lead research in Polaris is a task assigned to an AI worker with the web search connection. The worker reads the company's public material, finds what changed recently, and delivers a brief as a comment: what they sell, who they sell to, the trigger worth mentioning, and the questions to ask, each with a link. The rep decides whether the account is worth the call. - **Connections:** web search · HubSpot - **Delivered as:** A sourced brief per account - **Runs:** Overnight, on a cloud machine ## Research is the cheapest thing to buy back Account research has a rare property: it is genuinely useful, entirely public, and completely mechanical. Everything a rep would find in ten minutes is on the open web, and finding it requires no relationship and no judgement until the last step. It is also the first thing dropped when the day fills up, which is why so many first calls open with a question the buyer answered on their homepage. ## A recorded research delivery The competitor-pricing brief in the Polaris delivery demo, a real machine session, sped up and not staged. - **~7 min** — Machine time with live web research - **$2.80** — Billed for that delivery - **~1h24m** — The equivalent human clock. About $70 at $50 an hour ## What a usable brief contains The skill file fixes this shape so every brief reads the same way and comparison is possible. - **What they actually sell** — In one paragraph, in their words, with the page it came from. - **The trigger** — The thing that changed recently: funding, a launch, a hire, a new office, a pricing change. A brief with no trigger says so. - **The people** — Public roles relevant to the deal, and what those roles are typically measured on. - **Three questions** — Questions the rep could not have asked without reading this, which is the honest test of whether the research was worth doing. - **The disqualifiers** — What suggests this account is a poor fit, stated first rather than buried. > **It runs while nobody is awake** > > Assign fifteen account briefs at six in the evening and close the laptop. A cloud machine claims each job from the queue and works through them, and the deliveries are waiting as comments in the morning. This is the difference between an agent on your laptop and an agent on a machine. ## Questions people ask **Where does the research come from?** The open web, through the web search connection, with links in the delivery so every claim is checkable. If a fact could not be verified, the brief says so rather than filling the gap. **Can it build the target list too?** It can research the accounts you name and report what it found. Deciding which companies belong on the list is a positioning judgement, and it stays with the person who owns the number. **Does it work while my laptop is closed?** Yes. Tasks assigned to a worker are claimed by a runtime on a cloud machine, so work continues after you close the lid and the delivery is waiting when you come back. **How is a batch of briefs billed?** Per session, in human-equivalent hours at about $2 each. Searches are one of the observable inputs to the open formula, so a deeply researched account costs more than a shallow one, and every line is on the work log. ## Related - https://www.polarishq.co/use-cases/sales - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/use-cases/executive/strategic-research - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed - https://www.polarishq.co/glossary/human-equivalent-hours --- --- title: "CRM Hygiene with an AI Worker | Polaris" description: "An AI worker audits your CRM every week and delivers a record-by-record correction list. A human approves every change, so the forecast stops drifting quietly." url: https://www.polarishq.co/use-cases/sales/crm-hygiene section: Use cases updated: 2026-08-21 --- # CRM cleanup as a standing job instead of a quarterly panic Nobody sets out to let the CRM rot. It rots because cleaning it is a four-hour job with no owner and no deadline. ## The short answer CRM hygiene in Polaris is a recurring task assigned to an AI worker with the HubSpot connection. The worker audits records against the rules in its skill file, then delivers a correction list as a comment: missing owners, stages that contradict activity, close dates in the past, duplicates. A person approves each change, because a field on a deal is a forecast input. - **Connection:** HubSpot - **Cadence:** Weekly recurring task - **Approval:** Every correction reviewed by a person ## Dirty data is a forecast problem before it is a data problem A deal with no owner does not get worked. A close date three weeks in the past keeps a number in the quarter that will not land. Two records for the same company mean two reps calling the same buyer in the same week. None of this is dramatic on any single record, which is why it accumulates. It surfaces at the end of a quarter, as a surprise, in a meeting where nobody can explain it. ## What the audit looks for | Check | Why it matters | What the worker delivers | | --- | --- | --- | | No owner | Unowned deals are unworked deals | The record, the amount, a proposed owner | | Stage contradicts activity | A negotiation with no meeting is not a negotiation | The stage, the last activity date, the mismatch | | Close date in the past | Keeps dead pipeline in the number | The record and how many days overdue | | Probable duplicate | Two reps, one buyer | Both records and what matches | | Empty required fields | Reporting silently degrades | The field, the record, and what it should probably be | ## The weekly cycle 1. **The rules live in the skill file** — What each stage requires, which fields are mandatory, what counts as a duplicate. Written once, edited as a document when the rules change. 2. **The worker audits** — A cloud machine wakes for the task, reads the records through the connection, and works down the checklist, ticking each acceptance criterion. 3. **The correction list arrives** — One comment, grouped by check, with the record and the proposed fix side by side rather than a count of problems. 4. **A person approves** — The rep or the operations owner applies what is right and rejects what is not. Approving a list takes minutes; finding the list took the hours. 5. **The trend gets visible** — Same task every week, same structure, so a check that keeps producing the same failures tells you the process is broken, not the data. > **Why the worker does not just fix it** > > Automatic correction of CRM fields quietly rewrites your forecast, and there is no way to tell afterwards which change came from a person and which from a machine. Polaris keeps the delivery as a proposal and keeps the audit trail intact: the machine never marks its own work done. ## Questions people ask **Can it fix records automatically if we want that?** The Polaris delivery model is that work arrives as a comment for a person to close. Corrections are proposals, and a human applies them. That is a deliberate constraint, not a missing feature. **How long does the first audit take?** The first pass is the largest, because every historical problem surfaces at once. Sessions are clamped between five minutes and eight hours, and the work log shows what each session covered and what it cost. **Does this need our CRM to be tidy already?** No. A messy CRM produces a longer correction list, which is the useful output. What it does need is written rules, because an audit without a definition of correct produces opinions. **Who should own the approval?** Whoever answers for the forecast. In small teams that is the founder or the sales lead; in larger ones, sales operations. Polaris assigns the approval as a task with a date like any other. ## Related - https://www.polarishq.co/use-cases/sales - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/use-cases/data/data-quality-monitoring - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/work-log --- --- title: "Sales Enablement Content with AI Workers | Polaris" description: "An AI worker keeps battle cards, objection handling and one-pagers current in versioned docs, checks competitor claims on the web, and flags what went stale." url: https://www.polarishq.co/use-cases/sales/sales-enablement-content section: Use cases updated: 2026-08-21 --- # Battle cards that are still true this quarter Enablement content is written once, used constantly, and updated never. The updating is the assignable part. ## The short answer Sales enablement content in Polaris lives in versioned docs with file review and comments, maintained by an AI worker. The worker drafts battle cards and objection responses from won and lost deal notes, checks competitor claims through web search, and posts a list of documents that no longer match reality. Sales leadership approves what the team is allowed to say. - **Connections:** Google Drive · web search - **Lives in:** Docs, versioned, with file review - **Reviewed by:** Sales leadership, before use ## The half-life of a battle card A battle card is accurate the week it is written. Then the competitor changes their pricing page, ships the feature you said they lacked, and your card starts costing you deals in a way nobody can trace. Reps rarely report this. They quietly stop using the card and improvise, which is worse, because now every rep says something different and none of it is reviewed. ## What the worker maintains - **Battle cards** — One per named competitor, with claims dated and linked so a rep can see when each line was last checked. - **Objection responses** — Drafted from what actually came up in lost-deal notes, not from a list of objections somebody imagined. - **One-pagers per segment** — The same product argument written for the buyer who cares about a different thing. - **The stale list** — Documents whose claims no longer match what the worker found on the web this month, named individually. ## Two ways to keep enablement current **The quarterly refresh** - Someone owns it in addition to their job - It slips, then it slips again - Every card updated on the same day whether it needed it or not - Nobody knows which claim was checked when **A standing task** - A worker checks the claims monthly - Only what changed comes back - Each line carries the date it was verified - Sales leadership approves changes in file review > **Never trash a competitor with a machine** > > A worker will happily write whatever the skill file tells it to. Put your rules of engagement in that file: compare on facts you can link, describe what the other product genuinely does well, and never make a claim a rep would have to defend on a call. Approving the tone is leadership's job, once, in the file. ## Questions people ask **Where does the material come from?** From your own deal notes and docs, plus open web research on public competitor pages. The worker links what it read, so leadership can check any claim before a rep repeats it. **How does versioning work?** Docs in Polaris are versioned, and files support review and comments, so an enablement document has a history and an approver rather than being overwritten silently. **Can reps request updates?** Yes. A rep creates a task in the enablement workstream describing what a buyer said, assigns it to the worker, and the answer comes back as a delivery comment for leadership to approve. **Does this replace enablement headcount?** It replaces the production part of the job. Deciding what the company claims, and training people to say it well, is still a person's work, and the worker's output is only as good as the rules in its skill file. ## Related - https://www.polarishq.co/use-cases/sales - https://www.polarishq.co/use-cases/sales/proposal-writing - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/skill-file --- --- title: "Polaris for Customer Support Teams | AI Workers" description: "Triage, help-center upkeep, escalation tracking and feedback themes handled by AI workers on the same board as your team. Every reply approved by a person." url: https://www.polarishq.co/use-cases/customer-support section: Use cases updated: 2026-08-21 --- # Support work, with an AI teammate on the queue The queue is a conveyor belt of small decisions. Sorting them is mechanical; making them is not. ## The short answer Polaris gives a support team AI workers that sort, draft and maintain rather than answer on their own. A worker is hired in chat, given a skill file with your tone rules and severity definitions, and connected to Gmail, Slack or Linear. It triages incoming mail, drafts replies, keeps help-center articles current and tracks escalations. A person sends every customer-facing message. - **Typical connections:** Gmail · Slack · Linear · WhatsApp - **Software cost:** $0, unlimited agents - **Work cost:** ~$2 per human-hour delivered - **Customer replies:** Always sent by a person ## What actually costs a support team its day Not the hard tickets. The hard tickets are the job, and they are why good support people are worth what they cost. The day goes to the sorting: reading forty messages to find the six that matter, rewriting the same explanation for the ninth time, chasing engineering for an update on an escalation from Tuesday, noticing that the help article contradicts the product. Each of those is a production task with a shape, and each one is assignable. What is not assignable is the sentence a person reads when they are already angry. ## Five support jobs and how they run | Job | Connections | Worker delivers | Person decides | | --- | --- | --- | --- | | Ticket triage | Gmail · Slack | Sorted queue with severity, plus drafted replies | What gets sent | | Help center maintenance | Linear · GitHub · Google Drive | The stale-article list with rewritten drafts | What is published | | Escalation tracking | Linear · Slack | A daily status per open escalation | What gets promised to a customer | | Customer feedback loops | Gmail · Slack · WhatsApp | Monthly themes with quotes and counts | What product does about it | | Response templates | Google Drive | Drafted templates in versioned docs | The voice the team uses | ## What goes in a support worker's skill file Written once during hiring, edited afterwards like any document. - **Severity definitions** — What makes something urgent in your product specifically. Data loss and a billing failure are not the same urgency in every company. - **Tone rules** — How you apologise, whether you use the customer's first name, what you never say about a competitor or a bug. - **The refund and credit boundary** — What a draft may offer and what must be escalated to a person, stated as a number. - **Where the truth lives** — Which docs are authoritative, so a drafted answer cites the help center rather than inventing behaviour. > **The rule that keeps this safe** > > A Polaris worker delivers work as a comment on a task, and a human closes it. Applied to support, that means no message reaches a customer without a person reading it. Deploy a worker to make the queue legible, not to answer for you. ## The five support jobs in detail - [use-cases/customer-support/ticket-triage](https://www.polarishq.co/use-cases/customer-support/ticket-triage) - [use-cases/customer-support/help-center-maintenance](https://www.polarishq.co/use-cases/customer-support/help-center-maintenance) - [use-cases/customer-support/escalation-tracking](https://www.polarishq.co/use-cases/customer-support/escalation-tracking) - [use-cases/customer-support/customer-feedback-loops](https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops) - [use-cases/customer-support/response-templates](https://www.polarishq.co/use-cases/customer-support/response-templates) ## Questions people ask **Is this a helpdesk product?** No. Polaris is a work platform with tasks, docs and chat, plus AI workers you assign work to. Support tickets arrive through connections such as Gmail or Slack and become tasks on your board; the conversation with the customer happens in your own channel. **Can a worker reply to customers directly?** Gmail and WhatsApp are in the connection catalog, but deliveries arrive as comments for a person to approve and close. Treat sending as a human step, deliberately. **How does the team see what the worker did?** Every delivery is a comment on a task, and the workers area has an activity feed and insights. There is a record of what was produced, when, and who closed it. **What does this cost for a support team of six?** The software is free at any headcount. You pay about $2 per human-equivalent hour of delivered work, itemised job by job on the worker's work log, where any line can be challenged. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/customer-support/escalation-tracking - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/ai-workers/customer-success-manager - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "AI Ticket Triage for Support Teams | Polaris" description: "An AI worker reads the support inbox, sorts by severity against your own definitions, drafts replies for the repeatable ones, and leaves every send to a person." url: https://www.polarishq.co/use-cases/customer-support/ticket-triage section: Use cases updated: 2026-08-21 --- # Ticket triage that hands you a sorted queue and drafted replies Reading forty messages to find the six that matter is the most expensive hour in a support team's day. ## The short answer Ticket triage in Polaris means an AI worker with the Gmail connection reads incoming support mail, classifies each message against the severity definitions in its skill file, and delivers a sorted queue as a comment: what is urgent, what is a duplicate of a known issue, what has a drafted reply ready. Support agents send the replies and handle the ones that need a human. - **Connections:** Gmail · Slack - **Cadence:** Per batch, several times a day - **Output:** Sorted queue plus drafted replies ## Triage is a classification problem wearing a costume Nearly every message in a support queue is one of a small number of things: a question already answered in the docs, a bug already known, a billing problem, an angry customer who needs a person, or something genuinely new. A human doing triage is running that classification forty times before lunch. The classification is mechanical once the definitions exist. The definitions are the part only your team can write, because urgent means something different in a payments product than in a design tool. ## How a batch is processed 1. **Messages arrive as work** — The Gmail connection gives the worker access to the support address; Slack signals arrive in the Inbox as prefilled task suggestions rather than silent tasks. 2. **The worker classifies** — Each message is matched against the severity definitions and the known-issue list in the skill file. Anything it cannot classify is marked as needing a person, not guessed. 3. **Repeatable ones get drafts** — Where the answer exists in your help center, the worker drafts a reply that cites the article rather than paraphrasing product behaviour from memory. 4. **One delivery, ordered by severity** — The comment leads with what is urgent. Duplicates are grouped. Every draft sits with the original message so the agent can check it in one read. 5. **Agents send** — A person approves each reply and closes the task. The rating on the delivery is how the worker's classification gets corrected over time. ## What makes triage output trustworthy - **It shows its reasoning** — Each classification names the rule it matched, so an agent can see immediately when a rule is wrong rather than when a customer complains. - **It flags uncertainty** — A message the worker could not classify arrives marked as such. Confident misclassification is the failure mode that costs you a customer. - **It never invents behaviour** — Drafts cite the help article. If no article covers it, the draft says the answer needs a person. - **It groups duplicates** — Fifteen reports of the same outage arrive as one item with fifteen customers attached, which is also how you find out an outage started. > **Angry customers are not a triage category** > > Set the skill file so anything with an emotional signal, a legal word, or a public threat goes straight to a named person with no draft attached. A drafted apology that a machine wrote and a human sent under pressure is how a small problem becomes a screenshot. ## Questions people ask **Does the worker have access to our whole inbox?** It uses the Gmail connection you authorise, once, for the organisation. Credentials are verified live and stored server-side, so workers use them and browsers cannot read them back. **How does it learn our severity definitions?** You write them during hiring and they become a SKILL.md file on the worker. Refining triage means editing that document, which is also the audit trail for why a message was classified the way it was. **Can it close tickets?** No. Polaris workers deliver as comments and humans close tasks. In support that boundary is the safety mechanism, not an inconvenience. **What does triage cost per day?** It is billed in human-equivalent hours at about $2 an hour, computed from observable effort per session and clamped between five minutes and eight hours. Busy days cost more than quiet ones, and every session is a line on the work log. ## Related - https://www.polarishq.co/use-cases/customer-support - https://www.polarishq.co/use-cases/customer-support/response-templates - https://www.polarishq.co/use-cases/customer-support/escalation-tracking - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/skill-file --- --- title: "Help Center Maintenance with AI Workers | Polaris" description: "An AI worker compares your help center against what shipped in Linear or GitHub, lists the articles now wrong, and delivers rewritten drafts to approve." url: https://www.polarishq.co/use-cases/customer-support/help-center-maintenance section: Use cases updated: 2026-08-21 --- # Help articles that keep up with the product Every shipped change quietly makes a help article wrong, and the customer finds out before you do. ## The short answer Help center maintenance in Polaris is a recurring task assigned to an AI worker connected to Linear or GitHub and your docs. The worker compares what shipped against what the articles claim, delivers a list of documents that are now wrong, and attaches a rewritten draft for each. A support lead reviews the drafts in file review and publishes what is correct. - **Connections:** Linear · GitHub · Google Drive - **Cadence:** Monthly, or after each release - **Output:** Stale list plus rewritten drafts ## Documentation rots at the speed of shipping Nobody writes a wrong help article. Articles become wrong because a button moved, a plan was renamed, a limit changed, and the person who made the change had no reason to think about a page written eleven months ago. Support finds out through tickets. The article says one thing, the product does another, and an agent spends the afternoon explaining the difference to six people. The check itself is comparison work: what shipped, what the article claims, where they disagree. That is a task with a shape, so it can be owned. ## The three kinds of stale Worth separating, because they need different responses. - **Factually wrong** — The article describes behaviour the product no longer has. This one costs you tickets today and goes first in the delivery. - **Incomplete** — The behaviour is right and something new is undocumented, so customers cannot find a feature you built. - **Structurally stale** — Six articles now overlap because the product grew into them, and the customer cannot tell which one to read. ## How the worker checks | Source | What it reads | What it can conclude | | --- | --- | --- | | Linear or GitHub | Shipped issues and merged changes since the last run | What changed, and when | | Your help articles | The claims each article makes | Which claims are now contradicted | | Support tasks | Tickets tagged as documentation confusion | Which pages customers actually stumble on | | Web search | Your own public pages | Where a public page disagrees with the help center | > **Drafts go through file review** > > Docs in Polaris are versioned and support file review and comments, so a rewritten article arrives as a proposed version with a history rather than an overwrite. The person who approves it is the person who answers for what customers read. ## Questions people ask **Does it need access to our codebase?** GitHub and Linear are both in the connection catalog, and either is enough to see what shipped. A worker with only your help articles and support tickets can still find contradictions, but it will find fewer of them and later. **Can it publish the corrected article?** No. The delivery is a comment with the rewritten draft attached, and a person publishes. Documentation is customer-facing text, so it goes through the same approval as any other customer-facing text. **How often should this run?** After each release if you ship in batches, monthly if you ship continuously. Attach it to the release task so the check happens when the risk is created rather than on a calendar that drifts. **What does a monthly documentation audit cost?** It is billed in human-equivalent hours at about $2 each, and the formula counts observable effort including the drafts produced. A month with a large release costs more than a quiet one, and the work log shows exactly which session was which. ## Related - https://www.polarishq.co/use-cases/customer-support - https://www.polarishq.co/use-cases/customer-support/response-templates - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/github - https://www.polarishq.co/glossary/workstream --- --- title: "Escalation Tracking, Support to Engineering | Polaris" description: "An AI worker keeps every open escalation visible: what changed in Linear, what has had no update, and who is still waiting. Delivered daily as a comment." url: https://www.polarishq.co/use-cases/customer-support/escalation-tracking section: Use cases updated: 2026-08-21 --- # Escalations that do not go quiet after the handoff The customer's question is not what the bug is. It is whether anyone is still looking at it. ## The short answer Escalation tracking in Polaris means an escalation workstream where each customer issue is a task, linked to the engineering issue in Linear or GitHub. An AI worker posts a daily status comment: what moved, what has had no update past your threshold, and who is waiting on a reply. Support decides what to promise the customer; the worker makes sure nothing goes silent. - **Connections:** Linear · GitHub · Slack - **Cadence:** Daily status comment - **Lives in:** One escalation workstream, Focus-pinned ## The handoff is where trust is lost A support agent escalates a bug, engineering picks it up, and the thread ends. Two weeks later the customer asks for an update and the honest answer is that nobody knows. Not because anyone is hiding anything, but because the escalation existed in a Slack thread that scrolled away. The fix is boring and it works: every escalation is a task, in one workstream, with a customer attached and a status somebody is responsible for refreshing every day. ## What the daily comment says - **Moved** — Escalations where the linked engineering issue changed state, with what it changed to. - **Silent** — Escalations with no update longer than your threshold, named with the number of days and the engineer who owns them. - **Waiting on us** — Issues resolved in engineering where nobody has told the customer yet, which is the most embarrassing category and the easiest to fix. - **Ageing** — Open escalations sorted by how long the customer has been waiting rather than by severity, because those are different lists. ## Two ways an escalation ends **In a chat thread** - Reported once, discussed for an afternoon - No owner after the discussion - Status is whatever the last message said - The customer chases you for it **As a tracked task** - One task per customer issue, linked to the engineering issue - An owner and a date, like all work in Polaris - A daily status comment from the worker - Nothing resolved in engineering stays untold > **Signals become suggestions** > > When an escalation starts life as a Slack message, the Polaris Inbox turns it into a prefilled task suggestion with bucket, lane, labels and owner already set, waiting for one click. That is the moment escalations usually get lost, and it is the moment worth catching. ## Questions people ask **Does this replace our issue tracker?** No. Linear and GitHub are both connections, and the engineering issue stays where engineers work. The escalation task in Polaris is the customer-facing side of it: who is waiting, what they were told, and when. **Can the worker update the engineering issue?** Its deliverable is a status comment for a person to act on. Changing state in an issue tracker is a decision an engineer or a support lead makes, and Polaris keeps that boundary. **What threshold should we use for silent?** Start with two working days and adjust in the skill file once you see how noisy that is. The threshold is a line in a document, so tuning it takes a minute and leaves a history. **How do we see the history of one escalation?** It is a task, so the comments, the deliveries and the activity feed hold the whole record in one place, including which person closed it and how the delivery was rated. ## Related - https://www.polarishq.co/use-cases/customer-support - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/engineering/bug-triage - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/focus-lane --- --- title: "Turning Customer Feedback into Product Tasks | Polaris" description: "An AI worker clusters support conversations into themes with counts and quotes, then opens a task for the top ones so product sees evidence, not anecdotes." url: https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops section: Use cases updated: 2026-08-21 --- # Customer feedback that reaches product as evidence Support already knows what is wrong with the product. The problem is the format the knowledge arrives in. ## The short answer Customer feedback loops in Polaris run as a monthly task assigned to an AI worker with access to your support conversations. The worker clusters what customers said into themes, delivers each theme with a count and verbatim quotes, and proposes a task in the product workstream for the largest ones. Product decides what to build; support stops arguing from memory. - **Connections:** Gmail · Slack · WhatsApp - **Cadence:** Monthly theme report - **Output:** Themes with counts, quotes and proposed tasks ## Anecdote loses to roadmap every time A support lead says customers keep complaining about the export. A product manager hears one person's impression against a roadmap built from a quarter of planning. The impression loses, correctly, because it has no weight behind it. The same claim with a number and eleven quotes attached is a different conversation. The work of producing that number is reading three hundred conversations, which is exactly why it never happens. ## How the monthly report is built 1. **Sources are named in the skill file** — Which channels count as customer feedback: the support address, a shared Slack channel, WhatsApp conversations with named accounts. 2. **The worker clusters** — A cloud machine wakes for the task and groups conversations by what the customer was trying to do, not by the words they used. Two people describing the same failure differently belong in one theme. 3. **Each theme gets evidence** — How many conversations, over what period, with verbatim quotes attached rather than paraphrase. A theme with three mentions is reported as three. 4. **The top themes become proposals** — Prefilled task suggestions for the product workstream, each carrying the count and the quotes into the task itself. 5. **Product answers in the same place** — The decision lands as a comment on the task, so support can tell a customer what happened without asking anyone. ## Rules that keep the report honest - **Count conversations, not mentions** — One customer who wrote nine times about one problem is one conversation, otherwise the loudest account sets your roadmap. - **Quote, do not summarise** — A verbatim sentence from a customer survives the meeting. A paraphrase becomes someone's opinion by the second retelling. - **Report the boring themes** — The theme nobody feels strongly about is often the one costing the most support hours. - **Say what did not come up** — Silence about a feature you just shipped is information too. > **The loop closes on the board** > > Because support tasks, product tasks and the worker's delivery comments live in one product, a customer complaint from March and the decision made in April are two comments on the same object. Nobody has to reconstruct the story from four tools. ## Questions people ask **How does it handle private customer data?** Connections are authorised once for the organisation and credentials are stored server-side; workers use them and browsers cannot read them back. Beyond that, decide in the skill file which channels a worker may read at all. **Can it survey customers?** It reports on conversations that already happened. Asking customers something new is outbound contact, and outbound contact is sent by a person. **Does product have to use Polaris for this to work?** The proposed tasks land in a Polaris workstream, so the product side of the loop lives there. If product works in Linear, the connection lets a worker read that context, but the evidence and the decision are easiest to keep together on one board. **How far back can the first report go?** As far as the connected channels retain, though sessions are clamped at eight hours, so a large backlog is best split into several tasks. Each session is itemised on the work log. ## Related - https://www.polarishq.co/use-cases/customer-support - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/product/user-research-synthesis - https://www.polarishq.co/ai-workers/customer-success-manager - https://www.polarishq.co/integrations/whatsapp - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/delivery-comment --- --- title: "Support Response Templates, Maintained by AI | Polaris" description: "An AI worker drafts and maintains support response templates in versioned docs, built from replies your team actually sent and approved before anyone uses them." url: https://www.polarishq.co/use-cases/customer-support/response-templates section: Use cases updated: 2026-08-21 --- # A template library that stays in your team's voice Templates go stale the same way documentation does, except a stale template gets sent to a customer. ## The short answer Support response templates in Polaris live in versioned docs maintained by an AI worker. The worker drafts new templates from replies your team already sent, rewrites ones that no longer match the product, and flags any template contradicting a help article. A support lead approves each version in file review, and agents use them as a starting point rather than a script. - **Connection:** Google Drive - **Lives in:** Docs, versioned, with file review - **Approval:** Support lead, per version ## Why template libraries decay The library starts as twelve good replies. Then a plan is renamed and four of them are wrong, an agent writes a better version of one and keeps it in a personal note, and a new hire finds the original because it is the one in the shared folder. The library is not maintained because maintaining it produces nothing visible. Nobody thanks you for a template that was already fine. ## What the worker does with the library - **Promotes what works** — Reads the replies agents actually sent, finds the ones sent repeatedly, and drafts them into templates instead of leaving them in one person's habits. - **Rewrites what is wrong** — Any template naming a plan, a limit or a screen that changed comes back rewritten, with the change explained. - **Removes duplicates** — Three templates for the same situation means three different answers to the same customer question. - **Checks against the help center** — A template that contradicts a published article is worse than no template, and it is a specific thing to look for. ## What belongs in a template and what does not | Part of the reply | Template it | Why | | --- | --- | --- | | The explanation of what happened | Yes | It is the same every time and it needs to be accurate | | Links to the right article | Yes | Agents should not be hunting for the URL | | The apology | No | A pasted apology reads exactly like a pasted apology | | Refund or credit amounts | No | That is a decision, and it has a boundary a person owns | | The next step | Yes, with a blank | The shape is fixed, the specifics are not | > **Templates are a floor, not a script** > > The point of the library is that an agent never starts from a blank box on a Tuesday afternoon. It is not that every customer gets the same paragraph. Put that in the skill file, and reject drafts that read like they were written to be sent unedited. ## Questions people ask **Where do the templates live?** In Docs in Polaris, in a nested tree with the rest of your support material. Files are versioned and support review and comments, so each template has a history and an approver. **Can the worker use the templates to answer tickets?** It can draft a reply using them during triage, and that draft still arrives as a comment for an agent to approve and send. The library and the drafting are two different tasks. **How do we keep the voice consistent as the team grows?** The tone rules live in the worker's SKILL.md, which is a file anyone on the team can read. A new agent can see what the voice is meant to be rather than inferring it from whichever old thread they found first. **What triggers a template review?** Attach the task to your release cadence, so that anything that changes the product also queues a check of what your templates say about it. ## Related - https://www.polarishq.co/use-cases/customer-support - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/customer-support/help-center-maintenance - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/glossary/skill-file --- --- title: "Polaris for Founders and Executives | AI Workers" description: "Weekly business reviews, board packs, OKR tracking and research assembled by AI workers, so the leadership hour is spent deciding rather than collecting." url: https://www.polarishq.co/use-cases/executive section: Use cases updated: 2026-08-21 --- # Running the company with an AI teammate on the reporting The information you need to run the company exists. Assembling it every week is what nobody has time for. ## The short answer Polaris gives founders and executives AI workers that assemble the reporting layer. A worker is hired in chat, given a skill file with your metric definitions, and connected to Stripe, HubSpot, Linear or web search. It builds the weekly review, the board pack, the OKR roll-up and research memos, each delivered as a comment with the file attached. Leadership keeps the commentary and the decisions. - **Typical connections:** Stripe · HubSpot · Linear · web search - **Software cost:** $0, unlimited people - **Work cost:** ~$2 per human-hour delivered - **Runs:** Overnight, on a cloud machine ## Two hours of collecting for twenty minutes of thinking Every leadership meeting has a preparation cost, and somebody pays it: pulling revenue from one place, pipeline from another, shipping status from a third, then formatting all of it into something readable before Monday. The collecting is deterministic. The thinking is not. Handing the first part to a worker does not make the second part easier, but it means the second part happens with the whole picture in front of you instead of the half somebody had time to assemble. In small companies the person paying the preparation cost is usually the founder, at eleven at night. ## The five executive jobs | Job | Connections | Delivered as | Leadership keeps | | --- | --- | --- | --- | | Weekly business review | Stripe · HubSpot · Linear | One brief before the meeting, same shape weekly | What the numbers mean | | Board reporting | Stripe · HubSpot · Google Drive | A drafted pack with the data assembled | The narrative and the ask | | OKR tracking | Linear · Slack | A roll-up naming every key result with no movement | Whether to change the objective | | Strategic research | web search | A sourced memo, usually overnight | The decision | | Meeting follow-ups | Google Calendar · Slack | Every commitment as an owned, dated task | What was actually agreed | ## Why a Chief of Staff ships with every org Polaris itself is on your roster from day one, before you hire anyone. - **It keeps work owned and dated** — Unowned tasks are the leading indicator of a plan that will not happen, and the copilot names them rather than waiting for you to notice. - **It builds workstreams** — When a pile of related work has no home, it creates the container instead of leaving it in your head. - **It hires what is missing** — You describe the gap in chat and the interview produces a worker with a skill file, in about a minute. - **It turns noise into approved tasks** — Slack signals arrive in the Inbox as prefilled suggestions rather than becoming tasks silently. ## What leadership overhead costs here - **$0** — For the software, at any headcount. No seats, no tiers - **~$2** — Per human-equivalent hour delivered. Itemised on the work log - **≈12×** — Faster than the human clock it replaces. From the recorded delivery demo > **Every hour on the bill is challengeable** > > Human-equivalent hours come from an open formula based on observable effort: pickup time, searches, finished prose, checklist items, comments addressed and files produced, clamped between five minutes and eight hours per session. Each job appears on the worker's work log, and you can dispute a line from the log itself. That is the answer to how you know the hours are real. ## The five executive jobs in detail - [use-cases/executive/weekly-business-review](https://www.polarishq.co/use-cases/executive/weekly-business-review) - [use-cases/executive/board-reporting](https://www.polarishq.co/use-cases/executive/board-reporting) - [use-cases/executive/okr-tracking](https://www.polarishq.co/use-cases/executive/okr-tracking) - [use-cases/executive/strategic-research](https://www.polarishq.co/use-cases/executive/strategic-research) - [use-cases/executive/meeting-follow-ups](https://www.polarishq.co/use-cases/executive/meeting-follow-ups) ## Questions people ask **Is a worker allowed to see revenue data?** Stripe is one of the connections in the catalog and you decide whether to authorise it. Connections are org-wide, verified live and stored server-side, so workers use the credentials and browsers cannot read them back. **Can it write the board narrative?** It can draft one from the assembled data, and you should rewrite it. A board pack is an argument a founder makes about their own company, and the reason the machine never closes its own task applies most strongly here. **How is this different from a BI dashboard?** A dashboard shows the numbers. A worker assembles the numbers, writes the comparison to last week, names what moved and why it might have, and delivers it as a document on a task where the discussion happens. **What does an executive worker cost per month?** There is no subscription. Billing is about $2 per human-equivalent hour of delivered work, so a weekly brief and a monthly research memo cost what those deliveries measure, itemised job by job. ## Related - https://www.polarishq.co/use-cases - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/use-cases/executive/board-reporting - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/for/solo-founders --- --- title: "Weekly Business Review, Assembled by an AI Worker | Polaris" description: "An AI worker pulls revenue, pipeline and shipping status into the same brief every Monday and delivers it as a comment, so the meeting starts at the discussion." url: https://www.polarishq.co/use-cases/executive/weekly-business-review section: Use cases updated: 2026-08-21 --- # The weekly review, written before the meeting starts A meeting that begins with everyone reading the same document is a different meeting. ## The short answer A weekly business review in Polaris is a recurring task assigned to an AI worker connected to Stripe, HubSpot and Linear. Each Monday the worker assembles the same brief: revenue and its change, pipeline movement, what shipped, and what slipped, with each number sourced. It delivers the brief as a comment before the meeting. Leadership supplies the interpretation and the decisions. - **Connections:** Stripe · HubSpot · Linear - **Cadence:** Weekly, delivered before the meeting - **Format:** Identical structure every week ## Same shape, every week, on purpose The value of a weekly review comes from comparison, and comparison requires that the document does not change shape. When one week's brief has four sections and the next has seven, nobody can tell what moved. This is precisely what a worker with a fixed skill file is good at and what a busy human is bad at. The person assembling the brief improvises depending on what they had time for. The worker produces the same sections in the same order whether the week was calm or terrible. ## The sections worth fixing in the skill file - **The number and its delta** — Revenue this week against last week and against the same week last month, each sourced from the connection it came from. - **Pipeline movement** — Deals that changed stage in either direction, and what the movement does to the month. - **Shipped** — What actually closed in Linear or GitHub, named, rather than a count of issues. - **Slipped** — Commitments from last week's brief that did not land, carried forward with their age. This section is the one that makes the review honest. - **Unowned** — Work in the leadership workstreams with no assignee, which is usually where next month's slip is already visible. ## How Monday morning goes 1. **Sunday evening: the task fires** — A cloud machine claims the job. Nobody has to be online for this, which is the point of running workers on a machine rather than a laptop. 2. **It reads the sources** — Each connection you authorised, in the order set in the skill file, ticking acceptance criteria as sections complete. 3. **It delivers before the meeting** — One comment on the recurring task, with the document attached and anything it could not read stated plainly rather than left blank. 4. **Everyone reads first** — The brief is a comment on a shared task, so there is no distribution step and no version confusion. 5. **The meeting starts at the argument** — Decisions taken become tasks in the same workstream, owned and dated, and appear as slipped or delivered in next week's brief. > **The worker reports; it does not explain** > > A machine can tell you revenue fell 8% and that two deals slipped. It cannot tell you that both deals slipped because of the same broken onboarding step. Keep the interpretation section human and empty until the meeting fills it. ## Questions people ask **What if a data source is unavailable?** The delivery says so. A section marked as unread is more useful than a number the worker inferred, and stating what it could not verify is part of how deliveries are written. **Can different teams get their own version?** Yes. Assign the worker one recurring task per team, each with its own acceptance criteria, and each delivery is a separate comment with its own history. **Does this need us to move our data into Polaris?** No. The worker reads through connections you authorise once for the organisation, and Stripe, HubSpot and Linear stay where they are. **How much does a weekly brief cost?** It is billed on human-equivalent hours at about $2 each, calculated from observable effort per session. A brief is a short session, and the line appears on the worker's work log where you can challenge it. ## Related - https://www.polarishq.co/use-cases/executive - https://www.polarishq.co/use-cases/executive/okr-tracking - https://www.polarishq.co/use-cases/executive/board-reporting - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/glossary/work-log --- --- title: "Board Reporting with an AI Worker | Polaris" description: "An AI worker gathers the quarter's numbers, builds the recurring sections of the board pack and delivers it as a file. The founder writes the argument." url: https://www.polarishq.co/use-cases/executive/board-reporting section: Use cases updated: 2026-08-21 --- # Board packs where the assembly is not your weekend A board pack is 80% data you already have and 20% the story only you can tell. The 80% is what eats the week. ## The short answer Board reporting in Polaris means assigning the pack's assembly to an AI worker connected to Stripe, HubSpot and your docs. The worker collects the quarter's numbers, builds the recurring sections against last quarter, notes what changed in the metric definitions, and delivers a draft as a file on the task. The founder writes the narrative, the asks and anything a director will push back on. - **Connections:** Stripe · HubSpot · Google Drive - **Delivered as:** A generated file on the task - **Written by a human:** Narrative, asks, risks ## The week before the board meeting It is the same week every quarter. Numbers from four places, a deck rebuilt from last quarter's, a metric that was defined differently in March, and a founder writing the narrative at midnight because the assembly took four days. The assembly is the part with no judgement in it. Handing it over does not mean handing over the meeting: it means arriving at the narrative with the data already correct and the definitions already checked. ## Where the line sits in a board pack **The worker assembles** - Revenue, growth and the quarter-over-quarter comparison - Pipeline and conversion, pulled from the CRM - Product shipped against what was committed last quarter - Headcount and the hiring plan as it stands - The appendix nobody enjoys building **The founder writes** - What the quarter actually means - The risk you would rather not put in writing, in writing - The ask, and what happens if the answer is no - Anything a director will challenge ## Acceptance criteria for a board draft Set these on the task; the worker ticks them as it goes. - **Every number sourced** — Each figure names the connection and the date it was read. A board number with no provenance is a question you will be asked live. - **Definitions restated** — How each metric is calculated, in the pack, so a change in definition cannot be mistaken for a change in performance. - **Last quarter's commitments** — What was promised at the previous meeting, and whether it happened. Directors remember; packs frequently do not. - **Gaps stated** — Anything the worker could not verify appears as an open item rather than as a confident figure. > **Do not let a machine phrase your risks** > > A drafted risk section reads like every drafted risk section, and a board can tell. Write that part yourself, in your own words, including the sentence you would rather not write. That is the part of the pack that earns trust. ## The reporting layer around it - [use-cases/executive/weekly-business-review](https://www.polarishq.co/use-cases/executive/weekly-business-review) - [use-cases/executive/okr-tracking](https://www.polarishq.co/use-cases/executive/okr-tracking) - [use-cases/executive/strategic-research](https://www.polarishq.co/use-cases/executive/strategic-research) ## Questions people ask **Can it produce the deck itself?** Deliveries can include generated files, including documents and PDFs, attached to the task. Treat the output as the assembled draft, then take it into whatever format your board expects. **How do we handle sensitive financials?** Decide which connections a worker is given. Credentials are verified live and stored server-side, so a worker uses them and browsers cannot read them back, but the narrower question of which data a worker may read is yours to set. **Does it remember last quarter's pack?** It reads the previous pack from your docs, which is how the comparison and the commitment check are produced. Keeping the packs in Docs, versioned, makes each quarter cheaper than the last. **What does assembling a quarterly pack cost?** It is billed in human-equivalent hours at about $2 each, and a quarterly assembly is a longer session than a weekly brief. Sessions are clamped at eight hours, and every line is itemised on the work log where you can challenge it. ## Related - https://www.polarishq.co/use-cases/executive - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "OKR Tracking Without the Spreadsheet | Polaris" description: "Objectives become workstreams and key results become dated tasks. An AI worker posts a fortnightly roll-up naming every key result that has not moved." url: https://www.polarishq.co/use-cases/executive/okr-tracking section: Use cases updated: 2026-08-21 --- # OKRs that stay visible after the offsite OKRs do not fail at the writing. They fail in week three, when nobody has looked at them since the offsite. ## The short answer OKR tracking in Polaris means each objective is a workstream and each key result is a task with an owner and a date. An AI worker posts a fortnightly roll-up as a comment: which key results moved, which have had no activity, and which are dated to complete in a window that no longer looks realistic. Leadership decides what to cut, and the cut is recorded. - **Connections:** Linear · Slack - **Structure:** One workstream per objective - **Cadence:** Fortnightly roll-up comment ## The quarter-three problem Objectives are written with real care in week zero, live in a document for two weeks, and then compete with the actual work. By week six the document is a historical artifact and the team is busy with something else, which may well be the right something else. The failure is not the change of direction. The failure is that nobody decided to change direction. Putting key results on the same board as the work makes the competition visible. If a key result has had no task activity in three weeks, the company has already deprioritised it, and somebody should say so out loud. ## How OKRs map onto the board | OKR concept | In Polaris | Why | | --- | --- | --- | | Objective | A workstream | It is a container of work, which is what a workstream is | | Key result | A task with an owner and a date | An unowned key result is a wish | | Initiative | Tasks in the same workstream | The work and the measure sit together | | Check-in | A recurring task assigned to a worker | The check-in happens whether or not anyone remembers | | This quarter's focus | The Focus lane | Pinned first in every view, does not scroll away | ## What the roll-up names - **Moved** — Key results with task activity since the last roll-up, with what changed. - **Static** — Key results with no activity past your threshold, named with the owner and the number of days. This is the whole point of the report. - **Unowned** — Key results with no assignee, which happens more often than any leadership team expects. - **Arithmetically doomed** — Key results whose remaining work will not fit the remaining days, based on the dates already on the tasks. > **Cutting an objective is a decision, not a failure** > > The useful output of tracking is not a percentage. It is a fortnightly moment where leadership either recommits to a key result or drops it on purpose. Record the drop as a comment on the task so that next quarter's planning starts from what actually happened. ## Questions people ask **Do we need a separate OKR tool?** Not for this shape. Objectives are workstreams, key results are tasks, and the roll-up is a delivery comment. The advantage is that the measure and the work sit in the same product, so the roll-up reads real activity rather than a self-reported percentage. **How does the worker know if a key result moved?** It reads task activity in the objective's workstream, plus any connected sources such as Linear. Movement means work happened, not that somebody updated a confidence score. **Can it score progress as a percentage?** It can report what the tasks say, but a percentage from a machine invites false precision. The static list and the dates are the parts leadership actually acts on. **How often should the roll-up run?** Fortnightly is frequent enough to catch a stalled key result while there is still quarter left, and rare enough that people read it. Weekly roll-ups on a quarterly cycle become wallpaper. ## Related - https://www.polarishq.co/use-cases/executive - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/use-cases/executive/meeting-follow-ups - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/glossary/focus-lane --- --- title: "Strategic Research Memos, Delivered Overnight | Polaris" description: "Assign a research question at six in the evening. A cloud machine works the open web overnight and the sourced memo is a comment on the task by morning." url: https://www.polarishq.co/use-cases/executive/strategic-research section: Use cases updated: 2026-08-21 --- # Research memos waiting for you in the morning The questions worth researching are the ones you never have a free afternoon for. ## The short answer Strategic research in Polaris means assigning a question to an AI worker with the web search connection. A cloud machine wakes for the task, runs a live tool loop across the open web, ticks the acceptance criteria you set, and delivers a memo as a comment with the file attached and every claim linked. The work continues after you close your laptop, and a person decides what to do with the answer. - **Connection:** web search - **Runs:** On a cloud machine, after you log off - **Delivered as:** A sourced memo file on the task ## The questions that never get answered Should we enter this adjacent market. What are the three companies doing that we keep hearing about in deals. What does regulation in that country actually require. How do other companies in this category price. Each of these is answerable with several hours of careful public reading. Each of them loses every day to something urgent, which is why founders make these decisions on impressions rather than on reading. This is the strongest case for a cloud agent rather than one on your laptop: the work happens during the hours you were never going to spend on it anyway. ## A recorded research delivery The competitor-pricing memo in the Polaris delivery demo. A real machine session with live web research, sped up rather than staged. - **~7 min** — Machine time, end to end - **$2.80** — Billed for the delivery - **~1h24m** — The human clock it replaced. About $70 at $50 an hour ## How to write a research task that comes back useful - **Ask a decidable question** — How three named competitors price their team tier produces evidence. What is the future of our market produces an essay. - **Name the sources you trust and distrust** — In the skill file, once. Whether vendor blogs count, whether you want primary regulatory text, what to do with a number that appears only in a press release. - **Set acceptance criteria** — Every claim linked, contradictions between sources reported rather than resolved silently, an explicit list of what could not be found. - **Ask for the argument against** — A memo that only supports the direction you already lean is worse than no memo. > **Assign it and close the laptop** > > Tasks assigned to a Polaris worker are claimed off a queue by a runtime on a cloud machine. That machine keeps working while you sleep, posts progress as it goes, and the delivery is a comment waiting in the morning. The difference between this and an agent on your laptop is the machine, not the model. ## Questions people ask **How do I know the memo is not invented?** Every claim carries a link, and the acceptance criteria include an explicit list of what the worker could not verify. Reading the sources for the two claims that matter most is a five-minute check, and it is the right check to keep doing. **Can it read documents we already have?** Yes, if the material is in Docs in Polaris or in a connected Google Drive. A memo that combines public research with your own numbers is usually the more useful one. **How long can a single research task run?** Sessions are clamped between five minutes and eight hours, so a very large question is best split into several tasks with narrower scopes. That also produces better memos. **What does a research memo cost?** About $2 per human-equivalent hour, computed from observable effort including how many searches ran and how much finished prose was produced. The recorded demo memo was billed at $2.80, and every job is itemised on the work log. ## Related - https://www.polarishq.co/use-cases/executive - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/use-cases/sales/lead-research - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/glossary/human-equivalent-hours --- --- title: "Meeting Follow-Ups That Become Owned Tasks | Polaris" description: "An AI worker turns meeting notes into proposed tasks with owners and dates, then names the commitments from previous meetings that never moved at all." url: https://www.polarishq.co/use-cases/executive/meeting-follow-ups section: Use cases updated: 2026-08-21 --- # Every commitment from the meeting, owned and dated The decisions were good. The problem is the eleven commitments that existed only in a document nobody reopened. ## The short answer Meeting follow-ups in Polaris start from notes in Docs. An AI worker with the Google Calendar connection reads the notes, proposes one task per commitment with an owner and a due date, and delivers them as prefilled suggestions in the Inbox for one-click acceptance. It also reports commitments from earlier meetings that never moved, which is the part everyone avoids. - **Connections:** Google Calendar · Slack - **Output:** One proposed task per commitment - **Also reports:** Older commitments with no movement ## Notes are not follow-up A good note-taker produces a document. A document is a record, not a mechanism. Nothing in it has an owner, a date, or a place on anybody's board, so the commitments depend entirely on each person remembering their own. The conversion step, reading the notes and creating eleven tasks with the right owners, takes twenty minutes and is nobody's job. So it happens after important meetings and not after ordinary ones, which is where most commitments live. ## From notes to a board 1. **Notes land in Docs** — Written by whoever was in the room, in the nested doc tree next to the workstream the meeting belongs to. 2. **The task fires after the meeting** — The worker reads the notes and the calendar entry, which tells it who was present and therefore who can own something. 3. **Commitments become suggestions** — One prefilled task per commitment, with bucket, lane, owner and a due date already set. Anything the notes left ambiguous is flagged as needing an owner rather than assigned to a guess. 4. **One click each** — Whoever ran the meeting accepts, corrects or rejects. Ten seconds a task, and nothing entered the board unapproved. 5. **The next report includes the ghosts** — Commitments accepted three meetings ago with no activity since are listed by name, which is the mechanism that makes the whole loop work. ## Two versions of the same meeting **Notes only** - Eleven commitments in a document - Owners implied by whoever was talking - No dates - Rediscovered when the same topic comes back in six weeks **Notes plus a follow-up task** - Eleven proposed tasks, each with an owner - Dates that appear in the Focus lane - Ambiguous commitments flagged rather than assigned - Old commitments that never moved, named out loud > **Suggestions, not silent tasks** > > Polaris deliberately keeps machine-generated work in the Inbox as prefilled suggestions until a person accepts them. A meeting produces plenty of statements that sound like commitments and are not, and the one-click approval is where that gets sorted out. ## Questions people ask **Does Polaris transcribe the meeting?** No. It works from notes in Docs and the calendar entry. Record the meeting however you already do, and put the notes where the worker can read them. **What if the notes are vague about who owns something?** The worker flags it as needing an owner instead of assigning it to whoever spoke last. An incorrectly assigned task is worse than an unassigned one, because it looks handled. **Can this run for every meeting?** It can, and the cost tracks the volume rather than a seat count. Start with the recurring meetings that generate commitments, because those are where things quietly disappear. **Who sees the list of commitments that never moved?** It is a comment on a task in the workstream, so everyone in that workstream sees it. Making it visible rather than private is what stops the same commitment being re-agreed three times. ## Related - https://www.polarishq.co/use-cases/executive - https://www.polarishq.co/use-cases/executive/okr-tracking - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/glossary/workstream --- --- title: "Finance Use Cases for AI Workers | Polaris" description: "Five finance jobs an AI worker prepares in Polaris: invoice chasing, expense coding, month-end close, budget reporting, vendor renewals. You still sign off." url: https://www.polarishq.co/use-cases/finance section: Use cases updated: 2026-08-21 --- # Finance work with an AI worker on the roster The chasing, sorting and assembling that fills a finance week, prepared by a worker you brief once and review every time. ## The short answer Polaris gives a finance team an AI worker that prepares the repetitive half of finance work: chasing unpaid invoices, sorting receipts into categories, assembling the month-end checklist, drafting budget variance notes and tracking vendor renewals. The worker reads Stripe, Gmail and Google Drive through connections authorized once for the org, and posts each result as a comment on the task. A qualified human reviews the numbers and signs off. - **Connections used:** Stripe · Gmail · Google Drive - **Software cost:** $0 - **Work cost:** ~$2 per human-hour delivered - **Who closes the task:** You, never the machine ## Finance work is mostly retrieval, and retrieval is what a worker is good at Look at where a finance week actually goes. Someone opens Stripe, opens a mailbox, opens a folder of PDFs, and moves facts between them until a number agrees with another number. Then someone qualified decides what the number means. An AI worker in Polaris does the first part. It has the same task, the same comment thread and the same due date as a human teammate, because humans and AI workers are the same kind of record in Polaris. It reads the connections you gave it, works through a checklist you wrote, and posts the result back as a comment with the files attached. It does not sign anything. It does not close its own task. The delivery arrives as a comment and stays open until a person reads it and closes it. ## The five finance jobs covered here Each one has its own page with the brief, the connections and the review step written out. - **Invoice tracking** — An aged list of who owes what, matched against Stripe, with a reminder email drafted per account and none of them sent. - **Expense categorization** — Receipts pulled out of Gmail, coded against your own chart of accounts, with anything ambiguous escalated instead of guessed. - **Month-end close** — The close checklist assembled and chased across the people who owe items, so week two stops being a surprise. - **Budget reporting** — Plan against actuals, with the variances that exceed your threshold written up in prose a non-finance reader can follow. - **Vendor management** — Every contract's renewal date, notice window and current spend in one list, with the notice deadlines dated in the Focus lane. ## What each job needs and what comes back | Job | Connections | Delivered as | | --- | --- | --- | | Invoice tracking | Stripe, Gmail | Aged receivables list + drafted reminder per account | | Expense categorization | Gmail, Google Drive | Coded transaction file + an escalation list | | Month-end close | Google Drive, Slack, Docs | Live checklist with owners, plus a daily status comment | | Budget reporting | Google Drive, Stripe | Variance memo as a PDF, with the working shown | | Vendor management | Gmail, Google Calendar, Stripe | Vendor register + dated notice deadlines | ## The division of labor, written down This is not a hedge. It is how the product works: agents deliver, humans close. **The worker prepares** - Pulling records out of Stripe, Gmail and Drive - Matching payments to invoices and flagging what does not match - Applying your written coding rules to transactions - Chasing checklist owners and reporting who has not responded - Drafting the memo, the reminder, the summary - Ticking its own acceptance criteria as it goes **A qualified person decides** - Whether a treatment is correct - Anything with a tax consequence - Which late accounts get chased and how hard - Whether the books are ready to close - What the variance means and what to do about it - Closing the task and rating the work > **Polaris is not an accountant and does not give tax advice** > > An AI worker in Polaris prepares and presents information from the systems you connected it to. It has no professional qualification, no view on your jurisdiction, and no authority over your books. Anything with a tax, audit or statutory consequence needs a qualified human, and the work log exists so that person can see exactly what was done before they put their name on it. ## What this costs The software is free. You pay for work that was delivered. - **$0** — Polaris software, forever. Unlimited people, tasks, workstreams and docs - **~$2** — Per human-hour delivered. Itemized job by job on the worker's work log - **0** — Charge for a job that delivered nothing. Nothing delivered, nothing billed ## Finance use cases in detail - [use-cases/finance/invoice-tracking](https://www.polarishq.co/use-cases/finance/invoice-tracking) - [use-cases/finance/expense-categorization](https://www.polarishq.co/use-cases/finance/expense-categorization) - [use-cases/finance/monthly-close-checklist](https://www.polarishq.co/use-cases/finance/monthly-close-checklist) - [use-cases/finance/budget-reporting](https://www.polarishq.co/use-cases/finance/budget-reporting) - [use-cases/finance/vendor-management](https://www.polarishq.co/use-cases/finance/vendor-management) ## Questions people ask **Can an AI worker in Polaris close the books?** No. A worker can assemble the close checklist, chase the people who owe items and report what is still outstanding. Declaring a period closed is a judgment with accountability attached, and it stays with the person who holds that accountability. **Does the worker get access to our bank account?** Only if a bank is in the connection catalog, and it is not. The catalog is fixed: Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and web search. A finance worker typically gets Stripe, Gmail and Google Drive, and nothing else. **How do we know the hours on the bill are real?** Every job writes a work log entry with the observable effort behind it: how many searches it ran, how much finished prose it produced, how many checklist items and files. Hours are computed from that with a published formula and clamped between five minutes and eight hours. Any line can be challenged from the log itself. **Can two people review the same delivery?** Yes. A delivery is a comment on a task, so it sits in the thread where anyone with access can read it, reply and attach their own notes. The task stays open until a human closes it. **What happens if the worker gets a number wrong?** You see it before it matters, because the delivery is a draft in a comment rather than an action taken in your ledger. Reopen the thread, say what was wrong, and the correction becomes the next task. Persistent corrections belong in the worker's SKILL.md so the rule sticks. ## Related - https://www.polarishq.co/use-cases/finance/invoice-tracking - https://www.polarishq.co/use-cases/finance/monthly-close-checklist - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/use-cases/executive/board-reporting --- --- title: "AI Invoice Tracking and Chasing | Polaris" description: "Set up an AI worker in Polaris to match Stripe payments against invoices, age the receivables, and draft a reminder email per late account. You review and send." url: https://www.polarishq.co/use-cases/finance/invoice-tracking section: Use cases updated: 2026-08-21 --- # Invoice tracking that produces a chase list, not a dashboard A worker matches Stripe payments to what you invoiced, ages the gap, and drafts the reminder for each account. You decide who actually gets chased. ## The short answer An AI worker in Polaris tracks unpaid invoices by reading Stripe payments and invoice mail in Gmail, matching them, and posting a weekly aged list as a comment on the task: who owes what, how many days late, and a drafted reminder email per account. Nothing sends automatically. The finance owner reads the drafts, changes the tone for the accounts that matter, and sends them. - **Connections:** Stripe · Gmail - **Cadence:** Weekly recurring task - **Delivered as:** Comment with aged list + drafts ## The problem is not knowing who is late. It is writing five emails you do not want to write. Most teams can already tell you their receivables total. What they cannot do on a Friday afternoon is work through fourteen overdue accounts, remember which one already promised to pay on the 15th, which one is a friend of the founder, and which one has ignored two reminders and needs a harder note. So the chase slips a week. Then two. The receivables number is not a reporting problem, it is an unwritten-email problem. A worker is good at the part that stalls you: the matching, the ageing, and the first draft of every message. It is bad at knowing which client you cannot afford to annoy. So it writes all fourteen, marks the ones where the history looks complicated, and hands you a thread you can clear in ten minutes. ## Setting it up 1. **Hire the worker in chat** — Tell the Chief of Staff that invoices are going unchased. It runs a short interview, suggests a name and a role, and asks what the worker should be great at. Answers are option pills, so this takes about a minute. 2. **Connect Stripe and Gmail** — Both come from the fixed connection catalog and are authorized once for the whole org. Credentials are stored server-side. The worker uses them; nobody's browser can read them back. 3. **Write the escalation ladder into the SKILL.md** — Capabilities are a real skill file you can open and edit. Put your actual policy in it: friendly at 7 days, firm at 30, stop and escalate to a named human at 60. The worker follows what the file says. 4. **Create a recurring task with acceptance criteria** — Something like: every invoice over 7 days late appears in the list; each has a drafted email; any account with a payment plan in the mail thread is flagged rather than chased. The worker ticks these off as it works. 5. **Read the delivery and send** — A cloud machine wakes for the task, does the run, and posts the aged list and the drafts as a comment. You edit what needs editing, send from your own mailbox, and close the task. ## What the worker does at each stage of lateness These bands come from your SKILL.md. The example below is the default a lot of small teams start with. | Days late | Worker prepares | Human step | | --- | --- | --- | | 0–7 | Nothing. It stays off the list. | None | | 7–30 | A short reminder, invoice attached, polite | Skim and send | | 30–60 | A firmer note with the full history of prior contact | Read the history, decide the tone | | 60+ | No draft. A flagged summary of every contact attempt. | Decide: call, pause service, or write it off | | Any age, payment plan found in the thread | Flagged, not chased, with the promised date quoted | Decide whether the promise still holds | ## What it will get wrong, and how you find out - **Partial payments** — If a client pays 60% against one invoice and Stripe records it without a reference, the match is a guess. The worker marks guesses as guesses rather than presenting them as fact. - **Invoices raised outside Stripe** — It only sees what the connections show it. A PDF invoice someone sent from a personal mailbox is invisible until that mail is in the connected account. - **Credit notes and disputes** — A dispute lives in a conversation, not in a payment record. Tell the worker in the task thread and it will exclude the account from the next run. - **Currency** — Mixed-currency receivables should be stated in the acceptance criteria: which currency the list is denominated in and what rate source to cite. > **The worker never sends the email** > > Reminder emails carry your relationship with a customer. They are drafted in a comment on the task and sent by a person from their own mailbox. A worker that could send collection mail on its own would be a liability, not a feature. ## Questions people ask **Can the worker send the reminders automatically if we want it to?** The delivery model in Polaris is that agents deliver and humans close, so the drafted mail arrives as a comment for a person to send. That is deliberate. Sending collection mail without a human read is how a good customer relationship gets damaged over a bookkeeping error. **What does one weekly run cost?** It is billed on human-equivalent hours from the work log, at roughly $2 per hour, computed from the observable effort in the run: the records it read, the prose it wrote and the files it produced. A short weekly chase list is a small job, and a session is clamped at five minutes minimum and eight hours maximum. **Does this replace our accounting system?** No. Polaris does not hold your ledger and does not write to it. The worker reads Stripe and Gmail, prepares a list and drafts, and posts them on a task. Your accounting system stays where it is. **How do we handle a customer who should never be chased?** Name them in the SKILL.md with the reason. The skill file is plain text you edit directly, so exceptions are readable by anyone on the team rather than buried in a hidden prompt. ## Related - https://www.polarishq.co/use-cases/finance - https://www.polarishq.co/use-cases/finance/monthly-close-checklist - https://www.polarishq.co/use-cases/finance/vendor-management - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work --- --- title: "AI Expense Categorization for Small Teams | Polaris" description: "An AI worker codes receipts from Gmail and Drive against your chart of accounts, and hands back everything ambiguous as an escalation list instead of guessing." url: https://www.polarishq.co/use-cases/finance/expense-categorization section: Use cases updated: 2026-08-21 --- # Expense categorization that escalates instead of guessing A worker pulls receipts out of the mailbox, codes them against your own chart of accounts, and puts anything ambiguous in a pile for you rather than picking a category to look finished. ## The short answer An AI worker in Polaris categorizes expenses by reading receipt mail in Gmail and statement exports in Google Drive, matching each line to your own chart of accounts, and returning a coded file plus a separate escalation list of transactions it could not place with confidence. The rules live in an editable SKILL.md. A person reviews the escalations and approves the coding before anything reaches the books. - **Connections:** Gmail · Google Drive - **Rules live in:** An editable SKILL.md - **Output:** Coded file + escalation list ## Categorization tools fail on the 6% that matter Rules-based coding gets the easy lines. The airline is travel, the hosting bill is infrastructure, the coffee shop near the office is meals. That part was solved years ago. The lines that cost you time are the ones with no obvious home: a single payment covering three things, a subscription that is half marketing and half tooling, a reimbursement to a contractor with no receipt attached, a supplier whose legal name looks nothing like the brand you know. A classifier that assigns those a category with fake confidence has not saved you work. It has hidden work. The worker's job here is to split the pile honestly. Coded and confident on one side, escalated on the other, with the reason it stalled written next to each escalation. ## How it handles the awkward cases | Transaction shape | What the worker does | | --- | --- | | Clear match to a rule in the skill file | Codes it, cites the rule it applied | | Merchant name unrecognised | Runs a web search on the merchant, proposes a category, marks it as proposed | | One payment, several cost types | Escalates with the receipt attached and the suggested split | | No receipt found in the mailbox | Codes nothing. Lists it under missing documentation with the amount and date | | Amount above your review threshold | Escalates regardless of confidence | | Personal-looking spend on a company card | Escalates to the named human in the skill file. Never categorized silently | ## Where the boundary sits **The worker** - Finds and reads the receipts - Applies rules you wrote - Researches unknown merchants on the open web - Produces a coded file and a missing-documentation list - Explains every proposed category **Your bookkeeper or accountant** - Owns the chart of accounts - Rules on deductibility and on anything with a tax consequence - Resolves the escalations - Posts to the ledger - Closes the task ## Getting the rules right the first month The skill file is where the accuracy comes from. Write it like you would brief a new bookkeeper. - **Name your accounts exactly** — Paste your actual chart of accounts into the SKILL.md. A worker guessing at your account names is a worker generating rework. - **Give the ten merchants you argue about** — Every team has a handful of recurring lines that get miscoded. Write those down with the correct account and the reason. - **Set the escalation threshold in money, not in confidence** — Anything over an amount you would want to see personally goes in the escalation pile no matter how obvious it looks. - **Say who owns escalations by name** — The file names a human. Unowned escalations sit in a comment thread and rot. > **Categorization is not a tax opinion** > > Coding a line to an account is bookkeeping. Deciding whether it is deductible, in which jurisdiction, and under which rule is professional advice, and Polaris does not provide it. The worker prepares the file and shows its reasoning so a qualified person can decide quickly, not so they can skip deciding. ## Questions people ask **Can it write directly into our accounting software?** No. Accounting platforms are not in the Polaris connection catalog. The worker delivers a coded file as a comment on the task, and a person imports or posts it. That keeps a human between the classification and the ledger. **How does it learn our preferences over time?** By you editing the skill file. When a category is corrected, the corrected rule goes into the SKILL.md as a line anyone can read. There is no hidden model of your preferences drifting in the background. **What if a receipt is a photo rather than a PDF?** The worker reads what arrives in the connected mailbox and in Drive. Where it cannot extract a reliable amount or date, it lists the transaction under missing documentation rather than inventing values from a blurry image. **Is our financial data used to train a model?** Polaris stores your workspace data in its own Postgres database with row-level security, and connection credentials are held server-side where browsers cannot read them back. Treat the specifics as a question for the privacy policy at polaris-legal.pages.dev rather than for a marketing page. ## Related - https://www.polarishq.co/use-cases/finance - https://www.polarishq.co/use-cases/finance/invoice-tracking - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "AI-Assisted Monthly Close Checklist | Polaris" description: "Run month-end close in Polaris: an AI worker rebuilds the checklist, chases owners in Slack, reports blockers. A human declares the period closed." url: https://www.polarishq.co/use-cases/finance/monthly-close-checklist section: Use cases updated: 2026-08-21 --- # A monthly close that stops drifting into week two The close is a chase, not a calculation. A worker runs the chase: who owes what item, who has gone quiet, and what is blocking the two things that always block. ## The short answer In Polaris, a monthly close checklist is a task list with an AI worker assigned to keep it moving. The worker builds the checklist from the previous month, assigns each item to its named owner, chases people who have not delivered, and posts a daily status comment naming what is outstanding and who it is waiting on. The controller decides when the period is actually closed. - **Connections:** Slack · Google Drive · Gmail - **Cadence:** Daily during close week - **Lives in:** A close workstream, Focus lane pinned ## Close does not slip because the maths is hard Ask a controller why the close ran long and you will not hear about accruals. You will hear that the marketing lead never sent the agency invoice, that the payroll file landed on day four, and that nobody noticed one bank line was unreconciled until day six. Close is a coordination problem wearing an accounting costume. It has twenty to sixty items, eight owners, and one person doing the chasing while also doing their own items. That chasing is the part to hand over. The worker knows what the list was last month, who owned each line, and who has not answered. It posts the same status every morning so the blockers surface on day two instead of day six. ## Running a close in Polaris 1. **Make the close its own workstream** — A workstream is any container of work. Give the close one, with lanes for the stages you actually use. Lanes are shared between list and board view, so the controller can work in a list while everyone else looks at the board. 2. **Pin the deadline items to Focus** — Focus is a time-based lane across Today, This week and Next 30 days, pinned first in every view. The three items that always run late belong there, not buried in a list of forty. 3. **Assign the chase to a worker** — Assignment works the same for a worker as for a person. Give it the close workstream, Slack for the chasing and Google Drive for where the supporting files live. 4. **Write the acceptance criteria for the daily status** — For example: every open item named with its owner and days outstanding; every item that moved since yesterday listed; anything blocked stated with what would unblock it. The worker ticks these as it goes. 5. **Read the morning comment, unblock, then close** — The worker posts, you act. When the last item is done, a human closes the task. The machine never marks the close complete. ## What moves and what does not The worker changes the tempo of the close. It does not change who is responsible for it. **Handed to the worker** - Rebuilding this month's checklist from last month's - Assigning each item to the owner it had before - Chasing in Slack, on a schedule, without getting embarrassed about it - Reporting status every morning in the same shape - Collecting supporting files into the close folder **Stays with the controller** - Any accrual, estimate or cut-off judgment - Reviewing reconciliations - Deciding an item can be waived this month - Signing off the period - Closing and rating the task ## The shape of the job Sizes vary by team. These are the dimensions worth measuring on your own close before and after. - **3** — Focus horizons. Today, This week, Next 30 days - **1** — Status comment per day. Same shape every morning, so drift is visible - **$0** — Cost of the workspace. You pay only for the hours the worker delivers > **The worker cannot declare the period closed** > > A close carries a representation about the numbers. The machine posts progress, ticks criteria and delivers files, and it stops there. Marking the close done is a human action in Polaris by design, and the work log shows exactly what the worker did before you take that step. ## Questions people ask **Will the worker chase our CEO for a missing receipt?** It chases whoever is named as the owner of that item, in Slack, on the cadence you set. If you want a different escalation path for certain people, write it in the SKILL.md and the worker follows it. **Can it rebuild the checklist from last month automatically?** Yes, because last month's close lives in the same workspace as a set of tasks with owners and outcomes. That is the practical argument for keeping the close in the same tool as everything else rather than in a spreadsheet nobody can query. **What happens if nobody reads the morning comment?** The task stays open, which is the point. Deliveries are comments on tasks, so an unread status is visible as an open item with an assignee rather than a message that scrolled past in a channel. **Do we need every finance person to have a paid seat?** There are no seats. Polaris is free for unlimited people, tasks, workstreams and docs. The only thing billed is delivered work from AI workers, at roughly $2 per human-equivalent hour, itemized on the work log. ## Related - https://www.polarishq.co/use-cases/finance - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/use-cases/finance/invoice-tracking - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/integrations/slack --- --- title: "AI Budget Variance Reporting | Polaris" description: "An AI worker compares plan to actuals from Drive and Stripe, writes the material variances in plain prose, and delivers a PDF for finance to review." url: https://www.polarishq.co/use-cases/finance/budget-reporting section: Use cases updated: 2026-08-21 --- # Budget reports written for the people who did the spending Plan against actuals is easy to produce and hard to read. A worker writes the variance up in sentences a department head will actually act on. ## The short answer An AI worker in Polaris produces budget reporting by reading the actuals export in Google Drive alongside Stripe revenue, comparing them to the plan, and writing up every variance above your threshold in plain sentences. It delivers a PDF as a comment on the task with the working shown line by line. Finance reviews the numbers and decides what the variances mean before anything circulates. - **Connections:** Google Drive · Stripe - **Delivered as:** PDF attached to a delivery comment - **Cadence:** Monthly, or on demand ## The report that gets ignored is the one that only has numbers in it A department head opens a variance table, sees their line is red by eleven thousand, and has no idea whether that is the contractor they hired in week two or an invoice that landed a month late. So they ask finance. Finance answers the same four questions every month. The useful version of that report has a paragraph per material variance: what moved, the most likely reason based on what is actually in the transaction data, and what would need to be true for it to keep moving. Writing four paragraphs is a twenty-minute job that finance does after the numbers are already right, at the end of a long week, which is exactly why it does not get done. ## What the worker writes for each variance One block per line that breaches your materiality threshold. Lines inside the threshold get a single summary sentence, not a paragraph each. | Element | Source | Example of what it looks like | | --- | --- | --- | | The number | Actuals export, plan file | Marketing is $11.4k over plan for the month | | The composition | Transaction detail in the export | $9.2k of that is one agency invoice covering two months | | The check against revenue | Stripe | Revenue for the same period came in 4% under plan | | The open question | Written for the reader | Is the second month of that agency invoice already in next month's plan? | | What it will not say | Deliberately absent | No recommendation to cut a budget. That is not the worker's call | ## Making the report trustworthy - **Show the working** — The delivery includes the arithmetic behind each figure, so a reviewer can check a number in seconds rather than rebuilding it. - **Cite the source file and its version** — Docs in Polaris are versioned. A report that names the actuals export it read is a report you can reproduce next month. - **Set materiality explicitly** — Give a currency amount and a percentage in the acceptance criteria. Without one, the worker will either write about everything or about nothing. - **Keep opinions out of the draft** — Brief the worker to describe and to ask, not to recommend. Recommendations coming from a machine get either over-trusted or dismissed, and neither is useful. ## Who does what **The worker prepares** - Reading the plan and the actuals - Computing variance by line - Pulling the transaction detail behind material lines - Writing the narrative and generating the PDF - Flagging any line where the source data looked incomplete **Finance decides** - Whether the actuals are final - What a variance actually means - What gets said to the board or the bank - Whether a budget changes - When the report is fit to circulate ## Questions people ask **Where does the plan come from?** Wherever you keep it. In practice that is a file in Google Drive or a doc in Polaris. The worker reads what you point it at and names the file and version in the delivery so the report can be reproduced. **Can it forecast the rest of the year?** It can extend a trend and say so plainly, but a forecast is a set of assumptions somebody owns. Ask for the arithmetic and the stated assumptions in the delivery, then have a person decide whether the assumptions hold. **Does the PDF come out of Polaris or do we make it?** The worker produces it. Deliveries arrive as comments on the task and can include generated files, so a finished PDF is attached to the thread where the review happens. **How is this different from a BI dashboard?** A dashboard shows the variance. This writes the explanation next to it, in the same place the follow-up task will live. If you already have a dashboard that people read and act on, keep it, and use the worker for the narrative that nobody has time to write. ## Related - https://www.polarishq.co/use-cases/finance - https://www.polarishq.co/use-cases/finance/monthly-close-checklist - https://www.polarishq.co/use-cases/executive/board-reporting - https://www.polarishq.co/use-cases/data/reporting-automation - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "AI Vendor and Renewal Tracking | Polaris" description: "Keep a vendor register in Polaris: an AI worker reads contracts in Drive and charges in Stripe, and dates every notice deadline before it passes." url: https://www.polarishq.co/use-cases/finance/vendor-management section: Use cases updated: 2026-08-21 --- # Vendor management that catches the notice window Most money lost on vendors is lost by missing a cancellation deadline nobody had written down. A worker keeps the register and dates the deadlines. ## The short answer An AI worker in Polaris keeps a vendor register by reading contract files in Google Drive, renewal mail in Gmail and actual charges in Stripe, then dating each renewal and notice deadline into Google Calendar and the Focus lane. It posts the register as a comment on the task. A person decides what to renew, renegotiate or cancel before each deadline arrives. - **Connections:** Google Drive · Gmail · Stripe · Google Calendar - **Tracked per vendor:** Renewal date, notice window, real spend - **Reviewed:** Quarterly, or before each deadline ## Auto-renewal is a tax on not having a list A thirty-day notice window on an annual contract is a real deadline with a real cost, and it is almost never in anyone's calendar. It is in a PDF, in clause 11.2, in a Drive folder, signed by someone who has since left. The second failure is subtler. Teams track what they agreed to pay rather than what they are actually charged, and the two drift: a seat count that grew, an overage nobody watches, a plan that upgraded itself. A vendor worker fixes both by reading the contract and the charge, putting them side by side, and putting the notice date in a place with a person's name on it. ## The register the worker maintains | Field | Where it comes from | Why it earns its place | | --- | --- | --- | | Renewal date | The contract file in Drive | The thing everyone assumes is known and usually is not | | Notice window | The contract clause, quoted verbatim | The deadline that actually matters, and it is earlier | | Contracted amount | The contract file | What you agreed | | Actual charge | Stripe, or the receipt in Gmail | What you pay. The gap between these two is the finding | | Internal owner | You state it | An unowned vendor line never gets cancelled | | Last price change | Renewal mail in Gmail | Tells you whether a renegotiation is overdue | ## What the worker flags without being asked - **The gap between contracted and actual** — Where the charge does not match the agreement, the line is flagged with both figures and the source of each. - **A notice window closing inside 30 days** — It lands in the Focus lane under Today or This week, with the clause quoted so nobody has to reopen the PDF. - **A vendor with no internal owner** — Listed separately. These are the ones that quietly renew for three more years. - **Duplicate tools** — Two vendors billing for the same category is a question for a human, and the worker asks it rather than answering it. ## The boundary **The worker** - Reads contracts and quotes the clause - Reconciles agreed price against actual charge - Dates deadlines into the calendar and Focus - Drafts the cancellation or renegotiation email - Researches list pricing for alternatives on the open web **You** - Interpret a contract term that is genuinely ambiguous - Decide to renew, renegotiate or leave - Send anything that constitutes notice - Sign - Close the task > **Reading a contract is not interpreting one** > > The worker quotes the clause it found and names the file and page. It does not tell you what the clause means as a matter of law, and Polaris does not give legal advice. Where a term is ambiguous or the money is significant, the register exists to get the question in front of a lawyer earlier, not to answer it. ## Questions people ask **Can it cancel a vendor for us?** No. Serving notice is a contractual act. The worker drafts the notice email and dates the deadline; a person sends it from their own mailbox and closes the task. **What if our contracts are scattered across three Drive folders and two mailboxes?** Point the worker at all of them in the brief. It reads what the connections expose. Anything it cannot find is listed as a gap in the register, which is usually the most useful part of the first delivery. **How does this connect to legal's version of the same list?** It should be the same list. Vendor agreement management is the legal-side view of the same contracts, with different fields mattering. Keeping both in one workstream stops finance and legal from maintaining two registers that disagree. **Does the worker need access to our bank?** No, and it could not have it. The connection catalog is fixed and contains no banking connection. Actual spend comes from Stripe or from the receipts in the connected mailbox. ## Related - https://www.polarishq.co/use-cases/finance - https://www.polarishq.co/use-cases/legal/vendor-agreement-management - https://www.polarishq.co/use-cases/operations/vendor-procurement - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/glossary/focus-lane --- --- title: "HR Use Cases for AI Workers | Polaris" description: "Five HR jobs an AI worker prepares in Polaris: scheduling, onboarding, policy docs, application summaries, review cycles. Decisions about people stay human." url: https://www.polarishq.co/use-cases/hr section: Use cases updated: 2026-08-21 --- # HR work an AI worker can prepare, and where it must stop Scheduling, onboarding logistics, policy drafts and review-cycle admin, prepared by a worker. Every decision about a person stays with a person. ## The short answer Polaris gives a people team an AI worker for the administrative half of HR: scheduling interviews across calendars, running the onboarding checklist, drafting policy documents, summarizing applications against written criteria, and chasing a performance review cycle to completion. The worker uses Gmail, Google Calendar, Slack and Google Drive. Hiring, firing, pay and performance decisions are made by humans and are never delegated to a worker. - **Connections used:** Gmail · Google Calendar · Slack · Drive - **Decisions delegated:** None - **Software cost:** $0, no seats - **Work cost:** ~$2 per human-hour delivered ## HR is two jobs and only one of them can be handed over One job is judgment about people: who to hire, what someone is paid, whether a performance concern is real, how to handle a grievance. That job carries legal exposure and moral weight, and it belongs to a named human being every time. The other job is logistics. Finding a slot that works across four calendars. Making sure the laptop, the accounts and the first-week meetings exist before someone's first Monday. Chasing eleven managers for review forms. Keeping a policy document current and versioned. The second job is enormous, unglamorous, and the reason the first job gets done badly when a people team is stretched. That is what a worker takes. ## The five people-operations jobs covered here - **Candidate screening** — Structured summaries of applications against the criteria you wrote down, with no score, no ranking and no rejection. - **Employee onboarding** — The pre-start checklist run to completion so week one is not an apology. - **Interview scheduling** — The calendar puzzle solved and the invites drafted, including the reschedules. - **Policy documentation** — Drafts and updates in versioned docs, with the diff visible and a named approver. - **Performance review cycles** — The cycle chased and the packets assembled. The ratings and the conversations are the manager's. ## Written into every HR worker's brief This split is not advisory. Put it in the SKILL.md so it is visible to everyone who reads the worker's capabilities. **The worker may** - Summarize what a document says - Check whether stated criteria are evidenced - Schedule, chase, remind and assemble - Draft text for a human to approve and send - Report who has not responded **The worker may not** - Score, rank or shortlist a person - Reject, hire, promote or set pay - Assess performance - Send anything to a candidate or an employee - Close its own task ## What each job needs and what comes back | Job | Connections | Delivered as | | --- | --- | --- | | Candidate screening | Gmail, Google Drive | One structured summary per application, unranked | | Employee onboarding | Slack, Google Calendar, Gmail | A per-hire checklist with owners and a daily status | | Interview scheduling | Google Calendar, Gmail | Proposed slots plus drafted invites and confirmations | | Policy documentation | Docs, Google Drive, web search | A versioned draft with the changes marked | | Performance review cycles | Google Calendar, Gmail, Docs | Completion tracker plus an assembled packet per person | > **Polaris does not give employment-law or HR-compliance advice** > > Employment rules differ by country, by state and by contract, and getting them wrong is expensive for the employer and unfair to the employee. A worker in Polaris prepares documents and runs process. It has no view on whether your policy is lawful where you operate. Have a qualified adviser review anything that becomes policy, and keep the work log so they can see what was prepared and by what instruction. ## HR use cases in detail - [use-cases/hr/candidate-screening](https://www.polarishq.co/use-cases/hr/candidate-screening) - [use-cases/hr/employee-onboarding](https://www.polarishq.co/use-cases/hr/employee-onboarding) - [use-cases/hr/interview-scheduling](https://www.polarishq.co/use-cases/hr/interview-scheduling) - [use-cases/hr/policy-documentation](https://www.polarishq.co/use-cases/hr/policy-documentation) - [use-cases/hr/performance-review-cycles](https://www.polarishq.co/use-cases/hr/performance-review-cycles) ## Questions people ask **Can an AI worker decide who to interview?** No, and it should not be asked to. It can summarize each application against the criteria you published for the role, in the same shape for every candidate, so a human comparing them is comparing like with like. The shortlist is a human decision with a human's name on it. **Do employees know when a worker touched their file?** Everything a worker does is a task with a comment thread and a work log entry showing what it read and produced. That record is the honest answer to the question, and it is more of an audit trail than most manual HR admin leaves behind. **Is a people team big enough to justify this?** The teams that get the most out of it are the ones with no dedicated people team at all, where a founder or an office manager is doing onboarding between other work. Polaris is free for unlimited people, so the cost is only the hours the worker delivers. **What connections does an HR worker actually need?** Usually Google Calendar and Gmail for scheduling, Slack for chasing, and Google Drive for documents. The catalog is fixed and there is no HR information system in it, so a worker never has direct access to a payroll or HRIS record. ## Related - https://www.polarishq.co/use-cases/hr/candidate-screening - https://www.polarishq.co/use-cases/hr/employee-onboarding - https://www.polarishq.co/use-cases/hr/interview-scheduling - https://www.polarishq.co/ai-workers/recruiter - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/for/startups --- --- title: "AI Candidate Screening Summaries | Polaris" description: "An AI worker reads every application against your published criteria, quoting the evidence and naming the gaps. No scores, no ranking, no rejections." url: https://www.polarishq.co/use-cases/hr/candidate-screening section: Use cases updated: 2026-08-21 --- # Application summaries, not candidate scores A worker reads every application in the same shape against the criteria you published. It does not rank anyone, and it does not reject anyone. ## The short answer An AI worker in Polaris reads applications from a connected Gmail inbox and Google Drive and returns one structured summary per candidate against the criteria written for the role: what evidence exists for each requirement, what is missing, and the exact quote it came from. It produces no score, no ranking and no rejection. A hiring manager reads the summaries and decides who to interview. - **Connections:** Gmail · Google Drive - **Produces:** One evidence summary per candidate - **Does not produce:** Scores, rankings, rejections ## The reason screening goes wrong is that it is done inconsistently, late at night Two hundred applications arrive. The first thirty get read carefully. The next hundred get eight seconds each. By the last seventy the reader has an unspoken heuristic that has nothing to do with the job description, and nobody can reconstruct why anyone was cut. Automated scoring makes that worse rather than better, because it takes an unexamined judgment and gives it a number, which makes it look defensible. The useful intervention is consistency of reading, not automation of deciding. A worker reads all two hundred with the same attention and the same questions, quotes what it found, and says plainly where the application does not answer a requirement. Then a person makes every call, with the evidence in front of them. ## What a summary contains and what it deliberately omits **In every summary** - Each published requirement, with the evidence quoted verbatim - Requirements with no evidence in the application, named as such - Dates and durations as the candidate stated them - Anything the application says that does not fit the stated criteria but is factually notable - A link back to the source document **Never in a summary** - A score or a percentage match - A rank against other candidates - A yes or no recommendation - Inference about age, nationality, health, family or any protected characteristic - Anything sourced from outside the application unless you explicitly asked for it ## How to brief the worker so the output is fair The quality of the screening is set by the criteria, and the criteria are yours. - **Write the criteria before the first application arrives** — Put them in the SKILL.md. Criteria invented halfway through a pile are the mechanism by which bias enters a process. - **Make each criterion evidence-shaped** — "Has shipped a production system they were on call for" can be evidenced or not. "Strong engineer" cannot, and asking a worker to assess it produces noise dressed as a finding. - **Instruct it to quote, not to characterize** — A quote can be checked against the source in two seconds. A characterization cannot. - **Name the human who decides** — The task has an assignee. Make the shortlist a task assigned to a person, and keep it separate from the summarizing task. > **Automated decisions about candidates carry legal duties in many places** > > Several jurisdictions regulate automated decision-making in hiring, including notice, explanation and audit requirements. Polaris does not tell you what applies to you. It is built so the question is easier to answer honestly: the worker summarizes, a human decides, and the work log records what the worker read and produced. Get your own legal advice before any part of a hiring process becomes automated. ## Where the work goes | Stage | Who | Output | | --- | --- | --- | | Define criteria | Hiring manager | Criteria in the role's SKILL.md brief | | Read applications | Worker | One structured summary per candidate | | Flag gaps | Worker | Requirements with no evidence, named per candidate | | Shortlist | Human, always | A named list with reasons | | Reply to candidates | Human | Sent from a person's mailbox | ## Questions people ask **Can we ask it to rank the top ten?** You can ask, and you should not. A ranking from a worker gets treated as an assessment even when everyone agrees it is only a suggestion, and that is precisely the failure mode this page exists to avoid. Ask instead for the summaries sorted by application date and do the ranking yourself. **Will it search the web for a candidate?** Only if you explicitly instruct it to, and that is worth thinking hard about before you do. Unsolicited background research on applicants creates fairness and data-protection problems that a hiring process does not need. **What does a screening run cost?** It is billed on human-equivalent hours from the work log at roughly $2 per hour, computed from observable effort: documents read, prose produced, files generated. A large pile of applications is a real job and the log itemises it line by line so you can check the arithmetic. **Can several people see the same summaries?** Yes. The delivery is a comment on a task in a shared workspace, so the whole hiring panel reads the same document rather than four people forming four impressions from four different reads. ## Related - https://www.polarishq.co/use-cases/hr - https://www.polarishq.co/use-cases/hr/interview-scheduling - https://www.polarishq.co/use-cases/hr/employee-onboarding - https://www.polarishq.co/ai-workers/recruiter - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work --- --- title: "AI-Run Employee Onboarding Checklists | Polaris" description: "An AI worker builds a per-hire onboarding checklist, chases each owner in Slack, books week-one meetings, and reports what is missing before day one." url: https://www.polarishq.co/use-cases/hr/employee-onboarding section: Use cases updated: 2026-08-21 --- # Onboarding that is finished before the first Monday Nobody's first day should start with an apology about accounts. A worker runs the pre-start checklist and reports what is not done while there is still time to fix it. ## The short answer An AI worker in Polaris runs employee onboarding by creating a per-hire checklist from your template, assigning each item to the person who owns it, chasing them in Slack, booking the first-week meetings in Google Calendar, and posting a daily status naming what is still outstanding. IT access, contracts and pay are actioned by the humans who own those systems, not by the worker. - **Connections:** Slack · Google Calendar · Gmail - **Starts:** When the start date is confirmed - **Reports:** Daily until every item is closed ## Onboarding fails in the gap between eight owners The checklist is not the problem. Almost every company has one. The problem is that finishing it needs IT, finance, the hiring manager, an office lead and whoever administers each tool, and none of them owns the outcome. Each does their bit when they see the message. So on Monday the laptop is there but the email is not, or the email is there but nobody booked the manager's first one-to-one, and the new hire spends day one watching people apologise. A worker is a reasonable owner of the chase precisely because chasing eight busy people every day is a job no human wants and no human is good at. ## How an onboarding run works 1. **Keep the template as a doc, not in someone's head** — Docs in Polaris are a nested tree with to-dos and sub-pages, and they are versioned. The onboarding template lives there and gets better every time someone finds a gap. 2. **Create the hire as a workstream** — One workstream per hire keeps the thread in one place: the checklist, the questions, the week-one plan and the delivery comments. 3. **Assign the worker and give it Slack and Calendar** — Slack so it can chase item owners where they already are, Calendar so it can put the week-one meetings in before the diary fills up. 4. **Set acceptance criteria with a hard date** — For example: every item has a named owner; anything not confirmed by two working days before the start date is escalated to the hiring manager; the week-one calendar is booked and accepted. 5. **Read the daily status and unblock** — The worker posts what is outstanding and who it is waiting on. A human closes the task once the new starter is actually set up. ## Items a worker can genuinely own versus items it can only chase - **Can own: the week-one calendar** — Booking the manager one-to-one, the team intro and the first review point across real calendars, and rebooking when someone declines. - **Can own: the welcome pack** — Assembling the docs the new starter needs into one place, and checking the links in them still resolve. - **Can own: the status** — Knowing, every morning, exactly which of the twenty-two items are done and which are not. - **Can only chase: account creation** — Creating accounts in your identity provider is an IT action in an IT system. The worker asks, reminds and escalates. It does not provision. - **Must not touch: contract and pay** — Employment paperwork and payroll are human actions with legal consequences, handled outside Polaris by the people who own them. ## A first-week plan the worker can assemble The specifics are yours. This is the shape it fills in from your template. | When | What the worker sets up | Who owns it on the day | | --- | --- | --- | | Day 1, morning | Manager one-to-one booked, welcome doc assembled | Hiring manager | | Day 1, afternoon | Team introduction slot, access checklist confirmation | Team lead | | Day 2 | Tooling walkthrough booked with whoever administers each tool | Tool owners | | Day 3–4 | First real task drafted into the hire's Focus lane | Hiring manager | | Day 5 | End-of-week check-in booked, plus a written prompt asking what was missing | Hiring manager | | Day 30 | Review point already in the calendar | Hiring manager | > **The last item on the list is asking what went wrong** > > Brief the worker to send the new starter a short written question at the end of week one: what did not exist when you needed it? Collect the answers into the template doc. An onboarding checklist that never changes is an onboarding checklist that nobody is reading. ## Questions people ask **Can the worker create accounts in our tools?** No. The connection catalog is fixed and contains no identity or device-management system. The worker chases the human who provisions accounts and reports whether it is done, which is the part that actually slips. **Does the new hire get a Polaris account?** If you want one. Polaris is free for unlimited humans with no seat cost, and sign-in is passwordless: an email address, then a code sent to it. There is no password to issue or reset. **What if we hire in batches?** One workstream per hire, one worker running all of them. The status comment names each hire separately so a stalled item on one person does not disappear into a batch summary. **Can it handle offboarding too?** The same pattern works: a checklist, named owners, a chase and a daily status. Be stricter about the boundary, since offboarding touches access revocation and final pay, both of which are human actions in systems the worker cannot reach. ## Related - https://www.polarishq.co/use-cases/hr - https://www.polarishq.co/use-cases/hr/interview-scheduling - https://www.polarishq.co/use-cases/hr/policy-documentation - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/glossary/workstream --- --- title: "AI Interview Scheduling Across Calendars | Polaris" description: "An AI worker finds interview slots across panel calendars, drafts the invites and handles reschedules. A person sends every candidate email." url: https://www.polarishq.co/use-cases/hr/interview-scheduling section: Use cases updated: 2026-08-21 --- # Interview scheduling, including the reschedules Four calendars, two time zones, a candidate who can only do early mornings, and a panellist who declines twice. This is the job. ## The short answer An AI worker in Polaris handles interview scheduling by reading panellist availability in Google Calendar, proposing slots that fit every constraint you stated, and drafting the invite and the candidate email for a person to send. When someone declines, it proposes the next set of options in the same task thread. The worker never emails a candidate directly. - **Connections:** Google Calendar · Gmail - **Handles:** Proposals, conflicts and reschedules - **Sends to candidates:** Never. A human does ## Scheduling is not hard. It is just relentless. A single onsite loop can take fifteen messages. Someone reads four calendars, picks three options, sends them, waits, loses one option to a meeting booked in the meantime, and starts again. Multiply by six candidates and a week is gone. The reason it stays manual is that the constraints are unwritten: this panellist should not do two interviews back to back, that one is genuinely unavailable on Fridays regardless of what the calendar shows, and candidates currently employed elsewhere should be offered early or late slots so they are not explaining an absence. Write those constraints down once, in the worker's skill file, and the relentless part goes away while the judgment stays with you. ## Constraints worth writing into the brief Each of these is an unwritten rule in most teams, and each is the reason a scheduling tool gets abandoned. - **Buffers, not just gaps** — Fifteen minutes either side of an interview for notes. A slot that touches another meeting is not a valid slot. - **Panel load** — No panellist does more than two interviews in a day, and never two in a row. - **Candidate-friendly hours** — Offer early morning or after 17:00 to candidates in another job unless they have said otherwise. - **Time zones stated explicitly** — Every proposed slot appears in both the candidate's local time and the panel's, written out, not implied by a calendar invite. - **A hard latest date** — If no slot is agreed within your target window, escalate to the hiring manager instead of continuing to propose. ## What happens at each turn | Event | Worker | Human | | --- | --- | --- | | Interview requested | Reads calendars, proposes three slots with time zones | Picks or amends | | Candidate replies with a preference | Checks it still works, drafts the invite | Sends it | | A panellist declines | Proposes replacement slots in the same task thread | Confirms | | Candidate goes quiet | Flags on the agreed day. No chasing mail sent | Decides whether and how to follow up | | Interview happens | Books the debrief slot if that is in the brief | Runs the interview and the decision | ## What you are actually buying Not a scheduling widget. The hours around one. - **$0** — For the workspace, per person. No seats, unlimited humans - **~$2** — Per human-hour delivered. Only when a job actually delivers - **0** — Emails sent to candidates by a machine. Every candidate message is sent by a person > **Why the candidate never hears from the worker** > > A scheduling exchange is the candidate's first real contact with how your company behaves. Drafts arrive in the task thread and a person sends them. It costs ten seconds and it means nobody's first impression of you is a machine that got a time zone wrong at 06:00. ## Questions people ask **Does it need access to every panellist's calendar?** It needs the Google Calendar connection authorized for the org and visibility of the calendars you want it to read. Connections are authorized once and stored server-side, so nobody is pasting credentials into a task. **What about candidates who are not in Google Calendar?** They never are. The worker reads your side and proposes times, and the exchange with the candidate happens over drafted email that a person sends. That is the same way a human coordinator does it. **Can it handle a full onsite loop with five panellists?** Yes, and that is where the constraint list earns its keep. Give it the buffer rules, the load limits and the order the panel should run in, and check its first proposed loop closely before trusting the next ten. **Where do reschedules live?** In the same task thread as the original request, as further comments. The whole history of a loop sits in one place instead of across a mailbox, a calendar and a chat channel. ## Related - https://www.polarishq.co/use-cases/hr - https://www.polarishq.co/use-cases/hr/candidate-screening - https://www.polarishq.co/use-cases/hr/employee-onboarding - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/ai-workers/recruiter - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/use-cases/executive/meeting-follow-ups --- --- title: "AI-Drafted HR Policy Documentation | Polaris" description: "Keep HR policies current in Polaris: an AI worker drafts, versions, collects reviewer comments and flags stale documents. Sign-off stays with your adviser." url: https://www.polarishq.co/use-cases/hr/policy-documentation section: Use cases updated: 2026-08-21 --- # Policy documents that stay current and show their changes A policy nobody has updated in three years is worse than no policy. A worker keeps the drafts moving and the versions visible. A qualified adviser signs them off. ## The short answer An AI worker in Polaris drafts and maintains HR policy documentation in versioned Polaris docs: writing a first draft from your notes, marking what changed between versions, collecting comments from reviewers, and flagging policies that have not been reviewed within the period you set. Whether a policy is lawful where you operate is a question for a qualified adviser, and Polaris does not answer it. - **Lives in:** Versioned Polaris docs - **Connections:** Google Drive · Slack · web search - **Sign-off:** A named human, every time ## The failure mode is silence, not error Most handbooks are not wrong on the day they are written. They go wrong by sitting still: the company moves to four countries, adopts a different working pattern, changes who approves expenses, and the document describes a company that no longer exists. Nobody updates it because updating it means finding the current version, working out what changed, writing the change, getting someone to approve it, and telling everyone. Five steps, no owner. A worker can carry four of those five. The approval stays with a person, and it should be a person who is qualified to give it. ## What the worker maintains - **The draft** — A first version written from your bullet points and your existing document, in your own house language rather than boilerplate lifted from somewhere else. - **The diff** — What changed since the last version, stated in a sentence per change, so a reviewer reads the delta and not the whole document. - **The review clock** — Any policy past its review date appears as a dated task in the Focus lane with an owner attached. - **The comment round** — Reviewers comment on the doc; the worker collects unresolved comments into a list and chases the people who have not responded. - **The publication note** — A drafted announcement of what changed and what people need to do differently, for a human to send. ## Draft versus decide **The worker prepares** - Wording, structure and consistency across documents - Cross-references between related policies - A summary of what an existing policy currently says - Research on the open web, with every source named - Version history and change lists **A qualified person decides** - Whether the policy is lawful in each place you employ people - What the company's position actually is - Which obligations you are accepting - When the policy takes effect - Approving and publishing it > **A drafted policy is not legal advice, and a web search is not a legal source** > > The worker can find and cite public material, and it will name every source. That is research, not advice. Employment obligations vary by jurisdiction and change without notice, so the sensible workflow is: worker drafts, adviser reviews, company adopts. The named-source requirement exists so your adviser can check where a sentence came from instead of guessing. ## Questions people ask **Can the worker tell us whether our policy is compliant?** No. It can tell you what your document says, what it does not cover, and what public sources it found on a topic, with each source named. Whether that meets your obligations in a specific jurisdiction is a judgment for a qualified adviser. **Why keep policies in Polaris rather than a wiki?** Because the review cycle is a task, and the task and the document then live in the same place. Docs are a nested tree with versioning and file review, and a policy past its review date can be a dated item in someone's Focus lane instead of a note in a spreadsheet. **How do we stop the worker inventing a policy position?** Brief it to draft only from material you supply, and to write an explicit open question wherever your notes do not settle a point. A visible list of open questions at the top of a draft is far more useful than a confident paragraph that quietly guessed. **Can it write the announcement too?** Yes, as a draft. Anything that goes to the whole company about how people are employed is sent by a person, from their own account, after they have read it. ## Related - https://www.polarishq.co/use-cases/hr - https://www.polarishq.co/use-cases/hr/employee-onboarding - https://www.polarishq.co/use-cases/legal/policy-updates - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "Running Performance Review Cycles | Polaris" description: "An AI worker chases a review cycle to completion and assembles each packet. Ratings, feedback and the conversation stay entirely with managers." url: https://www.polarishq.co/use-cases/hr/performance-review-cycles section: Use cases updated: 2026-08-21 --- # Review cycles that finish, without a worker forming an opinion The administration of a review cycle is enormous and the judgment inside it is entirely human. A worker takes the first part and touches none of the second. ## The short answer An AI worker in Polaris runs the administration of a performance review cycle: creating a task per participant, booking the conversations in Google Calendar, chasing managers and reviewers who have not submitted, and assembling each person's packet from what was submitted. It writes no assessment, forms no view on anyone's performance, and reads nothing beyond the material submitted for the cycle. - **Connections:** Google Calendar · Gmail · Docs - **Assembles:** One packet per participant - **Assesses:** Nobody. Ever ## Cycles fail on completion rates, not on the form design Every review cycle has the same shape: a self-assessment, some peer input, a manager write-up, a conversation, an outcome. Every review cycle also has the same failure: on the deadline, sixty percent of it is in, and the people chasing the other forty percent are the same people who owe write-ups themselves. So the cycle slips three weeks, the feedback goes stale, and the conversations happen with half the input missing. Chasing is the whole job. A worker can do it every day, to everyone, without anyone taking it personally, and can tell you on day three that four managers have not started. ## Running the cycle 1. **Set up the cycle as a workstream with a task per person** — One task per participant, owned by their manager, in a workstream for the cycle. Everything for that person, including the packet, ends up on their task. 2. **Restrict what the worker can read, in writing** — State in the SKILL.md that the worker reads only the documents submitted for this cycle. It should not be reading a person's chat history or their commits to form a picture of them, and the brief is where you make that explicit. 3. **Let it chase on a schedule** — Daily reminders to whoever has an outstanding item, escalating to the manager's manager on the date you set, and a running completion figure posted to the cycle task. 4. **Have it assemble the packet** — Self-assessment, peer input and the manager's write-up collected into one document per person, in a consistent order, with anything missing named at the top. 5. **Humans hold the conversations and close the tasks** — Ratings, feedback, pay and progression are decided and delivered by people. The manager closes the task when the conversation has happened. ## The line, drawn hard Performance is the clearest case in this whole cluster where an AI worker must not have an opinion. **The worker does** - Create, assign and chase the tasks - Book the conversations - Report completion rates and who is behind - Assemble each packet in a consistent order - Name what is missing from a packet **The worker does not** - Summarize or paraphrase someone's feedback about a colleague - Rate, rank or calibrate - Suggest an outcome - Read work systems to build evidence about a person - Send anything to the person being reviewed > **Do not let a worker summarize peer feedback** > > It is the most tempting request on this page and the one to refuse. Feedback about a person is written by a colleague in specific words, often carefully chosen, and a summary strips the care out and adds a distortion nobody can audit. Assemble it verbatim, in the order it was given, and let the manager read all of it. ## What to measure about your own cycle - **Completion on the deadline, not two weeks after** — The number that tells you whether the cycle is working. The worker reports it daily during the window. - **Days from deadline to last conversation held** — This is the figure that quietly stretches every year. - **Packets missing an input** — A packet delivered with a gap named at the top is honest. A packet that hides the gap is not. ## Questions people ask **Can the worker draft a manager's review for them?** No. A review is a manager's own assessment of a person they are responsible for, and outsourcing the words to a machine hollows out the one conversation of the year that is supposed to be personal. The worker chases the manager. The manager writes. **Is the review material private?** It lives on tasks in your workspace with the access controls you set, and Polaris stores workspace data in Postgres with row-level security. Treat access design as your decision: put the cycle in a workstream visible to the people who should see it, and no wider. **What does the worker actually cost for a cycle?** Hours are computed from the observable effort in each job and billed at roughly $2 per human-equivalent hour, itemized on the work log. Chasing and assembly are small repeated jobs, and any line can be challenged from the log. **Can it run a lightweight cycle for a ten-person company?** That is the case where it helps most, because a ten-person company has no HR function and the founder is both the chaser and one of the people being chased. A worker chasing on a schedule removes the awkwardness that stalls small-company cycles. ## Related - https://www.polarishq.co/use-cases/hr - https://www.polarishq.co/use-cases/hr/policy-documentation - https://www.polarishq.co/use-cases/executive/okr-tracking - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/integrations/google-calendar --- --- title: "Legal Operations Use Cases for AI Workers | Polaris" description: "Five legal-operations jobs an AI worker handles in Polaris: contract tracking, compliance checklists, policy updates, vendor agreements, mark monitoring." url: https://www.polarishq.co/use-cases/legal section: Use cases updated: 2026-08-21 --- # Legal operations work an AI worker can carry Tracking, chasing, assembling and watching. Everything a legal team spends time on that is not actually practicing law. ## The short answer Polaris gives a legal or operations team an AI worker for legal administration: tracking where each contract is in review, keeping compliance checklists current and chased, watching for changes to policies and public sources, maintaining a vendor agreement register, and monitoring the open web for uses of a mark. Every legal judgment, every piece of advice and every signature belongs to a qualified lawyer. - **Connections used:** Gmail · Drive · Calendar · web search - **Legal advice given:** None. Polaris is not a law firm - **Software cost:** $0, unlimited people - **Work cost:** ~$2 per human-hour delivered ## Most of what slows a legal team down is not legal work Ask an in-house counsel or the founder who is doing the legal work at a small company where the time goes. It goes on knowing which of eleven contracts is currently with the other side, on remembering that a policy needs re-reviewing, on finding the signed version, and on chasing a sales lead who promised to explain what the customer actually asked for. That is queue management with high stakes attached. The stakes are why it does not get delegated, and the volume is why it does not get done well. An AI worker takes the queue management. It tracks state, chases people, assembles what a lawyer needs to look at, and watches sources that change. It does not read a clause and tell you what it means. ## The five legal-operations jobs covered here - **Contract review tracking** — Where every agreement is, who is holding it up, and how long it has been sitting there. - **Compliance checklists** — A living checklist with named owners, evidence attached, and the gaps stated plainly. - **Policy updates** — Watching sources you nominate and reporting what changed, with the source quoted. - **Vendor agreement management** — The register of what you signed, with the terms that bind you pulled out and quoted. - **Trademark monitoring** — Regular sweeps of the open web and Instagram for uses of your mark, with evidence captured. ## The one distinction that governs this entire cluster **Legal operations** - Knowing where a document is - Knowing what a document says, quoted verbatim - Knowing who owes an answer and for how long - Assembling the file a lawyer needs - Noticing that a source changed **Practicing law** - Deciding what a clause means - Advising on risk or exposure - Deciding whether you are compliant - Deciding whether to enforce a right - Negotiating, approving and signing ## What each job needs and what comes back | Job | Connections | Delivered as | | --- | --- | --- | | Contract review tracking | Gmail, Google Drive | A status board plus an ageing report per agreement | | Compliance checklists | Docs, Google Drive, Slack | A checklist with owners, evidence links and named gaps | | Policy updates | web search, Docs | A change report quoting the source and the date it changed | | Vendor agreement management | Google Drive, Gmail, Google Calendar | A register with key terms quoted and deadlines dated | | Trademark monitoring | web search, Instagram, Google Drive | A dated evidence file per potential use of the mark | > **Polaris does not provide legal advice and is not a substitute for a lawyer** > > Everything on these pages is document handling and process. A worker quotes what a document says and names where it found it. It does not interpret, advise or assess risk, and nothing it produces should be relied on as advice. The reason to run legal operations this way is that a qualified person gets a clean file and a clear question sooner, with a work log showing exactly what was gathered and how. ## Legal use cases in detail - [use-cases/legal/contract-review-tracking](https://www.polarishq.co/use-cases/legal/contract-review-tracking) - [use-cases/legal/compliance-checklists](https://www.polarishq.co/use-cases/legal/compliance-checklists) - [use-cases/legal/policy-updates](https://www.polarishq.co/use-cases/legal/policy-updates) - [use-cases/legal/vendor-agreement-management](https://www.polarishq.co/use-cases/legal/vendor-agreement-management) - [use-cases/legal/trademark-monitoring](https://www.polarishq.co/use-cases/legal/trademark-monitoring) ## Questions people ask **Can an AI worker review a contract?** It can tell you what a contract says, quote the clause and name the file and page. It cannot tell you what the clause means for you, what risk it carries or whether to accept it. Those are legal judgments and Polaris does not make them. **Is this useful for a company with no lawyer?** It is useful for getting organized before you engage one, which is where small companies waste the most money. A worker that has already assembled the signed versions, quoted the relevant terms and listed the open questions turns an expensive hour of counsel time into a productive one. **Where does the audit trail live?** On the task. Every job a worker runs produces a comment thread and a work log entry listing what it read, what it produced and how long the job was assessed at. That record is queryable long after the person who ran it has moved on. **Can a worker sign or serve anything?** No. It has no signing capability and no ability to send mail on its own. Anything that binds the company or constitutes notice is drafted in a comment and sent by a person. ## Related - https://www.polarishq.co/use-cases/legal/contract-review-tracking - https://www.polarishq.co/use-cases/legal/compliance-checklists - https://www.polarishq.co/use-cases/legal/vendor-agreement-management - https://www.polarishq.co/ai-workers/paralegal - https://www.polarishq.co/use-cases/finance/vendor-management - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "Contract Review Tracking with an AI Worker | Polaris" description: "Track every contract in review inside Polaris: an AI worker maintains state, names who is holding each agreement, and ages the stalled ones." url: https://www.polarishq.co/use-cases/legal/contract-review-tracking section: Use cases updated: 2026-08-21 --- # Knowing where every contract is, without asking three people The question that eats a legal team's week is "where is that one now?". A worker keeps the answer current and ages every stalled agreement. ## The short answer An AI worker in Polaris tracks contract review by reading the negotiation threads in a connected Gmail account and the document versions in Google Drive, then maintaining one task per agreement with its current state, who is holding it, and how many days it has been there. It posts an ageing report as a comment. Lawyers do the reviewing, the negotiating and the signing. - **Connections:** Gmail · Google Drive - **One task:** Per agreement, from request to signature - **Reported:** State, holder, days waiting ## Contract state is knowledge that lives in one person's head A deal is waiting on redlines. Is it with us or with them? Did the customer's counsel respond last Tuesday? Did anyone tell sales that the indemnity question was answered a week ago? In most companies the answer to all three is that one person knows, and finding out costs a message and an hour of latency. When that person is on holiday, the answer costs a day. The information is not hidden. It is in a mail thread and in a folder of files named v3, v3-final and v3-final-ACTUAL. Reading both and writing down the state is a job with no judgment in it, and it is the job that makes a legal queue visible. ## The states a worker tracks Use your own names for these. What matters is that every agreement is in exactly one of them and the clock is running. | State | What the worker records | The question it makes answerable | | --- | --- | --- | | Requested | Who asked, what for, what the commercial terms are | How many requests came in this month, and from where | | With us | Which reviewer, since when | Are we the bottleneck right now | | With them | Which version was sent, on what date | How long has the other side been sitting on it | | Open point | The point, quoted from the thread, and who owes an answer | What is actually blocking this one | | Ready to sign | Which file is final and where it is | Is the version about to be signed the version that was agreed | | Signed | The executed file and its location | Can anyone find the signed copy in ten seconds | ## Tracking versus reviewing **The worker** - Reads threads and files to determine current state - Ages every agreement and flags anything past your threshold - Quotes open points verbatim from the correspondence - Chases the internal owner who owes an answer - Assembles the file before a review session **The lawyer** - Reads the document - Decides what is acceptable - Writes the redlines - Advises the business on the trade - Approves execution ## Two habits that make the tracking trustworthy - **Make the worker cite the evidence for the state it recorded** — "With them since 12 Aug" is a claim. "With them since 12 Aug, per the mail sent at 16:42 attaching v4" is a fact someone can check. - **Never let it infer that a point is closed** — An unanswered question in a thread is open until a human says otherwise. Brief it to list ambiguity rather than resolve it, because a wrongly-closed open point is the one error here that costs real money. > **A quoted clause is not an opinion on that clause** > > The worker will happily pull the limitation of liability out of six agreements and show them side by side with the file and page for each. That is retrieval. Deciding which of those six positions you can live with is legal advice, and it comes from a qualified person, not from Polaris. ## Questions people ask **Does this replace a contract lifecycle management system?** It is not a CLM and does not pretend to be one. There is no clause library, no template automation and no e-signature. What it gives a small team is the part they are missing: current state, ageing and a chase, in the same workspace as everything else. **Can it draft a first-pass redline?** Drafting text is something the worker can do, and whether you want that is a decision for your counsel, not for a marketing page. What Polaris guarantees is that nothing it drafts leaves the workspace without a person reading and sending it. **What if negotiations happen in Slack rather than email?** Give it the Slack connection as well as Gmail. Slack messages can also arrive in the Polaris Inbox as prefilled task suggestions, so a commitment made in a channel becomes a tracked item after one click rather than being lost in scrollback. **How far back can it reconstruct?** As far back as the connected mailbox and Drive folders go. The first run on an existing pile is the expensive one and usually the most revealing, because it surfaces the agreements nobody had realized were still open. ## Related - https://www.polarishq.co/use-cases/legal - https://www.polarishq.co/use-cases/legal/vendor-agreement-management - https://www.polarishq.co/use-cases/legal/compliance-checklists - https://www.polarishq.co/use-cases/sales/proposal-writing - https://www.polarishq.co/ai-workers/paralegal - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/delivery-comment --- --- title: "AI-Maintained Compliance Checklists | Polaris" description: "Run compliance checklists in Polaris: an AI worker chases owners, collects evidence and names every gap. What counts as compliant is your adviser's call." url: https://www.polarishq.co/use-cases/legal/compliance-checklists section: Use cases updated: 2026-08-21 --- # Compliance checklists with evidence attached and gaps named A checklist where every line is ticked and nothing is evidenced is not a control. It is a document that will fail an audit slowly. ## The short answer An AI worker in Polaris maintains a compliance checklist as tasks with named owners, collects the evidence each item requires into the workspace, chases owners who have not supplied it, and reports gaps explicitly rather than marking an unevidenced item complete. Whether the checklist is the right checklist, and whether the evidence is sufficient, is decided by a qualified adviser or auditor. - **Connections:** Google Drive · Slack · Docs - **Every item has:** An owner, an evidence link, a date - **Unevidenced items:** Reported as gaps, never ticked ## Checklists rot in a specific and predictable way Someone builds a good checklist for a certification or a customer security review. It is accurate on the day. Six months later half the items are ticked from memory, three of the owners have changed role, and the evidence links point at a folder that was reorganized in March. The rot is not laziness. Maintaining evidence is a continuous chase across people who have other jobs, and no single person is paid to do it. That chase is the piece to hand to a worker: ask the owner, collect the artifact, date it, and report what is missing. The judgment, which is whether any of it is sufficient, stays with whoever is qualified to say so. ## What each checklist item carries An item without all four of these is not really a control, and the worker will say so. - **A named human owner** — Not a team. A person, with a task assigned to them, because a control owned by a team is owned by nobody. - **An evidence artifact** — A file, a screenshot, an exported record, stored in the workspace where the checklist lives rather than linked to a folder that moves. - **A date** — When the evidence was produced, not when the item was last ticked. Old evidence is a finding. - **A re-check interval** — The interval turns the item into a recurring task in the Focus lane instead of a line nobody looks at until an auditor asks. ## Standing one up 1. **Put the checklist in a doc, one item per line, with owners** — Docs in Polaris support to-dos and sub-pages, and they are versioned, so the checklist itself has a history you can show someone. 2. **Turn each item into a recurring task** — Assigned to its human owner, with an interval. The Focus lane carries the ones due now across Today, This week and Next 30 days. 3. **Give the worker Slack and Drive** — Slack to chase owners where they work, Drive to find and collect the evidence artifacts. 4. **Write the acceptance criteria to forbid silent ticks** — State it directly: an item is only complete when a dated artifact is attached and its owner has confirmed. Anything else is reported as a gap with a reason. 5. **Review the gap list, decide, close** — The worker delivers the gap list. A qualified person decides what it means and what to do, and closes the task. ## The boundary here matters more than usual **The worker reports** - Which items have evidence and which do not - How old each artifact is - Who has not responded, and for how long - Where the checklist itself has not been reviewed - What changed since the last run **A qualified person judges** - Whether this is the right checklist for your obligations - Whether an artifact actually evidences the control - Whether a gap is material - What to tell a customer, an auditor or a regulator - Whether you are compliant > **A worker cannot tell you that you are compliant** > > It can tell you that nineteen of twenty-three items have dated evidence and name the four that do not. That is a status report, and it is genuinely useful the day before an audit conversation. It is not an assessment, an opinion or an assurance, and Polaris gives none of those. ## Questions people ask **Which frameworks does it support?** It supports whatever checklist you write. Polaris ships no framework content, no control library and no certification templates, because shipping a generic list would encourage exactly the false comfort this page argues against. Bring the checklist your adviser gave you. **Can it collect evidence automatically?** It can retrieve artifacts from the connections it holds, which for most teams means Google Drive. Anything living in a system outside the fixed catalog has to be supplied by its human owner, and the worker's job there is to ask and keep asking. **What does the auditor actually see?** A checklist with dated artifacts, plus a task history showing when each item was chased, by whom and what was returned. The work log adds what the worker itself did in each run. That is a stronger trail than a spreadsheet with ticks in it. **Does the worker need admin access to our systems?** No. Connections are authorized once at org level from a fixed catalog and stored server-side. A compliance worker generally needs Drive and Slack, and giving it more than the job requires is a bad idea regardless of what the product allows. ## Related - https://www.polarishq.co/use-cases/legal - https://www.polarishq.co/use-cases/legal/policy-updates - https://www.polarishq.co/use-cases/legal/vendor-agreement-management - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/ai-workers/paralegal - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/work-log --- --- title: "Monitoring Policy Changes with an AI Worker | Polaris" description: "An AI worker re-reads the public sources you nominate on a schedule, quotes what changed and dates it. Interpretation and action stay with a qualified human." url: https://www.polarishq.co/use-cases/legal/policy-updates section: Use cases updated: 2026-08-21 --- # Knowing that a source changed, on the week it changed A worker watches the public pages you nominate, quotes what changed, and dates it. What the change means for you is a question for your adviser. ## The short answer An AI worker in Polaris monitors policy updates by re-reading the public sources you nominate on a schedule using web search, comparing them to what it recorded last time, and reporting each change with the text quoted and the date it was observed. It delivers the change report as a comment on a task. Interpreting the change and deciding what to do about it is a human judgment. - **Connections:** web search · Docs - **Cadence:** Whatever schedule you set - **Every change:** Quoted, sourced and dated ## The expensive failure is not misreading a change. It is missing one for eight months. Platform terms, supplier policies, published guidance from a body you deal with, a partner's acceptable-use page: these change quietly, and the first anyone notices is when something you were doing stopped being allowed. Nobody has the job of re-reading twelve web pages every month. It is the definition of work that is important and never urgent, so it never happens. A worker with web search can genuinely do this. It reads, it diffs against what it saw before, and it reports what moved. It stays honest by quoting: the change report contains the actual text and the URL, so the person reading it is looking at the source rather than at a paraphrase. ## What goes in the watch list and what comes back | Source type | Example of what you would nominate | What a report contains | | --- | --- | --- | | Platform terms | The terms page of a service you build on | The changed clause, quoted, with old and new text | | Supplier policy | A supplier's published data or security policy | What changed and the date the page was observed | | Published guidance | Guidance from a body relevant to your sector | The passage, quoted, with the source URL | | Competitor public commitments | A competitor's published SLA or pricing terms | The delta, with a note that this is a public page only | | Your own published pages | Your privacy policy or terms as they are actually live | Drift between what you intended and what is published | ## Rules that keep a monitoring worker useful rather than noisy - **Quote, never summarize** — A summary of a legal change is a paraphrase of something whose exact words matter. The report carries the text. - **Report no-change explicitly** — A run that found nothing should say so with the date. Silence is indistinguishable from a broken job. - **Separate the material from the cosmetic** — Ask for formatting and navigation changes to be listed separately from changes to substance, or the report becomes unreadable within two months. - **Keep the watch list in a versioned doc** — The list of what you watch is itself a thing that gets stale. Put it in a doc with a review date and an owner. > **This is monitoring, not legal analysis** > > The worker reports that a source changed and quotes the change. It does not tell you whether the change affects you, creates an obligation, or requires you to do anything, and Polaris does not give legal advice. The value is that the question reaches your adviser in the week it arose instead of during a diligence process eighteen months later. ## Questions people ask **How does it know what changed if it cannot store the whole page?** It records what it observed in a doc in your workspace on each run, and compares against that record on the next. The history is yours, versioned, and readable, so you can see exactly what the worker believed the page said in March. **Can it monitor sources behind a login?** Only where a connection in the fixed catalog reaches them. Web search covers the public web. A supplier portal that requires a password is out of scope, and the honest answer is to keep that one on a human's calendar. **How often should it run?** Monthly is enough for most sources and cheap, since you pay by delivered human-equivalent hour rather than per check. Put anything genuinely fast-moving on a weekly schedule and accept that most weeks the report will say nothing changed. **Who should the report go to?** A named person, as an assigned task, not a channel. A monitoring report posted to a channel with nobody's name on it is read by nobody, which is the same outcome as not running it. ## Related - https://www.polarishq.co/use-cases/legal - https://www.polarishq.co/use-cases/legal/compliance-checklists - https://www.polarishq.co/use-cases/legal/trademark-monitoring - https://www.polarishq.co/use-cases/hr/policy-documentation - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks --- --- title: "Vendor Agreement Management with AI | Polaris" description: "An AI worker quotes the binding terms from each signed vendor contract, dates every deadline, and names the agreements with no signed copy on file." url: https://www.polarishq.co/use-cases/legal/vendor-agreement-management section: Use cases updated: 2026-08-21 --- # A register of what you actually signed Finance tracks what a vendor costs. Legal needs to know what you promised them, what they promised you, and which of those promises has a date on it. ## The short answer An AI worker in Polaris maintains a vendor agreement register by reading signed agreements in Google Drive and amendment mail in Gmail, pulling out the terms that bind you with each clause quoted and located, and dating every deadline into Google Calendar. It flags agreements it cannot find a signed version of. A lawyer decides what any term means and what to do about it. - **Connections:** Google Drive · Gmail · Google Calendar - **Per agreement:** Terms quoted, located, dated - **Biggest first finding:** Agreements with no signed copy ## Two registers, one set of contracts Finance keeps a vendor list because it cares about money and renewal dates. Legal needs a different cut of exactly the same documents: what liability was accepted, what data terms apply, what happens on termination, whether there is an auto-renewal, and which of those has been amended by a side letter nobody filed. When these are two lists in two tools they disagree within a quarter, and the disagreement is discovered during diligence, at the worst possible moment. Running both cuts off one register in one workspace is the cheap fix. The worker reads the same files and produces the fields each side needs. ## The legal fields on each register entry Each is quoted from the document with a file name and a location, never paraphrased. | Field | Why it is on the register | How the worker records it | | --- | --- | --- | | Term and renewal mechanism | Auto-renewal is the term that costs the most and is read the least | Clause quoted, renewal date computed and dated in the calendar | | Notice requirement | The real deadline sits before the renewal date | Quoted verbatim with the required form of notice | | Liability position | The number a lawyer will want to compare across agreements | Quoted, with the file and page | | Data terms | Where a data processing addendum exists, and where it does not | Present or absent, stated plainly | | Termination rights | What you can actually do if the relationship goes wrong | Quoted | | Amendments and side letters | The document that quietly changes everything above | Listed with dates, or flagged as none found | ## Retrieval and judgment **The worker** - Finds the executed version, or reports that it cannot - Quotes each register field with its location - Computes renewal and notice dates and puts them in the calendar - Cross-checks the mail for amendments - Reports agreements that differ from your standard position **The lawyer** - Decides what a term means - Decides whether a position is acceptable - Advises on exposure - Negotiates the change - Approves and signs ## The findings a first run usually produces In a company that has never done this, the first delivery is rarely about clause wording. - **Agreements with no signed copy anywhere** — The most common finding, and the one that matters most in diligence. - **The version in the folder is not the version that was signed** — Found by comparing the file to what the mail thread actually attached. - **A side letter that changes a term on the main agreement** — Filed separately, never cross-referenced, invisible until someone reads both. - **Notice windows already inside 30 days** — These go straight into the Focus lane with an owner. > **Quoting a clause is retrieval. Reading it is your lawyer's job.** > > The register is a finding aid. It tells a qualified person where to look and what the document says, so their time goes on the question that needs them rather than on hunting for a PDF. Nothing on the register is advice, and Polaris does not tell you what any term means. ## Questions people ask **How is this different from the finance vendor page?** Same contracts, different fields. Finance tracks spend, renewal dates and whether the charge matches the agreement. Legal tracks the terms that bind you and the notice mechanics. Run them as one register in one workstream so the two views cannot drift apart. **Can the worker compare a supplier's terms to our standard position?** It can put them side by side and name where the wording differs, quoting both. Whether a difference is acceptable is a judgment, and the register exists to get that question in front of someone qualified with the evidence already assembled. **What if agreements are scattered across mailboxes and drives?** Point it at all of them in the brief. It reads what its connections expose, and it reports gaps rather than filling them in. The gap list from the first run is usually the most valuable output. **Does anything get signed in Polaris?** No. There is no e-signature in the product and no signing capability for a worker. Execution happens wherever you do it today, and the executed file gets filed and registered. ## Related - https://www.polarishq.co/use-cases/legal - https://www.polarishq.co/use-cases/legal/contract-review-tracking - https://www.polarishq.co/use-cases/finance/vendor-management - https://www.polarishq.co/use-cases/operations/vendor-procurement - https://www.polarishq.co/ai-workers/paralegal - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/glossary/workstream --- --- title: "Trademark and Brand Use Monitoring | Polaris" description: "An AI worker sweeps the web and Instagram for uses of your brand name and captures the URL, context and date of each. Infringement calls stay with a lawyer." url: https://www.polarishq.co/use-cases/legal/trademark-monitoring section: Use cases updated: 2026-08-21 --- # Watching the open web for uses of your mark A worker sweeps public sources on a schedule, captures dated evidence of each use it finds, and hands you a file. Whether to act is a legal decision. ## The short answer An AI worker in Polaris monitors use of a brand name by running scheduled sweeps across web search and Instagram, capturing each apparent use with the URL, the surrounding text and the date observed, and delivering a dated evidence file as a comment on the task. It does not assess infringement, does not contact anyone, and does not decide whether a use is a problem. A lawyer does that. - **Connections:** web search · Instagram · Google Drive - **Captures:** URL, context, date observed - **Assesses infringement:** Never ## The problem with brand monitoring is that you find out too late to have options By the time someone in the company notices another business using a name close to yours, they have usually been trading under it for a year, have customers, and have a story about how they got there first. The options available at month two are very different from the options at month fourteen. Nobody runs the searches, because running them is dull and finding nothing is the normal result. That is a good description of a job to assign to a machine that keeps a log. The value is entirely in the evidence file: what was found, where, and on what date, captured at the time rather than reconstructed later from memory. ## What a sweep covers and what it deliberately does not - **Covered: the open web** — Web search across the terms and variants you nominate, including common misspellings and the phrase forms you actually use in market. - **Covered: Instagram** — Account names and public posts using the mark, which is where a consumer brand issue usually shows up first. - **Covered: your own historical use** — Evidence of your own public use over time, which is worth capturing while it is easy. - **Not covered: trademark registers** — Official register searching is a specialist task that belongs with a trademark attorney or an agent. Polaris does not search registers and does not pretend the web is a substitute. - **Not covered: any contact with the other party** — The worker never sends a message. A letter about a mark is a legal act with consequences, and it comes from a lawyer. ## What each captured item records | Element | Purpose | | --- | --- | | The URL and account handle | So it can be re-checked and, if needed, produced later | | The surrounding text, quoted | Context is what distinguishes a passing mention from a commercial use | | The date observed | The single most important field, and the one memory gets wrong | | What the worker searched to find it | So a sweep can be reproduced or widened | | A note that nothing was assessed | The item is a finding, not an allegation | > **Nothing here is a view on infringement** > > Whether a use of a mark is infringing depends on classes, territories, registration status, similarity and use in trade, and it is a legal assessment. The worker records that a use exists and when it was seen. Polaris does not give legal advice, does not search official registers, and does not tell you whether to act. ## Why it is affordable to run this monthly A sweep that finds nothing is cheap, and finding nothing is the outcome you want. - **$0** — Cost of the workspace holding the evidence. Unlimited docs and files - **~$2** — Per human-equivalent hour delivered. A quiet monthly sweep is a short job - **0** — Messages sent to a third party. The worker contacts nobody ## Questions people ask **Can it search trademark registers?** No. Register searching is not in the connection catalog and it is genuinely specialist work. This is public-web and Instagram monitoring, which is complementary to a proper watch service, not a replacement for one. **What should we do when it finds something?** Read the evidence, and if the use looks commercially relevant, put it in front of a trademark lawyer with the dated file attached. The point of capturing evidence on the day is that the lawyer's options are wider when the record is contemporaneous. **Will it produce a lot of false positives?** At first, yes, particularly if your brand name is an ordinary word. Tighten the search terms in the SKILL.md over the first two or three runs, and keep the excluded terms visible in the file so future readers know what was deliberately filtered out. **Where is the evidence stored?** As files attached to the delivery comment on the task, and in Google Drive if you connect it. Docs and files in Polaris are versioned, so the record of what was found and when is itself auditable. ## Related - https://www.polarishq.co/use-cases/legal - https://www.polarishq.co/use-cases/legal/policy-updates - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/use-cases/legal/compliance-checklists - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/integrations/instagram - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks --- --- title: "Operations Use Cases for AI Workers | Polaris" description: "Five operations jobs an AI worker handles in Polaris: process docs, procurement, inventory alerts, SOP maintenance and cross-team dependency tracking." url: https://www.polarishq.co/use-cases/operations section: Use cases updated: 2026-08-21 --- # Operations work with an AI worker doing the chasing Ops is the function that holds the seams together. Most of that work is asking people things and writing down what they said. ## The short answer Polaris gives an operations team an AI worker for the work that holds a company's seams together: writing down processes that only exist in someone's head, running a procurement comparison, watching stock levels against sales, keeping standard operating procedures accurate, and tracking dependencies between teams. The worker chases people in Slack and reports what it found. Operational decisions stay with the person accountable for them. - **Connections used:** Slack · Linear · Supabase · Drive - **Software cost:** $0, unlimited people - **Work cost:** ~$2 per human-hour delivered - **Who closes the task:** A human, always ## Operations is the function most damaged by tool sprawl An ops lead's job is to know the state of things across every other function. When tasks live in one tool, documents in a second, and the conversation where the decision was actually made in a third, that job becomes a full-time act of reassembly. Polaris puts the three in one place, which removes some of the reassembly. The worker removes the rest: it goes and asks, it reads what the tools already said, and it writes the answer where the task is. What it does not do is decide. An ops decision has consequences for people and money, and the person accountable makes it. ## The five operations jobs covered here - **Process documentation** — Getting a process out of one person's head and into a doc, by interviewing them and reading the threads where it actually happens. - **Vendor procurement** — A comparison built from real quotes and public pricing, with the questions you have not asked yet listed. - **Inventory tracking** — Stock against sales, reorder points watched, and an alert that arrives before you are out. - **SOP maintenance** — Finding where the written procedure and the real procedure have drifted apart. - **Cross-team coordination** — Dependencies between teams tracked as items with owners and dates instead of as promises made in a meeting. ## What each job needs and what comes back | Job | Connections | Delivered as | | --- | --- | --- | | Process documentation | Slack, Google Drive, Docs | A versioned process doc with open questions listed | | Vendor procurement | web search, Gmail, Google Drive | A comparison table plus the unasked questions | | Inventory tracking | Supabase, Stripe, Slack | A stock report with reorder flags and a days-of-cover figure | | SOP maintenance | Docs, Slack, Google Drive | A drift report naming steps that no longer match reality | | Cross-team coordination | Slack, Linear, Google Calendar | A dependency list with owners, dates and what is blocked | ## The trade an ops team is actually making **Before** - State reassembled by hand from four tools every Monday - Processes documented once, then never again - Dependencies tracked as things people said in a meeting - Chasing done by whoever is least embarrassed to chase - A per-seat bill for each tool holding a piece of the picture **After** - Tasks, docs and the conversation in one workspace - A worker that re-reads and re-asks on a schedule - Dependencies as dated items with named owners - Chasing done daily by something that never gets tired of it - Free software, and a bill only for delivered work > **Ops workers earn their place by asking, not by knowing** > > The useful pattern in every job on this page is the same: the worker goes and asks a person a specific question, waits, asks again, and writes down what came back with the source attached. That is unglamorous and it is exactly the work that does not happen when everyone is busy. ## Operations use cases in detail - [use-cases/operations/process-documentation](https://www.polarishq.co/use-cases/operations/process-documentation) - [use-cases/operations/vendor-procurement](https://www.polarishq.co/use-cases/operations/vendor-procurement) - [use-cases/operations/inventory-tracking](https://www.polarishq.co/use-cases/operations/inventory-tracking) - [use-cases/operations/sop-maintenance](https://www.polarishq.co/use-cases/operations/sop-maintenance) - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) ## Questions people ask **Do we need to move everything into Polaris first?** No. Workers read the connections you give them, so a team still running Linear and Slack can put an ops worker on top of both. The teams that get the most out of it tend to consolidate over time because the tasks, the docs and the deliveries end up in one place anyway. **How many workers does an ops team need?** Start with one and give it one job for a fortnight. Hiring a worker takes about a minute in chat, so there is no reason to plan a roster in advance. Add a second when you can name the job it would own. **What happens to the worker's output if we stop using it?** It is docs, tasks and files in your workspace. The deliveries are comments on tasks and the documents are versioned, so the record does not depend on the worker still being on the roster. **Is there a per-seat cost for the ops team?** No seats and no tiers. Polaris is free for unlimited humans, tasks, workstreams and docs. The only bill is delivered work from AI workers at roughly $2 per human-equivalent hour, itemized on the work log. ## Related - https://www.polarishq.co/use-cases/operations/process-documentation - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/for/small-business --- --- title: "AI Process Documentation from Real Work | Polaris" description: "An AI worker reads the Slack threads where a process really happens, interviews the owner, and drafts a versioned doc with the open questions at the top." url: https://www.polarishq.co/use-cases/operations/process-documentation section: Use cases updated: 2026-08-21 --- # Getting a process out of one person's head The process exists. It is in someone's habits and in six months of Slack threads. A worker interviews and reads until it is on a page. ## The short answer An AI worker in Polaris documents a process by reading the Slack threads where it actually happens, asking the person who owns it a short list of specific questions, and writing a versioned process doc with every unresolved point listed as an open question at the top. The owner reads the draft, answers the open questions, and approves it. What the process should be remains their decision. - **Connections:** Slack · Google Drive - **Source material:** Real threads, not a blank template - **Every draft opens with:** The questions it could not answer ## Nobody writes documentation because writing it means admitting you are not sure Ask the person who owns a process to document it and they will start, get four steps in, hit the part where it depends on what the customer said, and stop. That branch has never been written down because writing it means deciding it, and deciding it is a bigger job than they had time for today. So the process stays undocumented, and it leaves with them. A worker gets past this by starting from evidence rather than from a blank page. It reads the last ten times the process ran in Slack, drafts what it observed, and puts the branch it could not resolve at the top as an explicit question. Answering a specific question is a two-minute job. Writing a document is not. ## How a documentation run goes 1. **Name the process and where it happens** — The brief needs one thing to be precise: which Slack channels and which folders is this process visible in. A worker pointed at everything produces a document about nothing. 2. **The worker reads the last several real instances** — It looks for the actual sequence, the people involved, the decisions taken and the exceptions, and it notes how often each exception occurred rather than treating it as a rule. 3. **It asks the owner a short list of specific questions** — Posted as a comment on the task. Specific means answerable in a line: does the finance check happen before or after the customer is told, and who does it when the owner is away. 4. **It drafts the doc with open questions at the top** — Docs in Polaris are a nested tree with markdown shortcuts, to-dos and sub-pages, and they are versioned, so the draft has a history from the first version onward. 5. **The owner answers, approves and the doc becomes the reference** — The human closes the task. The doc is then the input to SOP maintenance, which is what stops it going stale. ## What makes the resulting document actually get used - **Frequencies, not just steps** — "This exception happened in 3 of the last 10 runs" tells a reader far more than a step listed as if it were routine. - **Named people for the branches** — Every process has a step that in practice means asking a particular person. Naming them is honest, and it identifies your single points of failure. - **The open questions kept visible** — Do not delete the question list when the document is approved. Move it to the bottom as the record of what was decided and when. - **A link to the real threads** — A documented step that cites the conversation where it was decided survives an argument six months later. ## Describing versus designing **The worker** - Describes what actually happens today - Counts how often exceptions occur - Names who is involved at each step - Asks about the parts it could not resolve - Writes and versions the document **The process owner** - Decides what the process should be - Resolves the branches - Approves the document - Decides who is accountable for each step - Closes the task > **The first draft being wrong is the point** > > A worker's description of your process will contain something that makes the owner say "no, we stopped doing that in March". That reaction is the fastest documentation review anyone has ever run, and it is much easier to get than a paragraph written from scratch. ## Questions people ask **What if the process does not happen in Slack?** Then the worker has less to read and the interview carries more weight. Say so in the brief and expect a longer question list. Processes that leave no trace anywhere are the ones most worth documenting and the slowest to capture. **Will it invent steps it did not observe?** Brief it to write only what it saw and to list everything else as an open question, and check the first draft against that instruction. A documentation worker whose gaps are invisible is worse than no documentation at all. **How is this different from SOP maintenance?** This creates the document from evidence for the first time. SOP maintenance keeps an existing document honest as reality drifts away from it. Most teams need this once per process and maintenance forever after. **Can the document live somewhere else?** It can be exported, but keeping it in Polaris docs means the review task, the process owner and the version history are in the same place. A process doc in a separate wiki is a process doc that will be out of date by the second quarter. ## Related - https://www.polarishq.co/use-cases/operations - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files --- --- title: "AI-Assisted Vendor Procurement Research | Polaris" description: "An AI worker researches vendors on the web, pulls real quotes out of Gmail, and builds a sourced comparison plus the questions nobody has answered." url: https://www.polarishq.co/use-cases/operations/vendor-procurement section: Use cases updated: 2026-08-21 --- # Procurement comparisons built from quotes, not from vendor websites A worker researches the field, collects what the vendors actually told you, and lists the questions you have not asked yet. You choose. ## The short answer An AI worker in Polaris supports procurement by researching the available options with web search, collecting the quotes and answers vendors sent to your Gmail, and building a comparison table where every cell cites its source. It separates published pricing from what you were actually quoted, and lists the questions no vendor has answered. The buying decision and the negotiation stay with a person. - **Connections:** web search · Gmail · Google Drive - **Every cell:** Cites its source - **Separated:** Published price vs quoted price ## The comparison spreadsheet is where procurement quietly goes wrong Someone builds a grid of five vendors across nine criteria. Three of the cells come from a marketing page, two come from a sales call, one is a guess, and none of them says which is which. The grid then gets shown to a decision-maker who treats every cell as equally solid. The fix is not a better grid. It is provenance: every cell carrying where it came from, so a reader can see instantly that the security answer came from a vendor's own website and the price came from a written quote. That is dull work to do by hand and easy for a worker, which reads, cites and stops short of forming a preference. ## How the comparison is built | Row source | How it is marked | How much it should count | | --- | --- | --- | | A written quote in your mailbox | Quoted, with the sender and date | The only pricing evidence worth planning on | | Public pricing page | URL and date observed | A starting point that often does not survive a call | | A claim on a vendor's marketing site | Quoted, marked as vendor-stated | Useful for what they choose to emphasize, not as fact | | An answer from a sales call | Only if someone wrote it down and it is in the thread | As good as the note taken | | No answer from anyone | Listed as an open question, not left blank | This is the row that decides the deal later | ## The questions a worker will surface that nobody asked Ask it to produce this list explicitly. It is usually more valuable than the comparison itself. - **What happens at renewal** — Published pricing rarely mentions the increase mechanism, and it is a question that costs money if you skip it. - **What the exit looks like** — Data export format, notice period and what happens to your records afterwards. - **Which stated capability is actually available on the plan you are quoted** — The gap between the feature page and the tier in the quote is a recurring source of unpleasant surprises. - **Who else in the company already pays this vendor** — The worker can check the vendor register and the charges before you buy a second contract with the same supplier. > **The worker does not have a favorite** > > Brief it to compare and to ask, not to recommend. A recommendation from a machine tends to be either over-weighted or dismissed as noise, and both outcomes waste the research. A sourced table plus a list of unanswered questions gives a human everything they need to make the call themselves. ## Questions people ask **Can the worker contact vendors for quotes?** It drafts the enquiry and a person sends it. Nothing is sent to a third party by a worker, which also means no vendor gets a machine-written enquiry with your company name on it that nobody read first. **How current is the pricing research?** It reads public pages at the time it runs and records the date it observed them. Published pricing changes without notice, which is why the comparison marks a written quote as a different class of evidence from a web page. **What happens after we choose?** The signed agreement goes into the vendor register, where the finance view tracks the spend and the renewal and the legal view tracks the terms. Running procurement in the same workspace means the decision, the quotes and the resulting contract stay linked. **Is this worth it for a small purchase?** Probably not for a small one-off. It earns its place on recurring commitments and anything with an auto-renewal, where the cost of an unasked question compounds annually. ## Related - https://www.polarishq.co/use-cases/operations - https://www.polarishq.co/use-cases/finance/vendor-management - https://www.polarishq.co/use-cases/legal/vendor-agreement-management - https://www.polarishq.co/use-cases/operations/inventory-tracking - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/integrations/gmail --- --- title: "AI Inventory Alerts from Your Own Data | Polaris" description: "An AI worker reads your Supabase stock table and Stripe sales, computes days of cover per item, and raises an owned reorder task before you run out." url: https://www.polarishq.co/use-cases/operations/inventory-tracking section: Use cases updated: 2026-08-21 --- # Stock alerts that arrive before you are out Polaris does not hold your inventory. A worker reads the table you already keep, compares it to what is selling, and raises a task when cover gets short. ## The short answer An AI worker in Polaris tracks inventory by reading the stock table you keep in Supabase alongside sales in Stripe, computing days of cover per item, and raising a task with an owner whenever cover falls below the threshold you set. It posts the stock report as a comment and alerts in Slack. Polaris is not an inventory system and does not write to your stock records. - **Connections:** Supabase · Stripe · Slack - **Computes:** Days of cover per item - **Writes to your stock records:** Never. Read only ## Most small operations already have the data and no alarm on it A stock table exists. Sales exist. What does not exist is anything that looks at both on a Tuesday morning and notices that the item selling four times faster than last month has eleven days of cover and a six-week lead time. That gap is why reorder decisions get made in a panic, at a worse price, with expedited shipping. A worker closes it by doing a boring calculation on a schedule and turning the result into an owned task rather than a notification. A notification gets dismissed. A task with a name and a date on it in the Focus lane does not. ## What the worker computes and reports - **Days of cover per item** — Current stock divided by the recent sales rate, using the window you specify rather than a lifetime average that hides a trend. - **Items where the rate changed sharply** — A doubling in velocity matters more than an absolute level, and it is the signal a static reorder point misses. - **Cover measured against lead time** — Thirty days of cover is comfortable at a one-week lead time and an emergency at a ten-week one. The lead time comes from your data, not from an assumption. - **Items with no movement** — The other half of the problem, and the half nobody asks about. ## Where each number comes from | Input | Source | What it must contain | | --- | --- | --- | | Current stock | Your Supabase table | An item identifier, a quantity, a timestamp | | Sales rate | Stripe, or your own sales table in Supabase | Enough history to cover the window you chose | | Lead time | Your own data or the vendor register | Days, per item or per supplier | | Threshold | You set it in the acceptance criteria | A number of days, per item class | | Owner | You set it | A named person who receives the reorder task | ## What Polaris is and is not here **What it does** - Reads your existing stock and sales data - Does the arithmetic on a schedule - Raises an owned, dated task when a threshold is crossed - Posts a readable stock report as a delivery comment - Alerts in Slack where the team already is **What it is not** - An inventory management system - A warehouse or fulfilment tool - A writer to your stock records - A purchase-order system - A replacement for counting what is on the shelf > **The arithmetic is only as good as your stock table** > > If the recorded quantity drifts from what is physically there, the worker will report cover figures with confidence and they will be wrong. Brief it to state the timestamp of the stock data in every report, so a reader can see immediately when they are looking at numbers from a count three weeks old. ## Questions people ask **Do we need Supabase to use this?** You need your stock data somewhere the worker can read it, and Supabase is the database connection in the fixed catalog. Teams already keeping stock in a spreadsheet in Google Drive can point the worker there instead, with the same caveat about how fresh the numbers are. **Can it place the reorder?** No. It drafts the purchase enquiry and raises the task. Committing company money to a supplier is a human action, and the same rule applies here as everywhere else in Polaris: agents deliver, humans close. **How often should it run?** Daily for fast-moving items, weekly for everything else. Billing is by delivered human-equivalent hour, so a short daily arithmetic job is inexpensive, and each run is logged with its own line you can check. **What if we have multiple locations?** Then the stock table needs a location column and the acceptance criteria need to say whether cover is computed per location or in aggregate. Getting that decision written down explicitly is worth more than any feature. ## Related - https://www.polarishq.co/use-cases/operations - https://www.polarishq.co/use-cases/operations/vendor-procurement - https://www.polarishq.co/use-cases/data/data-quality-monitoring - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/for/ecommerce-brands --- --- title: "Keeping SOPs Accurate with an AI Worker | Polaris" description: "An AI worker compares each SOP against how the work really ran in Slack and task history, then reports every step that has drifted. The owner decides." url: https://www.polarishq.co/use-cases/operations/sop-maintenance section: Use cases updated: 2026-08-21 --- # Finding where the SOP and the real procedure came apart Every written procedure starts accurate and drifts. A worker compares the document to how the work actually ran and reports the difference. ## The short answer An AI worker in Polaris maintains standard operating procedures by comparing each written document against how the work actually ran in Slack and in task history, then reporting drift: steps nobody performs, steps people perform that are not written down, and named owners who have changed role. The document owner decides whether to change the procedure or correct the practice. - **Connections:** Slack · Google Drive · Docs - **Reports:** Steps skipped, steps added, owners moved - **Cadence:** Per review interval you set ## Drift is invisible until the wrong person has to follow the document The people who run a procedure daily stopped needing the document years ago. They know the current version because they are the current version. The document sits there, quietly describing a step that was removed and a person who left. It gets discovered on the day it matters most: a new hire follows it, or someone covers a holiday, or an auditor asks to see it. Comparing a document to reality is a genuinely tedious job with a clear method, which is a good description of what to give a worker. It reads the document, reads how the work actually ran, and lists the differences without deciding which side is correct. ## The two kinds of drift, and why they are different problems Both show up in the same report, and each needs a different human response. **The document is behind** - A step in the doc that nobody has done in months - A tool named that the team stopped using - An owner who has changed role or left - An approval that was removed and never deleted from the page **The practice has slipped** - A check in the document that people skip when busy - A step performed in a different order than written - An approval routinely done after the fact - An exception that has quietly become the normal path ## What the drift report contains - **The step, quoted from the document** — With the version it came from, since docs in Polaris are versioned. - **The evidence, or the absence of it** — Links to the runs where the step appeared, or a plain statement that no instance was found in the window examined. - **The window examined** — "No instance in the last 20 runs" is a finding. "No instance found" without a window is not. - **A question for the owner** — Should the document change, or should the practice? The worker asks. It does not answer. ## Where the ownership sits | Action | Who | | --- | --- | | Detect the drift and evidence it | The worker | | Decide whether the step still matters | The document owner | | Update the document | The worker drafts, the owner approves | | Correct the practice | The manager responsible for the work | | Retire a procedure entirely | The owner, on the record, with a date | > **Report drift, never resolve it** > > A worker that silently updates the document to match what people do has just converted a control failure into a documented policy. Every drift item comes back as a question with evidence attached, and a human decides which side moves. That rule is worth writing into the SKILL.md in exactly those words. ## Questions people ask **How does the worker know how the work actually ran?** From the trace the work leaves: the task history in the workspace and the Slack threads where the steps get discussed. Where a procedure leaves no trace anywhere, the worker will say it could not evidence the step rather than assuming it happened. **How often should SOPs be re-checked?** Give each document a review interval and let it become a recurring task with an owner in the Focus lane. Quarterly suits most procedures; anything tied to a control or an audit commitment usually deserves more. **Can it write a new SOP from nothing?** That is the process documentation job rather than this one. Documentation creates the first version from evidence; maintenance keeps it honest afterwards. Most teams need the first once and the second forever. **What if two teams follow the same SOP differently?** The report will show it, which is usually the most useful thing it finds. Whether they should converge is a decision for the person who owns the procedure, and it is a much easier conversation with the evidence in front of both teams. ## Related - https://www.polarishq.co/use-cases/operations - https://www.polarishq.co/use-cases/operations/process-documentation - https://www.polarishq.co/use-cases/legal/compliance-checklists - https://www.polarishq.co/use-cases/hr/policy-documentation - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "Cross-Team Dependency Tracking | Polaris" description: "An AI worker turns cross-team commitments in Slack and Linear into owned, dated tasks, chases the late ones, and reports weekly on what is blocked." url: https://www.polarishq.co/use-cases/operations/cross-team-coordination section: Use cases updated: 2026-08-21 --- # Dependencies as dated items, not as things people said in a meeting Two teams agree something in a call. Neither writes it down. Three weeks later each is waiting for the other. A worker makes the promise a tracked item. ## The short answer An AI worker in Polaris tracks cross-team dependencies by turning commitments made in Slack and in Linear into tasks with a named owner and a date, chasing whoever is late, and posting a weekly list of what is blocked and by whom. Slack messages arrive in the Polaris Inbox as prefilled task suggestions that a human accepts. Nothing becomes a commitment without someone clicking accept. - **Connections:** Slack · Linear · Google Calendar - **Every dependency has:** An owner, a date, a blocked item - **Suggestions become tasks:** Only when a human accepts ## The most expensive delay in a company is two teams politely waiting for each other It starts as a reasonable exchange. Design says they will have the final assets once the copy is signed off. Marketing hears that as a commitment for next week. Nobody writes a date. Both teams move on to other work and are entirely certain they are not the blocker. The delay only becomes visible when a launch date is missed, at which point it is a conversation about blame rather than about a missing dependency. The intervention that works is unglamorous: someone writes the promise down with a name and a date on it, and asks about it before the date arrives. That is a worker's job description. ## How a commitment becomes a tracked item The Inbox catches what your tools hear. Signals become suggestions, never silent tasks. - **A Slack message arrives as a prefilled task suggestion** — With the workstream, lane, labels and owner already set. One click makes it a real task, and nothing is created without that click. - **The worker adds the missing date** — Or asks for it. A dependency without a date is a wish, and the worker's first question is always when. - **The blocked item is linked** — Knowing that something is late matters much less than knowing what stops moving because of it. - **The chase runs on a schedule** — Before the date, not after it. A dependency chased the morning it comes due is often still recoverable. ## The weekly dependency report One comment, one shape, every week, so the pattern over a quarter is readable. | Section | What it lists | The conversation it starts | | --- | --- | --- | | Due this week | Commitments with a date inside the window and their owners | Is anything about to slip | | Overdue | How many days late, and what it is blocking | Is this the same team every week | | No date agreed | Commitments that were made without one | Which of these are real and which were politeness | | Cleared since last week | What actually moved | The part teams never get told, and it matters | | Recurring blockers | The same dependency appearing across weeks | This is a structural problem, not a chasing problem | > **The worker chases the item, not the person** > > Brief it to name the commitment, the date and what is blocked, and to leave characterizations of anybody's reliability out of it. A chase that reads as neutral gets answered. A chase that reads as an accusation gets escalated into a meeting about tone, which is precisely the meeting nobody needed. ## Why this lives in one workspace Cross-team work is where tool sprawl does the most damage, because each team keeps its own version of the truth. - **3** — Focus horizons every team shares. Today, This week, Next 30 days - **1** — Click to turn a Slack signal into a task. The suggestion arrives prefilled - **$0** — Cost per additional person or team. No seats, unlimited humans ## Questions people ask **What if the other team uses Linear and we do not?** Give the worker the Linear connection. It reads what is there and tracks the dependency in Polaris, so the two teams do not have to agree on a single tracker before they can agree on a date. **Does every Slack message become a task?** No, and that distinction is deliberate. The Inbox turns a signal into a suggestion with the fields prefilled, and a human accepts it. A tool that silently creates tasks from chat produces a backlog nobody trusts within a fortnight. **Who should own the dependency report?** One named person, usually in ops, with the report as an assigned task rather than a message in a channel. Reports addressed to everyone are read by no one. **Can the worker escalate?** It can follow the escalation path you write in the SKILL.md: chase the owner, then their manager after a set number of days, then flag to the ops lead. Because the file is plain text, everybody can read the rules being applied to them, which is what makes escalation acceptable. ## Related - https://www.polarishq.co/use-cases/operations - https://www.polarishq.co/use-cases/operations/process-documentation - https://www.polarishq.co/use-cases/executive/meeting-follow-ups - https://www.polarishq.co/use-cases/engineering/sprint-planning - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/focus-lane --- --- title: "AI Workers: 18 Roles You Can Hire in Polaris" description: "Hire an AI worker in about 60 seconds, give it a role, tools and a first task, and pay $2 per human-hour it delivers. Eighteen roles, each with a job spec." url: https://www.polarishq.co/ai-workers section: AI workers updated: 2026-08-21 --- # AI workers you can hire, brief and assign Every role below is a teammate you hire in chat, brief with an editable skill file, and assign work to on the same board as your people. ## The short answer An AI worker in Polaris is a teammate you hire through a short chat interview that produces an editable SKILL.md file, connect to the tools it needs, and assign tasks to on the same board as your people. When a task is assigned, a cloud machine wakes, researches, drafts and posts the deliverable as a comment. A person closes the task and rates the work. - **Roles on this page:** 18 - **Time to hire one:** About 60 seconds - **Software cost:** $0 - **Delivered work:** $2 per human-hour ## What you are actually hiring A Polaris worker is a row in the same members table as your colleagues, with kind set to agent instead of human. Assignment, due dates, labels, comments and the Focus lane behave identically for both. That is the whole architectural trick, and it is why a worker can own a task rather than sit behind a chat box waiting to be prompted. The role you pick decides what goes into the worker's skill file. A research analyst is briefed to source every claim and to name what it could not verify. A bookkeeper is briefed to prepare a categorisation for a person to approve and never to touch a ledger of record. The brief is a file you can open, read and edit, not a hidden prompt. ## How hiring works 1. **Say what is falling through the cracks** — Tell your Chief of Staff, in chat, what nobody is picking up. It suggests a role and a name. 2. **Answer the interview in clicks** — A name, the role, what the worker should be great at, which tools it needs. Option pills and multi-select, no forms. 3. **Read the skill file** — The answers become a real SKILL.md. Open it, sharpen the wording, add a rule you would otherwise repeat every week. 4. **Authorise the connections** — Pick from the fixed catalog. Credentials are verified live and stored server-side, once, for the whole org. 5. **Assign the first task** — The worker card lands in chat ready for work. Give it acceptance criteria and a due date, exactly as you would a person. ## The eighteen roles - [Hire an AI research analyst](https://www.polarishq.co/ai-workers/research-analyst) — Assign the question on Tuesday afternoon and read the brief before your Wednesday call, with every claim carrying the URL it came from. - [Hire an AI support specialist](https://www.polarishq.co/ai-workers/support-specialist) — Hand it the backlog on Friday and read eighteen drafted replies, sorted by severity, before you send a single one. - [Hire an AI content writer](https://www.polarishq.co/ai-workers/content-writer) — Briefs go in as tasks, drafts come back as versioned docs you can comment on line by line, and your voice rules live in a file rather than in one person's head. - [Hire an AI bookkeeper](https://www.polarishq.co/ai-workers/bookkeeper) — It does the sorting, chasing and drafting that eats a finance afternoon, and it stops at the line where a qualified person has to sign. - [Hire an AI recruiter](https://www.polarishq.co/ai-workers/recruiter) — Every candidate is scored against the same written rubric, and the rubric is a file your hiring manager wrote and can change. - [Hire an AI SDR](https://www.polarishq.co/ai-workers/sdr) — It does the twenty minutes of account research nobody has time for, and hands you a first line that could only have been written about that company. - [Hire an AI QA engineer](https://www.polarishq.co/ai-workers/qa-engineer) — Fourteen vague reports go in, fourteen tickets with steps, expected behaviour and a severity come out, with the three duplicates already merged. - [Hire an AI data analyst](https://www.polarishq.co/ai-workers/data-analyst) — It settles what active user means, writes the query that matches the definition, and turns the numbers you hand it into a paragraph an executive can read. - [Hire an AI social media manager](https://www.polarishq.co/ai-workers/social-media-manager) — A month of posts planned, written and filed by Tuesday, so the only decision left is which ones you actually publish. - [Hire an AI executive assistant](https://www.polarishq.co/ai-workers/executive-assistant) — Every commitment made in a meeting becomes a task with a name and a date on it before the next meeting starts. - [Hire an AI project coordinator](https://www.polarishq.co/ai-workers/project-coordinator) — The unglamorous half of running projects, done every week without anyone having to be the person who nags. - [Hire an AI SEO specialist](https://www.polarishq.co/ai-workers/seo-specialist) — It looks at what is actually ranking today, then writes the brief that says what your page has to do differently to deserve the spot. - [Hire an AI technical writer](https://www.polarishq.co/ai-workers/technical-writer) — The documentation debt on your team is not a writing problem, it is a nobody-has-two-free-hours problem, and this is the worker for those two hours. - [Hire an AI competitive analyst](https://www.polarishq.co/ai-workers/competitive-analyst) — Assign it every Monday and you get a written record of what your market did last week, with a URL behind every line of it. - [Hire an AI customer success manager](https://www.polarishq.co/ai-workers/customer-success-manager) — It reads everything you know about an account and writes the two paragraphs you wish you had before the renewal call. - [Hire an AI paralegal](https://www.polarishq.co/ai-workers/paralegal) — It organises the paperwork and surfaces the dates and clauses, and it stops well before the point where anything becomes advice. - [Hire an AI financial analyst](https://www.polarishq.co/ai-workers/financial-analyst) — You supply the figures; it writes the explanation of what moved, why it matters and which assumption the whole thing rests on. - [Hire an AI ops coordinator](https://www.polarishq.co/ai-workers/ops-coordinator) — It writes down how your company does things, so the answer stops living in whichever colleague happens to be on holiday. ## The division of labour The same split shows up on every role page, because it is the rule the product enforces. **The worker does** - Research with live web search and a source trail - Draft, in a versioned doc you can comment on - Produce files: PDF, CSV, JSON, markdown - Tick the acceptance criteria it actually met - Post progress, then post the deliverable as a comment **A person does** - Set the acceptance criteria and the due date - Decide, approve, sign and send - Close the task, which the machine can never do - Rate the delivery, which is the worker's review - Challenge any line on the bill from the work log ## Hiring a worker versus buying another seat | | A new SaaS seat | A Polaris AI worker | | --- | --- | --- | | Time to onboard | Procurement, then a login | About 60 seconds, in chat | | What it costs | A monthly fee whether used or not | $2 per human-hour delivered, nothing if nothing ships | | Where the brief lives | In someone's head | In an editable SKILL.md | | Proof of work | A dashboard | A work log, itemised and challengeable | | Who signs off | Nobody in particular | The human who owns the outcome | > **Polaris is in free public beta** > > There are no customer counts to quote and no case studies yet. The software is free, the roles below are real, and the pricing is the same for everyone: $2 per human-hour of delivered work, itemised job by job. ## Questions people ask **How many AI workers can one team hire?** There is no seat limit and no per-worker fee, so a team can hire as many roles as it has work for. Cost is driven only by delivered work at $2 per human-hour. A worker that is never assigned a task never appears on the bill. **Can two workers share the same connection?** Yes. Connections are authorised once for the whole organisation and stored server-side, then granted to individual workers. A research analyst and a competitive analyst can both hold web search without anyone re-entering a credential. **What happens if a worker delivers something wrong?** You comment on the task the way you would with a colleague, and the latest human comment becomes the worker's new brief on the next run. The task stays open because a machine cannot close it. If the delivery was not worth the hours logged, challenge the line from the work log. **Do I need to know how to prompt an AI to use these?** No. The hiring interview is answered in clicks and produces the skill file for you. After that you write tasks the way you write them for people: a title, a short description, and a checklist that acts as the acceptance criteria. **Which role should a small team hire first?** Pick the work that is already late every week. For most small teams that is research briefs, support replies or written documentation, which map to the research analyst, support specialist and technical writer pages. Every org also starts with a Chief of Staff on the roster at no extra cost. ## Related - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "AI Research Analyst You Can Hire in 60 Seconds" description: "An AI research analyst that returns a sourced brief with a live-web citation trail, ticks its own acceptance criteria, and costs $2 per human-hour delivered." url: https://www.polarishq.co/ai-workers/research-analyst section: AI workers updated: 2026-08-21 --- # Hire an AI research analyst Assign the question on Tuesday afternoon and read the brief before your Wednesday call, with every claim carrying the URL it came from. ## The short answer An AI research analyst in Polaris is a worker you hire in chat to answer open questions with sourced written briefs. Given a question and acceptance criteria, a cloud machine runs live web searches, drafts the brief in a versioned doc, attaches a typeset PDF when the work is board-facing, lists its sources as plain URLs, and posts the result as a comment for a person to close and rate. - **Typical output:** Sourced brief, doc plus PDF - **Core connection:** Web search - **Illustrative task:** 3.4 human-hours, $6.80 ## What this worker is great at Turning a vague question into a written answer somebody can act on. You assign “what are the three cheapest ways to run scheduled jobs on Cloudflare, and what breaks at each price point” and you get back a brief that names the options, quotes the prices it found, and links each number to the page it came from. The brief that comes back has a shape, because the shape lives in the skill file. Claims carry sources. Conflicting sources are flagged rather than averaged into mush. There is a short section on what the analyst could not verify, which is the part most human briefs quietly omit and the part that stops you presenting a guess as a finding. Written work is drafted in Docs by default, so it lands as a versioned page with comments rather than a wall of chat text. When the brief is going in front of a board or a client, it also arrives as a typeset PDF. ## What to connect it to Connections come from the fixed catalog, are authorised once for the whole org, and are stored server-side. - **Web search** — The one this role cannot work without. It is what turns an assertion into a sourced claim, and it is the only connection in the catalog that needs no credential. - **Google Drive** — Give it Drive when the research has to sit alongside existing material your team already keeps there. - **Notion** — Useful when the prior art lives in a Notion wiki and you want the analyst working from the same page as everyone else. - **Slack** — Worth adding if the questions arrive as messages: the Inbox turns those signals into prefilled task suggestions you approve with one click. ## A realistic first week 1. **Monday: a competitor pricing brief** — Four named competitors, current published prices, what each price includes, and the URL for every figure. Acceptance criteria: one row per competitor, no unsourced number. 2. **Tuesday: the question your board asked** — Whatever came up in the last meeting that nobody had time to chase. Ask for a two-page answer with a stated confidence level per section. 3. **Thursday: a market primer for a new hire** — The regulations, acronyms and players in a market your team is entering, written for somebody who starts on Monday. 4. **Friday: a follow-up on Monday's brief** — Comment on the task with your three sharpest objections. The latest human comment becomes the brief for the next run, so the second pass answers you rather than restating the first. ## Where the human stays in charge **The analyst prepares** - The question decomposed into answerable parts - Findings with the URL behind each one - A note where two sources disagree - An explicit list of what it could not verify - The brief as a doc, and a PDF when it needs one **You decide** - Whether the finding is good enough to act on - What goes in front of a client or a board - Which recommendation the company actually takes - Closing the task and rating the delivery ## What a brief costs Hours are human-equivalent, estimated from observable effort by the published formula and logged job by job. Illustrative example, computed from that formula rather than measured on a customer. | Observed effort | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 5 live web searches at 12 min each | 60 | | 6,300 characters of finished prose at 90 chars/min | 70 | | 4 acceptance criteria ticked at 8 min each | 32 | | 1 progress comment, 1 doc, 1 PDF | 25 | | Total: 3.4 human-hours at $2 | $6.80 | > **What an AI research analyst is not good at** > > It cannot read what the open web will not show it: paywalled reports, gated PDFs, and anything behind a login are outside its reach, and it will say so rather than guess. It does not conduct interviews, run surveys or phone anyone. A single machine session is bounded to minutes and a handful of searches, so a question that needs forty sources should be split into several tasks. And it is a researcher, not a decision-maker: it will lay out the trade-off and leave the call to you. ## Work this role picks up - [use-cases/executive/strategic-research](https://www.polarishq.co/use-cases/executive/strategic-research) - [use-cases/marketing/competitor-monitoring](https://www.polarishq.co/use-cases/marketing/competitor-monitoring) - [use-cases/product/user-research-synthesis](https://www.polarishq.co/use-cases/product/user-research-synthesis) - [use-cases/sales/lead-research](https://www.polarishq.co/use-cases/sales/lead-research) ## Questions people ask **How does an AI research analyst cite its sources?** Every factual claim carries the URL it came from, listed as plain links in the delivery comment and inside the doc. The runtime records each web search in the work log, so the trail of what was searched and what was read sits next to the finished brief. **Can it research private or internal material?** It works from the open web plus whatever you connect it to and whatever you put in the task. It cannot reach paywalled databases, gated analyst reports or anything requiring a personal login that is not part of the connection catalog. **How long does a research brief take to come back?** A machine session runs in minutes rather than the hours the same brief takes a person, and progress is posted on the task while it works. Long questions are better split across several tasks than pushed into one session. **What stops the analyst from inventing a statistic?** The skill file requires a source for every claim and an explicit note where verification failed, and the work log shows how many live searches were actually run. A brief with no searches logged and confident numbers in it is a brief worth challenging. **Can I edit how the analyst writes briefs?** Yes. The role's capabilities live in a SKILL.md file you can open and edit, so a house rule such as always including a confidence level or never using a source older than eighteen months becomes permanent after you write it once. ## Related - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/use-cases/executive/strategic-research - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/use-cases/product/user-research-synthesis - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks --- --- title: "AI Support Specialist for Your Ticket Queue" description: "An AI support specialist that triages the queue, drafts replies in your tone, and escalates what it should not answer. You send. $2 per human-hour delivered." url: https://www.polarishq.co/ai-workers/support-specialist section: AI workers updated: 2026-08-21 --- # Hire an AI support specialist Hand it the backlog on Friday and read eighteen drafted replies, sorted by severity, before you send a single one. ## The short answer An AI support specialist in Polaris is a worker you hire to triage a support queue and draft replies. Given a batch of tickets, it sorts them by severity, drafts an answer for each in the tone recorded in its skill file, marks the ones a person must handle, and posts the batch as a comment. A human reviews and sends every message. - **Typical output:** Triaged batch with drafted replies - **Core connections:** Gmail, Slack - **Illustrative task:** 3.6 human-hours, $7.20 ## What this worker is great at The queue work that is repetitive without being mindless. - **Triage with a stated rule** — It sorts a batch by severity using the rule written in its skill file, not a private hunch, so you can argue with the rule instead of the outcome. - **Drafting in your tone** — Refund wording, apology wording, the line you always use when a customer is right and the invoice is not. Written once into the skill file, applied every time. - **Spotting the repeat** — Eleven tickets about the same broken export is a bug, not eleven tickets. It names the pattern and drafts the engineering task to go with it. - **Knowing when to stop** — Anything touching identity, money movement or a legal threat comes back flagged for a person rather than answered. - **The weekly theme summary** — A short doc on what customers complained about this week, which is the artefact product teams never get and always want. ## Where the human stays in charge Nothing reaches a customer without a person putting it there. **The specialist prepares** - A severity-sorted queue - A drafted reply per ticket - An escalation list with the reason attached - A bug task written from the repeat pattern **You do** - Read and send, or rewrite and send - Approve refunds, credits and goodwill - Verify identity before anything sensitive - Close the task and rate the batch ## A realistic first week 1. **Day one: the backlog** — Assign the oldest forty tickets with one acceptance criterion per outcome: answered, escalated, or duplicate of a known bug. 2. **Day two: the tone file** — Read the drafts, fix three sentences you would never write, and paste those corrections into the skill file. That is the whole training loop. 3. **Day three: response templates** — Ask for the eight replies you send most often, written as reusable templates and filed as a doc. 4. **Friday: the theme summary** — One page on what broke this week, with ticket counts per theme, ready to paste into the product channel. ## What a batch costs Illustrative: eighteen replies drafted in one session, priced by the published human-hours formula. Every line appears on the work log and can be challenged there. - **18** — Replies drafted. One session, one task - **3.6** — Human-equivalent hours. From prose length, ticks and comments - **$7.20** — Total cost. At $2 per human-hour ## What an AI support specialist is not good at It is not a live chat agent. A worker runs when a task is assigned to it, so it clears queues in batches rather than answering the customer who is typing right now. If your promise is a sixty-second first response, keep a person on the front line and give the batches to the worker. It cannot verify who somebody is, and it will not try. Account recovery, address changes and anything that moves money stay with a human who can check identity properly. It also has no view of your billing system unless you paste the facts into the task, so it drafts around what it knows and marks the gap. And it has no memory of a customer beyond what is in the task and its skill file. Long relationship history has to be handed to it, or it will write a technically correct reply to somebody who has already been apologised to twice. ## Work this role picks up - [use-cases/customer-support/ticket-triage](https://www.polarishq.co/use-cases/customer-support/ticket-triage) - [use-cases/customer-support/response-templates](https://www.polarishq.co/use-cases/customer-support/response-templates) - [use-cases/customer-support/escalation-tracking](https://www.polarishq.co/use-cases/customer-support/escalation-tracking) - [use-cases/customer-support/customer-feedback-loops](https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops) ## Questions people ask **Does the AI support specialist reply to customers directly?** No. It drafts replies and posts them as a delivery comment on the task, and a person sends them. That boundary is enforced by the product: a machine cannot close a task in Polaris, and nothing leaves for a customer without a human action. **How does it learn our tone of voice?** Tone lives in the worker's SKILL.md, which you can open and edit. The fastest way to train it is to correct two or three drafted sentences and paste your version into the skill file, where it applies to every batch from then on. **Can it handle refunds?** It can draft the refund message and flag the ticket as a refund case, but approving and issuing money stays with a person. Give it the rule you use for goodwill credits and it will tell you which tickets meet the rule. **What happens to tickets it cannot answer?** They come back on the escalation list with the reason attached, rather than receiving a vague holding reply. Legal threats, identity questions and anything the skill file marks as out of scope are always escalated. ## Related - https://www.polarishq.co/ai-workers/customer-success-manager - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/customer-support/response-templates - https://www.polarishq.co/use-cases/customer-support/escalation-tracking - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/skill-file --- --- title: "AI Content Writer That Drafts in Your Docs" description: "An AI content writer that drafts inside versioned docs, keeps your voice rules in an editable skill file, and bills $2 per human-hour of finished writing." url: https://www.polarishq.co/ai-workers/content-writer section: AI workers updated: 2026-08-21 --- # Hire an AI content writer Briefs go in as tasks, drafts come back as versioned docs you can comment on line by line, and your voice rules live in a file rather than in one person's head. ## The short answer An AI content writer in Polaris is a worker you hire to produce written drafts: articles, landing copy, release notes, newsletters. It drafts inside a versioned Polaris doc, applies the voice and structure rules stored in its editable skill file, cites anything factual it looked up, and posts the draft as a comment. An editor comments, the worker revises, a person publishes. - **Typical output:** Versioned draft in Docs - **Core connections:** Web search, Notion, Google Drive - **Illustrative task:** 2.8 human-hours, $5.60 ## What this worker is great at Producing the first eighty per cent of a piece against a real brief, in a place your editor can actually work. The draft arrives as a Polaris doc with version history and inline comments, not as a block of chat text somebody has to copy into a document before it can be edited. The interesting part is the voice rules. Every team has them and almost no team writes them down: no exclamation marks, never say solution, always open on the reader's problem rather than the company, British spelling, one em dash per page at most. In Polaris those rules go in the worker's skill file, which means the ninth draft obeys them as reliably as the first and nobody has to give the same feedback twice. It is also good at the writing nobody volunteers for. Release notes from a list of merged changes, the FAQ page nobody has updated since launch, forty product descriptions with a consistent structure and no two openings alike. ## What to connect it to Four connections from the fixed catalog cover most writing work. | Connection | Why this role needs it | | --- | --- | | Web search | Checking a claim before it goes in a draft, and citing the source next to it | | Notion | Working from the brand guide and the existing wiki instead of guessing at house style | | Google Drive | Where finished copy has to land when the rest of the team already lives in Drive | | Slack | Turning a request in a channel into a prefilled writing task you approve with one click | ## A realistic first week 1. **Monday: one article against a real brief** — Give it the audience, the angle, the three points it must make and the two things it must never claim. Checklist items become the acceptance criteria. 2. **Tuesday: write the voice rules down** — Read the draft, mark what grated, and paste those rules into the skill file. This is the hour that pays for the rest of the week. 3. **Wednesday: release notes** — Hand it the merged changes and ask for user-facing notes that say what changed and who should care. 4. **Friday: a second pass** — Comment on Monday's doc with your edits. The latest human comment is the brief for the next run, so the revision answers your notes rather than rewriting from scratch. > **What an AI content writer is not good at** > > It cannot have had your experience. Founder stories, customer quotes, opinions with a scar behind them and anything that depends on being in the room are not its work, and asking for them produces the flattest paragraphs it writes. It has no view of your traffic or conversion data unless you put the numbers in the task. It will not invent a testimonial, a statistic or a customer, which occasionally reads as a limitation and is in fact the point. ## Where the human stays in charge - **The brief** — Audience, angle and the claims that are off limits are yours to set. A vague brief produces a vague draft faster than a person would produce one. - **The edit** — The worker revises against comments; the editor decides when the piece is finished. - **Publication** — Nothing is published by a worker. The draft sits in a doc until a person moves it. - **The rating** — You close the task and rate the delivery, and that rating is the worker's performance review. ## Work this role picks up - [use-cases/marketing/content-calendar](https://www.polarishq.co/use-cases/marketing/content-calendar) - [use-cases/marketing/seo-content-production](https://www.polarishq.co/use-cases/marketing/seo-content-production) - [use-cases/product/release-notes](https://www.polarishq.co/use-cases/product/release-notes) - [use-cases/sales/sales-enablement-content](https://www.polarishq.co/use-cases/sales/sales-enablement-content) ## Questions people ask **Where does the AI content writer put the draft?** In a Polaris doc, filed under the page whose topic fits, with version history and inline comments. The delivery comment on the task links to the doc, so the review happens in a document rather than in chat scrollback. **How do I stop it writing like every other AI?** Write your banned words and structural rules into the worker's SKILL.md. A short list of forbidden phrases, a required opening move and a sentence-length instruction change the output more than any amount of prompting inside individual tasks. **Can it write in more than one voice?** Hire one worker per voice. Two workers with different skill files, one for the product blog and one for the sales deck, cost nothing extra because there are no per-worker fees, and the briefs stay clean. **What does a 1,100-word article cost?** As an illustrative example computed from the published formula, roughly 2.8 human-equivalent hours, which is $5.60 at $2 per human-hour. The line appears on the work log with the effort it was derived from, and you can challenge it there. ## Related - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/ai-workers/social-media-manager - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/use-cases/marketing/seo-content-production - https://www.polarishq.co/use-cases/product/release-notes - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/glossary/skill-file --- --- title: "AI Bookkeeper That Prepares, You Approve" description: "An AI bookkeeper that categorises transactions, chases missing receipts and prepares the monthly close, for a qualified human to review, approve and file." url: https://www.polarishq.co/ai-workers/bookkeeper section: AI workers updated: 2026-08-21 --- # Hire an AI bookkeeper It does the sorting, chasing and drafting that eats a finance afternoon, and it stops at the line where a qualified person has to sign. ## The short answer An AI bookkeeper in Polaris is a worker you hire to prepare bookkeeping work for human approval. It categorises transactions against rules you write, lists the ones it is unsure about, chases missing receipts as tasks, and drafts the monthly close checklist. A qualified bookkeeper or accountant reviews, approves and files everything. Polaris does not provide accounting or tax advice. - **Typical output:** Categorised CSV plus exceptions list - **Core connections:** Stripe, Gmail, Google Drive - **Illustrative task:** 2.2 human-hours, $4.40 ## What this worker is great at The sorting work. Given a month of transactions and the category rules your accountant actually uses, it returns a categorised file plus a separate list of the ones it would not guess at. The exceptions list is the valuable half: a short, honest column of what needs a human, instead of four hundred rows where a dozen are quietly wrong. It is also good at chasing. Missing receipts become individual tasks with the vendor, the amount and the date already filled in, assigned to whoever spent the money, due before close. The nagging that nobody enjoys becomes a set of dated tasks on a board rather than a series of awkward messages. And it keeps the close checklist honest. The same twenty-two steps every month, ticked as acceptance criteria as each one is genuinely done, with the state of the close visible to anyone who opens the task rather than living in the head of the person doing it. ## What to connect it to Read-only access is the right default for this role. A restricted, read-only Stripe key is safer than a full one. - **Stripe** — Payments, invoices and refunds, which is where most of the month's transactions originate for a software business. - **Gmail** — Receipts and vendor invoices arrive as email. Connect it so the chasing list is built from what is actually missing. - **Google Drive** — Where the categorised files and the close checklist have to end up for your accountant. - **Slack** — So a message about a strange charge becomes a prefilled task suggestion instead of scrolling out of sight. ## Where the human stays in charge This is a regulated field and the boundary is not negotiable. **The bookkeeper prepares** - A proposed category for each transaction - An exceptions list with the reason for each - Receipt-chasing tasks, owned and dated - A drafted close checklist with progress ticked **A qualified person decides** - Whether each category is correct - What is deductible and what is not - Every entry that reaches the ledger of record - Filings, returns and anything sent to a tax authority ## What a month of categorisation costs Illustrative, computed from the published human-hours formula rather than measured on a customer. | Observed effort | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 5,000 characters of written output at 90 chars/min | 56 | | 4 acceptance criteria ticked at 8 min each | 32 | | 1 progress comment, 2 files produced | 30 | | Total: 2.2 human-hours at $2 | $4.40 | ## What an AI bookkeeper is not good at It is not an accountant and Polaris does not provide accounting or tax advice. Treat everything it produces as a prepared draft for a qualified person to check, exactly as you would treat work from a junior who started last month. It does not write to your ledger. There is no path by which a Polaris worker posts an entry into your accounting system, and that is deliberate: the output is a file and a list, and a person moves it. It will also be wrong about anything jurisdiction-specific unless you tell it the rules. VAT treatment, allowable expenses and thresholds vary by country and change by year, so the category rules in the skill file need to come from your accountant rather than from the worker's general knowledge. Where it does not know, it should push the row to the exceptions list, and the skill file should say so explicitly. ## Questions people ask **Can an AI bookkeeper file my taxes?** No. It prepares and organises work for a qualified human to review, approve and file, and Polaris does not provide accounting or tax advice. Filings, returns and anything sent to a tax authority are a person's responsibility. **Does it write entries into our accounting system?** No. It produces files and lists that a person reviews and moves. Nothing a Polaris worker does reaches your ledger of record without a human putting it there. **How does it know our categorisation rules?** You write them into the worker's SKILL.md, ideally with your accountant. Rules written once are applied every month, and anything the rules do not cover goes to the exceptions list rather than being guessed at. **What access does it need to Stripe?** A restricted, read-only key is enough for categorisation and chasing, and it is the safer default. Credentials are verified live and stored server-side, so a browser cannot read them back after they are saved. **Is it cheaper than a bookkeeping service?** It is priced differently: nothing per month, and $2 per human-hour of work actually delivered, itemised on the work log. Whether that is cheaper depends entirely on your volume, and it does not remove the need for a qualified reviewer. ## Related - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/use-cases/finance/expense-categorization - https://www.polarishq.co/use-cases/finance/invoice-tracking - https://www.polarishq.co/use-cases/finance/monthly-close-checklist - https://www.polarishq.co/use-cases/finance/vendor-management - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/glossary/work-log --- --- title: "AI Recruiter That Screens Against Your Rubric" description: "An AI recruiter that screens applications against a written rubric, drafts outreach and prepares interview kits. Hiring decisions always stay with people." url: https://www.polarishq.co/ai-workers/recruiter section: AI workers updated: 2026-08-21 --- # Hire an AI recruiter Every candidate is scored against the same written rubric, and the rubric is a file your hiring manager wrote and can change. ## The short answer An AI recruiter in Polaris is a worker you hire to prepare hiring work: screening applications against a rubric you write, drafting outreach and rejection messages, building interview kits, and keeping a scorecard document current. People make every hiring decision, send every message and run every interview. Polaris does not provide employment or legal advice. - **Typical output:** Scorecard doc with evidence per criterion - **Core connections:** Gmail, Google Calendar, Google Drive - **Illustrative task:** 2.9 human-hours, $5.80 ## What this worker is great at Screening is repetitive, high volume, and exactly where inconsistency creeps in. - **Applying one rubric to everyone** — The criteria live in the skill file. Every candidate is assessed against the same list in the same order, with the evidence quoted next to each score. - **Showing its working** — A score with no evidence attached is useless. The scorecard cites the line in the application that produced each judgement, so a hiring manager can disagree with something specific. - **Interview kits** — Questions tied to the rubric, what a strong answer contains, and the two follow-ups to ask when the first answer is thin. - **Drafting the hard messages** — Rejections that are short, specific and human, drafted in your tone for a person to read and send. - **Keeping the pipeline current** — Candidates as tasks with owners and dates, so nobody sits unanswered for nine days because the spreadsheet was not opened. ## A realistic first week 1. **Write the rubric first** — Before any screening, have the hiring manager write the criteria and what evidence counts. Paste it into the skill file. This is the artefact the whole role depends on. 2. **Screen the current batch** — Assign the open applications with acceptance criteria: every candidate scored, every score evidenced, no candidate rejected by the worker. 3. **Build the interview kit** — One page per stage, tied back to the rubric, so two interviewers assess the same things. 4. **Draft the outreach** — Ten messages to people you want to approach, each referencing something real from their public work. ## Where the human stays in charge Hiring affects people's livelihoods, so the boundary here is drawn tightly. **The recruiter prepares** - Scored applications with quoted evidence - A ranked shortlist proposal - Interview questions and answer guides - Drafted outreach and rejection messages **People decide** - Who advances and who does not - Every message that reaches a candidate - The interview itself and the reference calls - The offer, the terms and the signature > **What an AI recruiter is not good at** > > It cannot judge whether someone will be good to work with, which is most of what an interview is for. It has no access to your applicant tracking system unless you paste the material into the task, and it does not auto-reject: a candidate leaves the process because a person decided so. Employment law varies by country and changes, and Polaris does not provide employment or legal advice, so a rubric that touches protected characteristics needs review by someone qualified in your jurisdiction before a single application is screened against it. ## What a screening round costs Twenty-five applications screened against a five-criterion rubric, delivered as one scorecard document, works out at roughly 2.9 human-equivalent hours under the published formula: fifteen minutes to pick the task up, about a hundred minutes of finished written assessment at ninety characters a minute, five acceptance criteria ticked, one progress comment and one document produced. At $2 per human-hour that is $5.80. The example is illustrative, computed from the formula rather than measured on a customer. What is not illustrative is where it appears: every job's hours land on the worker's work log with the effort they were derived from, and any line can be challenged from there. ## Work this role picks up - [use-cases/hr/candidate-screening](https://www.polarishq.co/use-cases/hr/candidate-screening) - [use-cases/hr/interview-scheduling](https://www.polarishq.co/use-cases/hr/interview-scheduling) - [use-cases/hr/employee-onboarding](https://www.polarishq.co/use-cases/hr/employee-onboarding) - [use-cases/hr/policy-documentation](https://www.polarishq.co/use-cases/hr/policy-documentation) ## Questions people ask **Can an AI recruiter reject candidates automatically?** No. It scores against the rubric and proposes a shortlist, and a person decides who advances. It also cannot send anything: rejection and outreach messages are drafts until a human sends them. **How do we keep screening fair?** The rubric is a written file rather than a hidden model preference, so it can be reviewed, argued with and changed before it is used. Have someone qualified in your jurisdiction check it, keep the evidence requirement on every score, and treat the output as a prepared assessment rather than a decision. **Does it integrate with our applicant tracking system?** Not directly. The connection catalog covers Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and web search, so material from an ATS is handed to the worker in the task or through Drive. **Can it schedule interviews?** It prepares the scheduling work, such as the candidate list, the panel and the proposed slots, and a person confirms and sends the invitations. Google Calendar is in the connection catalog for exactly this kind of preparation. ## Related - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/use-cases/hr/candidate-screening - https://www.polarishq.co/use-cases/hr/interview-scheduling - https://www.polarishq.co/use-cases/hr/employee-onboarding - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "AI SDR for Account Research and First Drafts" description: "An AI SDR that researches accounts, finds the real trigger event with a source, and drafts the first touch. You press send. $2 per human-hour delivered." url: https://www.polarishq.co/ai-workers/sdr section: AI workers updated: 2026-08-21 --- # Hire an AI SDR It does the twenty minutes of account research nobody has time for, and hands you a first line that could only have been written about that company. ## The short answer An AI SDR in Polaris is a worker you hire to prepare outbound: researching target accounts with live web search, documenting the trigger event that makes the timing right, drafting a first-touch message per account, and keeping the target list current as tasks. A person reviews every message and sends it. The worker does not send email or make calls. - **Typical output:** Account brief plus drafted first touch - **Core connections:** Web search, HubSpot, Gmail - **Illustrative task:** 3.5 human-hours, $7.00 ## What this worker is great at The research that makes personalisation real rather than a merge field. For each account it looks for something checkable and recent: a funding round, a job posting that reveals a new team, a pricing page that changed, a public repository that shows which tools they run. Each finding comes back with the URL behind it, so you can read the source before you quote it in an email. The output is two things per account. A short brief with the evidence, and one drafted opening that uses the strongest piece of it. Twelve accounts researched to that standard is a morning a person rarely gets, and it is the difference between a first line that could be sent to anyone and one that could only be sent to them. It is also useful for the housekeeping that decays fast. Which accounts have gone quiet, which contacts changed jobs, which deals have no next step written down. ## What to connect it to - **Web search** — The engine of the role. No search access, no evidence, and without evidence this worker writes the same email everyone else sends. - **HubSpot** — So account research is prepared against the deals and contacts that already exist rather than a separate list nobody reconciles. - **Gmail** — For drafting in context. Sending stays with a person, and that is a hard boundary of the product rather than a setting. - **Slack** — When a teammate drops an account name in a channel, the Inbox turns it into a prefilled research task waiting for one click. ## A realistic first week 1. **Monday: define the ideal account in writing** — Company shape, the signal that means now rather than later, and the two disqualifiers. Put it in the skill file before any research runs. 2. **Tuesday: twelve account briefs** — One acceptance criterion per account: evidence found, URL cited, or explicitly marked as no trigger found. 3. **Wednesday: the first touches** — One drafted message per qualified account, each opening on the specific evidence. Read them, cut half the second paragraph, send. 4. **Friday: pipeline hygiene** — Every open deal without a written next step becomes a dated task with an owner. ## What a research batch costs Illustrative example, computed from the published human-hours formula: twelve account briefs in one session. - **12** — Accounts researched. Each with a cited trigger or an explicit none - **3.5** — Human-equivalent hours. 6 searches, 7,200 characters, 3 criteria ticked - **$7.00** — Total cost. At $2 per human-hour, itemised on the work log ## What an AI SDR is not good at It does not send. No email leaves your account and no sequence starts because a worker decided to, which rules out the volume plays and is a deliberate limit rather than a missing feature. It cannot reach gated data. Contact databases, sales intelligence platforms and anything behind a login are outside the connection catalog, so email addresses and phone numbers have to come from a tool you already pay for. It reads what the open web publishes, which is enough for the trigger and not enough for the contact record. And it has no judgement about whether a prospect is worth your time beyond the criteria you wrote down. Give it a vague ideal customer profile and you get twelve briefs about companies you should never have looked at, delivered promptly. ## Questions people ask **Can an AI SDR send cold emails?** No. It drafts messages and posts them on the task, and a person reviews and sends. Polaris workers deliver work as comments and never take an outbound action on your behalf. **Where does it find trigger events?** From live web search: funding announcements, job postings, changed pricing pages, public repositories, press coverage. Every finding is delivered with the URL it came from, and accounts with no genuine trigger are marked as such rather than padded with generic observations. **Can it find email addresses?** Not reliably, and it should not be asked to. The connection catalog has no contact database in it, so verified addresses come from whatever prospecting tool you already use while the worker handles research and drafting. **How is this priced against an outsourced SDR?** There is no monthly retainer and no seat. You pay $2 per human-hour of delivered research and drafting, itemised job by job, and nothing at all in a week where no task is assigned. Meetings booked remain the work of the person sending the emails. ## Related - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/use-cases/sales/lead-research - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/use-cases/sales/crm-hygiene - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/ai-worker --- --- title: "AI QA Engineer for Bug Triage and Test Plans" description: "An AI QA engineer that turns messy bug reports into reproducible tickets, writes test plans and finds duplicates. It reasons about software; it does not run it." url: https://www.polarishq.co/ai-workers/qa-engineer section: AI workers updated: 2026-08-21 --- # Hire an AI QA engineer Fourteen vague reports go in, fourteen tickets with steps, expected behaviour and a severity come out, with the three duplicates already merged. ## The short answer An AI QA engineer in Polaris is a worker you hire to turn raw bug reports into engineering-ready tickets and to write test plans. It rewrites reports with reproduction steps, expected and actual behaviour and a severity, groups duplicates, drafts regression checklists, and posts the batch as a comment. It reasons about software from what it is given and does not execute code. - **Typical output:** Rewritten tickets with severity and steps - **Core connections:** GitHub, Linear, Slack - **Illustrative task:** 2.4 human-hours, $4.80 ## What this worker is great at The gap between what a user reports and what an engineer can act on. - **Making a report reproducible** — It rewrites the customer's story into numbered steps, expected behaviour, actual behaviour, environment and severity, and it marks the fields it could not fill rather than inventing them. - **Killing duplicates** — Fourteen reports are often nine bugs. It groups them and names the one ticket the engineer should read. - **Writing the test plan** — For a feature about to ship: the happy path, the edge cases that actually break things, and the regression list of what this change is most likely to have damaged. - **Severity with a reason** — A severity nobody can argue with is a severity nobody trusts. Each one comes with the impact sentence behind it. - **Release checklists** — Acceptance criteria as checklist items on the task, ticked as each one is genuinely verified by whoever verifies it. ## Before and after, on one report The value of this role is visible in a single row. | Field | What arrived | What the worker delivers | | --- | --- | --- | | Title | Export is broken | CSV export returns an empty file for orgs with more than 1,000 rows | | Steps | None | Numbered, from a clean session, with the data condition stated | | Expected vs actual | Missing | Both written explicitly, in the user's terms | | Severity | Unset | Set, with the impact sentence that justifies it | | Duplicates | Unknown | Three earlier reports linked, one canonical ticket named | ## A realistic first week 1. **Day one: the untriaged backlog** — Assign everything nobody has looked at. Acceptance criteria: every report rewritten, every duplicate grouped, every severity justified. 2. **Day two: the definition of severity** — Write your severity ladder into the skill file so the next batch uses your scale rather than a generic one. 3. **Day three: a test plan for the next release** — Feature description in, happy path plus edge cases plus regression list out. 4. **Friday: the pattern report** — One page on what this week's bugs have in common, which is the artefact that changes engineering priorities. ## Where the human stays in charge **The QA engineer prepares** - Reproducible, severity-rated tickets - Duplicate groupings with a canonical ticket - Test plans and regression checklists - A weekly pattern summary **Engineers decide** - Whether a bug is real after reproducing it - What gets fixed and in what order - Whether the release is safe to ship - Closing the task and rating the batch ## What an AI QA engineer is not good at It does not run your software. There is no browser, no test runner and no shell in the machine session, so it cannot reproduce a bug, take a screenshot or watch a suite go red. It writes the plan and the ticket; a person or your CI runs the test. That limit is worth stating plainly because the role name invites the opposite assumption. Everything it produces is derived from the reports, the descriptions and the code context you put in front of it, plus live web search for library behaviour and error messages. It also cannot tell you whether an intermittent failure is a race condition or a flaky test environment. It will list both hypotheses and the evidence that would separate them, which is genuinely useful, and then an engineer has to go and look. ## Questions people ask **Can an AI QA engineer run automated tests?** No. The machine session has web search, documents and file production, not a test runner or a browser. It writes test plans, cases and checklists that your engineers or your CI execute. **How does it decide severity?** By the ladder you write into its SKILL.md. Without one it uses a generic scale, which is why the first week should include ten minutes writing down what critical actually means at your company. Every severity is delivered with the impact sentence behind it. **Does it work with our issue tracker?** Linear and GitHub are both in the connection catalog, and the tickets it prepares are written to be pasted or moved into whichever tracker you use. Polaris tasks themselves can hold the work if you would rather not keep two systems. **What does a triage batch cost?** As an illustrative example from the published formula, fourteen reports rewritten in one session comes to about 2.4 human-equivalent hours, or $4.80 at $2 per human-hour. The effort behind that number is itemised on the work log and can be challenged there. ## Related - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/use-cases/engineering/bug-triage - https://www.polarishq.co/use-cases/engineering/code-review-workflow - https://www.polarishq.co/use-cases/engineering/incident-postmortems - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work --- --- title: "AI Data Analyst for Definitions and Reporting" description: "An AI data analyst that defines metrics precisely, specifies the query, interprets results you provide and writes the reporting nobody has time for." url: https://www.polarishq.co/ai-workers/data-analyst section: AI workers updated: 2026-08-21 --- # Hire an AI data analyst It settles what active user means, writes the query that matches the definition, and turns the numbers you hand it into a paragraph an executive can read. ## The short answer An AI data analyst in Polaris is a worker you hire for the written half of analytics: defining metrics without ambiguity, specifying the query that matches the definition, interpreting result sets you provide, and producing recurring reports as documents and CSV files. It writes and reasons about analysis rather than connecting to a warehouse and running it. - **Typical output:** Metric definitions doc plus CSV - **Core connections:** Supabase, Stripe, Google Drive - **Illustrative task:** 2.5 human-hours, $5.00 ## What this worker is great at Ending the argument about what a number means. Most reporting disputes are definitional: three dashboards say three different things about active users because nobody ever wrote down whether a login counts, whether internal accounts are excluded, and what the window is. This worker produces the definition document, one metric per section, with the edge cases named and the query logic spelled out beside it. It is also good at the writing that surrounds a number. You paste in the week's figures and get back the commentary: what moved, what did not, which movement is inside normal variance and which one deserves a task. That paragraph is usually the reason a dashboard existed in the first place. For recurring work it produces the same structure every time, which is what makes week nine comparable to week two. ## What to connect it to - **Supabase** — The product database when Polaris-shaped teams keep their data there. A restricted, read-only key is the sensible default. - **Stripe** — Revenue, refunds and invoice data, which is where half of the metric definitions in a software business end up pointing. - **Google Drive** — Where the recurring reports have to land so the rest of the company can find them. - **Web search** — For benchmark definitions and how a metric is conventionally calculated, cited rather than asserted. ## What a definitions task costs Illustrative, computed from the published human-hours formula. Six metrics defined and documented in one session. | Observed effort | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 2 live web searches at 12 min each | 24 | | 5,400 characters of finished prose at 90 chars/min | 60 | | 4 acceptance criteria ticked at 8 min each | 32 | | 1 progress comment, 1 doc produced | 20 | | Total: 2.5 human-hours at $2 | $5.00 | ## Where the human stays in charge **The analyst prepares** - Metric definitions with edge cases named - Query logic that matches each definition - Written commentary on a result set - Recurring report structure, held steady week to week **Your data team decides** - Whether the definition is the one the company adopts - Running anything against production - Whether a number is trustworthy enough to act on - What the business does about the movement > **What an AI data analyst is not good at** > > It does not connect to your warehouse and execute SQL. A machine session has web search, documents and files, so the analyst specifies the query and a person or a pipeline runs it. It cannot see a number you did not give it, which means every interpretation is only as good as the result set pasted into the task. Statistical significance on small samples is another place to be careful: it will tell you when a movement is inside normal variance, and it cannot rescue an experiment that never had the traffic to conclude anything. ## Questions people ask **Can the AI data analyst query our database?** It writes the query and explains what it returns; running it stays outside the machine session. Supabase and Stripe are in the connection catalog for the credential side, and the analytical output is documents, commentary and files. **How do I get useful commentary on weekly numbers?** Paste the result set into the task and state what a normal week looks like. The commentary is only as good as the context: without a baseline, any movement can be described and none of it can be judged. **Can it build a dashboard?** It writes the specification for one: the metrics, their definitions, the cuts that matter and the queries behind each tile. Building it in your BI tool is a person's job, and the specification is the part that usually takes longest anyway. **How does it handle a metric nobody has defined?** It proposes a definition, names the edge cases that could change the number, and marks the choices you have to make rather than quietly picking one. Those choices are what a data team should be arguing about. ## Related - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/use-cases/data/metric-definitions - https://www.polarishq.co/use-cases/data/reporting-automation - https://www.polarishq.co/use-cases/data/dashboard-requests - https://www.polarishq.co/use-cases/data/analysis-backlog - https://www.polarishq.co/integrations/supabase - https://www.polarishq.co/integrations/stripe --- --- title: "AI Social Media Manager for the Whole Calendar" description: "An AI social media manager that plans the month, writes captions per platform and drafts hook variants. Posting stays with a person who reads them first." url: https://www.polarishq.co/ai-workers/social-media-manager section: AI workers updated: 2026-08-21 --- # Hire an AI social media manager A month of posts planned, written and filed by Tuesday, so the only decision left is which ones you actually publish. ## The short answer An AI social media manager in Polaris is a worker you hire to plan and write social content. It builds a dated calendar as tasks, writes captions adapted to each platform's format and length, drafts several hook variants per post, and files the batch as a document with a CSV. A person reviews the copy and publishes it, because workers do not post. - **Typical output:** Dated calendar plus written captions - **Core connections:** Instagram, WhatsApp, web search - **Illustrative task:** 3.2 human-hours, $6.40 ## What this worker is great at Volume with structure, which is the part that collapses first when a founder is doing social themselves. - **The calendar as real tasks** — Each post is a dated task in a workstream with an owner, so the plan lives on the board rather than in a spreadsheet nobody opens after week two. - **Writing per platform, not once** — The same idea written three ways, because a caption that works on Instagram is not the post that works on LinkedIn and pretending otherwise is why cross-posting reads badly. - **Hooks in batches** — Five opening lines per post, so the choice is yours instead of the first thing that came out. - **Working from the pillars** — Your content pillars live in the skill file, so the month is balanced rather than five posts about the same feature. - **Recycling what worked** — Give it the posts that performed and the reasons, and it will write the next round in that shape. ## A realistic first week 1. **Write the pillars down** — Four or five recurring themes, the audience for each, and the things you will never post about. Straight into the skill file. 2. **Plan the month** — Twelve posts, dated, each as a task with the pillar in the labels. Acceptance criteria: no two consecutive posts from the same pillar. 3. **Write the captions** — Full copy for the first two weeks, with hook variants, delivered as a document and a CSV you can work from. 4. **Review and cut** — Read them in one sitting. Delete the three that sound like everyone else, comment on the task saying why, and the next run avoids that register. ## Where the human stays in charge **The manager prepares** - A dated calendar as tasks - Captions per platform, with hook variants - A shot list or asset note per post - A monthly plan balanced across your pillars **You do** - Publish, from your own accounts - Approve anything that makes a claim about the product - Reply to comments and messages as a person - Decide what the brand sounds like this quarter ## What it costs, and what it cannot do A month of planning plus twenty-four written captions comes to roughly 3.2 human-equivalent hours under the published formula, which is $6.40 at $2 per human-hour. Illustrative, computed from the formula rather than measured on a customer, and itemised on the work log where you can challenge it. The limits are worth being blunt about. It does not post: publishing is a human action from your own accounts. It cannot see your analytics, so it will not tell you what performed unless you paste the numbers into the task, and it will never invent an engagement figure to fill a report. It has no eye for imagery, so it writes the shot note and a person shoots or selects the asset. It is also a poor judge of taste in your specific community. A joke that lands in a niche audience usually depends on knowing the people in it, and that is not something a skill file captures well. ## Work this role picks up - [use-cases/marketing/social-media-scheduling](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) - [use-cases/marketing/content-calendar](https://www.polarishq.co/use-cases/marketing/content-calendar) - [use-cases/marketing/campaign-planning](https://www.polarishq.co/use-cases/marketing/campaign-planning) ## Questions people ask **Does the AI social media manager publish posts?** No. It writes and plans, and delivers the batch as a comment with a document and a file. Publishing happens from your own accounts, by a person, after reading the copy. **Can it report on what performed?** Only from numbers you provide in the task. It has no analytics view of its own and it will not estimate a figure, so paste in the real performance data and ask for the pattern behind it. **How does it avoid sounding like generic AI social copy?** By having rules to obey. Banned phrases, a required specificity level and examples of your best posts go into the SKILL.md, and the drafts you delete with a reason become the next set of rules. Without that file it writes the average of the internet, like everything else. **Can one worker handle several brands?** Hire one per brand. There are no per-worker fees, so separate skill files keep the voices from bleeding into each other, which is the failure mode when one worker carries two brand guides. ## Related - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/use-cases/marketing/social-media-scheduling - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/use-cases/marketing/campaign-planning - https://www.polarishq.co/integrations/instagram - https://www.polarishq.co/integrations/whatsapp - https://www.polarishq.co/glossary/workstream --- --- title: "AI Executive Assistant That Owns the Follow-Ups" description: "An AI executive assistant that turns meeting notes into owned, dated tasks, prepares briefing packs before calls and keeps the week's non-negotiables visible." url: https://www.polarishq.co/ai-workers/executive-assistant section: AI workers updated: 2026-08-21 --- # Hire an AI executive assistant Every commitment made in a meeting becomes a task with a name and a date on it before the next meeting starts. ## The short answer An AI executive assistant in Polaris is a worker you hire to convert meeting output into tracked work and to prepare you for what is next. It turns notes into owned, dated tasks, drafts the follow-up messages, assembles a briefing document before a call, and produces a weekly digest of what is drifting. A person sends, schedules and commits. - **Typical output:** Owned dated tasks plus a briefing doc - **Core connections:** Google Calendar, Gmail, Slack - **Illustrative task:** 2.1 human-hours, $4.10 ## What this worker is great at Turning talk into tracked work. Paste the notes from a meeting and it returns the commitments as individual tasks, each with the person who said they would do it and the date they said it by, filed into the right workstream. The commitments that had no owner or no date come back as a short list of questions, which is usually the most useful part of the delivery. Preparation is the other half. Before an external call it assembles a one-page brief: who is on the call, what was agreed last time, what is open, and the two things you said you would come back with. That page is the difference between a call that moves and a call that recaps. Weekly, it produces the digest: what is overdue, what is unowned, what has been pushed twice. Focus keeps the non-negotiables in view across today, this week and the next thirty days, and the assistant is what keeps the lane honest. ## What to connect it to - **Google Calendar** — So preparation is tied to what is actually in the diary rather than to a list somebody maintains separately. - **Gmail** — For drafting follow-ups in context. Sending stays with you. - **Slack** — The Inbox turns messages into prefilled task suggestions, which is where most unrecorded commitments are born. - **Google Drive** — For the briefing packs and the documents a meeting refers to. ## A realistic first week 1. **Monday: last week's meeting notes** — All of them, in one task. Acceptance criteria: every commitment has an owner and a date, or appears on the open-questions list. 2. **Tuesday: a briefing pack** — For your hardest call this week. Ask for one page, and for the two questions you should be ready to answer. 3. **Thursday: the follow-up drafts** — Short messages to the people who owe you something, drafted for you to send from your own account. 4. **Friday: the drift digest** — Overdue, unowned and twice-pushed, in one document, ordered by how much it will cost to keep ignoring. ## Where the human stays in charge **The assistant prepares** - Commitments as owned, dated tasks - Briefing packs before meetings - Drafted follow-ups and chasers - A weekly digest of what is drifting **You do** - Accept, decline and move meetings - Send every message that goes out - Decide what is genuinely a priority this week - Close the task and rate the work ## What an AI executive assistant is not good at It does not attend or transcribe meetings. There is no recorder, no bot in the call and no live audio, so the notes have to arrive in the task. Give it a transcript and it is excellent; give it nothing and it has nothing. It also cannot make commitments on your behalf. It does not accept invitations, book flights, negotiate a time with somebody's assistant or reply to anyone. Every one of those is a person's action, taken after reading what the worker prepared. And it will not protect your calendar from you. It can show that three deep-work blocks were eaten by meetings this week, in writing, on a Friday. The decision to stop agreeing to them is not a task anyone can delegate. ## Questions people ask **Can an AI executive assistant join my meetings?** No. It works from notes or a transcript you put in the task. There is no meeting bot and no audio capture in Polaris, which keeps it out of the room and out of your recording policy. **Does it schedule meetings for me?** It prepares the scheduling work, such as proposed slots and the message to send, and a person confirms. Google Calendar is in the connection catalog so the preparation is grounded in the real diary. **How does it decide what belongs in Focus?** By the rules you give it in the skill file plus the dates on the work. Focus is a time-based lane across today, this week and the next thirty days, pinned first in every view, and the assistant proposes what belongs there while you decide what stays. **What does a follow-up task cost?** Turning a meeting into eleven owned tasks and a digest is roughly 2.1 human-equivalent hours under the published formula, or $4.10 at $2 per human-hour. That example is illustrative; the real number for your job appears on the work log with the effort behind it. ## Related - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/ai-workers/recruiter - https://www.polarishq.co/use-cases/executive/meeting-follow-ups - https://www.polarishq.co/use-cases/executive/weekly-business-review - https://www.polarishq.co/use-cases/executive/okr-tracking - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/glossary/focus-lane --- --- title: "AI Project Coordinator That Chases the Board" description: "An AI project coordinator that finds unowned and undated work, prepares the sprint, writes the status roll-up and maps dependencies before they bite." url: https://www.polarishq.co/ai-workers/project-coordinator section: AI workers updated: 2026-08-21 --- # Hire an AI project coordinator The unglamorous half of running projects, done every week without anyone having to be the person who nags. ## The short answer An AI project coordinator in Polaris is a worker you hire to keep a board truthful. It finds tasks with no owner or no date, drafts the sprint plan from the backlog, writes the weekly status roll-up, and maps dependencies between workstreams. It proposes; the people who own the work confirm priorities, dates and scope. - **Typical output:** Status roll-up plus a list of gaps - **Core connections:** Linear, Slack, Notion - **Illustrative task:** 2.1 human-hours, $4.20 ## What this worker is great at Noticing what everyone else has stopped seeing. A board that has been running for three months contains work with no owner, dates that passed in April and two lanes that are quietly the same project. The coordinator lists all of it in one document, ordered by the damage it will do, and turns each item into a task pointed at a named person. It writes the status roll-up nobody wants to write on a Thursday afternoon: what shipped, what slipped, what is blocked and who is blocking it. Because it works from the board rather than from what people say in the meeting, the version it produces is the one that includes the awkward line. Sprint preparation is the third piece. Backlog in, a proposed sprint out, with each item carrying acceptance criteria as a checklist so the definition of done is written before the work starts rather than argued about at the end. ## What to connect it to - **Linear** — When engineering work lives there and the coordination has to reconcile two boards rather than pretend one exists. - **Slack** — Decisions get made in channels and never reach the board. The Inbox turns those messages into prefilled task suggestions you approve. - **Notion** — For the project documentation the roll-up has to link back to. - **GitHub** — Useful when what shipped is best answered by what merged. ## Where the human stays in charge **The coordinator prepares** - A list of unowned, undated and stale work - A proposed sprint with acceptance criteria - The weekly status roll-up - A dependency map across workstreams **Owners decide** - What the priority actually is - Who takes each piece of work - Whether a date can be committed to - What gets cut when the sprint is too full ## What a weekly pass costs Illustrative example computed from the published human-hours formula: one board audit and roll-up in a single session. - **2.1** — Human-equivalent hours. 4,500 characters, 5 criteria ticked, 1 doc - **$4.20** — Weekly cost. At $2 per human-hour delivered - **$0** — Software cost. Unlimited humans, tasks and workstreams ## What an AI project coordinator is not good at It cannot set priority. It can show that two lanes both claim to be the most important thing this quarter, and that is where its authority ends. Priority is a judgement about the business, and delegating it to a worker produces a plan nobody follows. It also has no read on morale. A person who is quietly stuck looks identical to a person who is quietly fine on a board, and the coordinator will report the task as in progress in both cases. That gap is what a standup is for. And its estimates are inherited, not earned. It has never built your product, so a date it proposes is a restatement of what somebody already said rather than an independent judgement of how long the work takes. ## Work this role picks up - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) - [use-cases/engineering/sprint-planning](https://www.polarishq.co/use-cases/engineering/sprint-planning) - [use-cases/product/roadmap-planning](https://www.polarishq.co/use-cases/product/roadmap-planning) ## Questions people ask **How is an AI project coordinator different from an automation rule?** A rule fires on a trigger and does one thing. This worker reads the whole board, writes a judgement about what is wrong with it, and delivers that as a document with tasks attached. It also explains why each item is on the list, which a rule cannot do. **Can it assign work to people?** It proposes owners and dates and creates the tasks, and the people involved confirm. Assignment behaves the same for humans and AI workers in Polaris because both are rows in the same members table. **Does it replace a project manager?** No. It removes the administrative half of the role, which is chasing, tidying and reporting. Prioritisation, negotiation and knowing when a team is struggling stay with a person. **How often should it run?** Weekly is the natural cadence, assigned as a recurring task you create. A worker runs when work is assigned to it, so the rhythm comes from the board rather than from a schedule inside the worker. ## Related - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/use-cases/engineering/sprint-planning - https://www.polarishq.co/use-cases/product/roadmap-planning - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/glossary/focus-lane --- --- title: "AI SEO Specialist for Briefs and Link Plans" description: "An AI SEO specialist that reads live results pages, writes briefs with the evidence behind them, plans internal links and shapes answers AI engines can quote." url: https://www.polarishq.co/ai-workers/seo-specialist section: AI workers updated: 2026-08-21 --- # Hire an AI SEO specialist It looks at what is actually ranking today, then writes the brief that says what your page has to do differently to deserve the spot. ## The short answer An AI SEO specialist in Polaris is a worker you hire to do the research and planning half of search work. It reads live results pages, writes content briefs with the evidence behind each recommendation, plans internal linking across a page set, and structures answer blocks so generative engines can quote them. It cannot access your analytics or crawl your site. - **Typical output:** Content brief with SERP evidence - **Core connection:** Web search - **Illustrative task:** 3.3 human-hours, $6.60 ## What this worker is great at Search work splits into research, planning and execution. This worker owns the first two. - **Reading the results page honestly** — What ranks now, what format each result uses, and what those pages have in common. The brief starts from evidence rather than from a keyword tool's difficulty score. - **Briefs somebody can write from** — The angle, the sections, the questions the page must answer, and the specific thing this page will contain that the current top result does not. - **Intent, stated plainly** — Whether a query wants a definition, a comparison or a purchase, and what that means for the page shape. - **Internal link plans** — Which page should link to which, and with what anchor, across a set. This is the work that decides whether a large page set behaves like a structure or a pile. - **Answer blocks for AI engines** — Self-contained passages that answer the query without pointing at earlier context, which is the shape a model can lift into a citation. ## A realistic first week 1. **Monday: five briefs** — Five queries you care about, one brief each, every recommendation carrying the URL of the result that justifies it. 2. **Tuesday: the intent audit** — Take ten existing pages and ask which query each one actually serves. The mismatches are usually the cheapest wins you have. 3. **Wednesday: the link plan** — A map of which pages should link to which, with anchors, delivered as a table. 4. **Friday: the answer-block pass** — Rewrite the opening block of your top pages so each answers the question on its own in a quotable paragraph. ## What this role does and does not touch | Search work | Who does it | | --- | --- | | Reading live results and writing the brief | The AI SEO specialist | | Deciding which queries the business chases | You | | Writing the page | A person, or the AI content writer | | Search Console and analytics data | You, and you paste the numbers in | | Crawling your site for technical issues | A crawler you already run | | Publishing and shipping the change | A person | ## Where the human stays in charge **The specialist prepares** - Briefs with SERP evidence attached - Intent classification per query - An internal link plan across the set - Quotable answer blocks for each page **You decide** - Which pages are worth existing at all - What ships and when - Whether a claim in the copy is true - How aggressive the indexation rollout is ## What an AI SEO specialist is not good at It cannot see your data. No Search Console, no analytics, no rank tracker, no crawl of your own site, because none of those are in the connection catalog. Everything it says about your performance comes from numbers you paste into the task, and it will not estimate traffic to fill a gap. It also cannot promise a ranking, and a brief that claims otherwise should be treated as a red flag regardless of who wrote it. What it can do is make the page genuinely better than the ones currently winning, which is the only lever anybody actually controls. The last limit is scale. A machine session runs a handful of live searches, so a hundred-query research project belongs in several tasks rather than one, and the specialist should be told which ten queries matter most before it starts. ## Questions people ask **Can an AI SEO specialist see our Search Console data?** No. It works from live web search plus whatever you put in the task. Paste in the queries, impressions and positions you care about and it will write the analysis; without that data it stays on what the open results page shows. **Does it write the pages too?** It writes briefs, and the AI content writer or a person writes the page. Splitting the roles keeps the brief honest, because the worker judging what the page must contain is not the one trying to finish it. **What is an answer block and why does it matter?** It is a short, self-contained passage that answers the page's question with no pronouns pointing at earlier context. Generative engines quote passages rather than pages, so a block that stands alone is the unit that gets cited. **Can it help with programmatic page sets?** Yes, and this is where it earns most. It plans the permutation dimensions, writes the uniqueness rules that keep pages out of doorway territory, and designs the internal link graph. It will also tell you which permutations have nothing specific to say and should not be generated. ## Related - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/use-cases/marketing/seo-content-production - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/generative-engine-optimization --- --- title: "AI Technical Writer for Docs Nobody Updates" description: "An AI technical writer that turns merged changes into release notes, writes runbooks and SOPs, and audits documentation for the parts that went stale." url: https://www.polarishq.co/ai-workers/technical-writer section: AI workers updated: 2026-08-21 --- # Hire an AI technical writer The documentation debt on your team is not a writing problem, it is a nobody-has-two-free-hours problem, and this is the worker for those two hours. ## The short answer An AI technical writer in Polaris is a worker you hire to produce and maintain documentation. It turns a list of merged changes into user-facing release notes, writes runbooks and standard operating procedures from what a person describes, audits existing docs for statements that no longer match reality, and delivers everything as versioned pages a reviewer comments on. - **Typical output:** Versioned doc, ready for review - **Core connections:** GitHub, Notion, Google Drive - **Illustrative task:** 1.8 human-hours, $3.60 ## What this worker is great at Release notes are the obvious win. Thirty-one merged changes in, a page that says what changed, who it affects and what they need to do about it out, with the internal refactors filtered out because nobody outside the team cares about them. That page normally gets written badly at five o'clock on a Thursday by whoever lost. Runbooks are the higher-value one. A person describes what they do when the queue backs up, in whatever order it comes out, and the worker returns numbered steps with the decision points marked and the commands quoted exactly as given. What was in one engineer's head becomes a document the next person can follow at two in the morning. Doc audits are the third. Point it at a set of pages and ask which statements contradict the current product, which links are dead, and which pages have not been touched since a rename. The answer is a task list, not a lecture. ## What to connect it to - **GitHub** — The changes, the pull request titles and the code context that release notes and reference docs come from. - **Notion** — When the existing documentation lives there and the audit has to read what is already published. - **Google Drive** — For the procedures that live outside engineering, where the rest of the company keeps its files. - **Web search** — For checking how a library or protocol actually behaves before documenting it, with the source cited. ## A realistic first week 1. **Day one: the release notes backlog** — Two or three releases that never got written up. One acceptance criterion per release: every user-facing change covered, no internal noise included. 2. **Day two: the on-call runbook** — Dictate the procedure into the task description in whatever order it comes out. The worker returns the ordered version with the gaps marked. 3. **Wednesday: the audit** — Twenty pages, one list of what is now false, ordered by how likely somebody is to be misled by it. 4. **Friday: the style rules** — Write your documentation conventions into the skill file so the next fifty pages match without anyone reviewing for tone. ## Where the human stays in charge **The writer prepares** - Release notes filtered to what users care about - Runbooks with decision points marked - Standard operating procedures, numbered - An audit list of statements that went stale **An engineer or owner decides** - Whether the procedure is actually correct - What is safe to document publicly - Which changes belong in the notes at all - Publishing the page and closing the task > **What an AI technical writer is not good at** > > It cannot verify that a procedure works, because it does not run anything. If the person describing the runbook skips a step, the document will skip it too, in confident numbered prose. It also has no way of knowing which internal details are sensitive, so a public-facing page needs a person to read it with that question in mind. Documentation about behaviour nobody wrote down anywhere is beyond it: there is no source to work from, and it will say so rather than invent the missing half. ## Questions people ask **Can an AI technical writer read our codebase?** GitHub is in the connection catalog, and code context you provide in the task is what it writes from. It reasons about code rather than executing it, so behaviour that is only observable at runtime has to be described by someone who has watched it. **Where do the documents end up?** In Polaris Docs by default, as versioned pages with comments and sub-pages, filed under the parent page whose topic fits. From there a person moves the content wherever it needs to be published. **How do we keep documentation style consistent?** Put the conventions in the worker's SKILL.md: heading depth, whether to use second person, how commands are formatted, what a runbook must always contain. Every document after that follows the same rules without a reviewer re-explaining them. **What does a set of release notes cost?** Notes for thirty-one merged changes work out at roughly 1.8 human-equivalent hours under the published formula, which is $3.60 at $2 per human-hour. That figure is illustrative; your actual line appears on the work log with the effort it came from. ## Related - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/qa-engineer - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/use-cases/product/release-notes - https://www.polarishq.co/use-cases/operations/process-documentation - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/notion --- --- title: "AI Competitive Analyst Watching Your Market" description: "An AI competitive analyst that tracks what rivals publish, writes up what changed with sources, and tells you which moves matter and which are noise." url: https://www.polarishq.co/ai-workers/competitive-analyst section: AI workers updated: 2026-08-21 --- # Hire an AI competitive analyst Assign it every Monday and you get a written record of what your market did last week, with a URL behind every line of it. ## The short answer An AI competitive analyst in Polaris is a worker you hire to track named competitors and write up what changed. It reads public pricing pages, product announcements, job postings and press coverage with live web search, records what moved since the last brief, separates a real strategic shift from routine noise, and delivers the write-up as a document with sources. - **Typical output:** Watch brief with source trail - **Core connection:** Web search - **Illustrative task:** 4.0 human-hours, $7.90 ## What this worker is great at Keeping a written record where none exists. Most teams know their competitors changed something and cannot say when, which means every strategy conversation restarts from memory. A weekly brief in a versioned doc gives you a timeline: the price moved in March, the positioning changed in May, the job postings for a data team started in June. Pricing pages are its best surface. It records the tiers, the limits, what moved into the paid plan and what quietly left it, and it quotes the page rather than paraphrasing it. Job postings are the second best, because a company hiring five people for a new team has told you what it is building whether it meant to or not. The judgement it adds is separation. A blog post is not a strategy change. A pricing page rewrite plus a new role plus a changed homepage headline usually is, and the brief says which of the three it thinks you are looking at. ## What to connect it to - **Web search** — The entire role. Everything it reports comes from public sources it can cite, which is also the reason this work stays defensible. - **Notion** — When the competitive wiki lives there and each brief has to extend a page the sales team already reads. - **Google Drive** — For the quarterly summary that goes into a board pack. - **Slack** — So the sighting somebody posts in a channel becomes a prefilled task rather than scrolling away. ## What a monthly watch costs Illustrative, computed from the published human-hours formula: four competitors reviewed in one session, delivered as a doc and a PDF. | Observed effort | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 8 live web searches at 12 min each | 96 | | 6,300 characters of finished prose at 90 chars/min | 70 | | 4 acceptance criteria ticked at 8 min each | 32 | | 1 progress comment, 1 doc, 1 PDF | 25 | | Total: 4.0 human-hours at $2 | $7.90 | ## Where the human stays in charge **The analyst prepares** - What changed, with the date and the source - A separation of signal from routine noise - A running timeline across months - A quarterly summary for a board pack **You decide** - Whether to respond at all - What your own pricing does next - Which competitor is genuinely the threat - What goes in front of customers or investors ## What an AI competitive analyst is not good at It only sees what is published. Private roadmaps, unannounced pricing experiments, deals lost on terms nobody wrote down and anything discussed in a customer call are invisible to it. Your sales team knows things this worker never will, and the brief is better when their notes are pasted into the task. It is also not continuous. A worker runs when a task is assigned, so competitive watching is a recurring task you create rather than a monitor that pings you the moment a page changes. Weekly or monthly is the honest cadence. And it should not be asked to guess motive. It can tell you a competitor moved their cheapest tier from twelve dollars to nineteen, and it cannot tell you whether that was confidence or desperation. Briefs that speculate about intent tend to be the ones that age worst. ## Questions people ask **Does the AI competitive analyst monitor competitors continuously?** No. It runs when you assign it a task, so the pattern that works is a recurring weekly or monthly task per competitor set. Between runs, nothing is watching, and the brief says what changed since the previous one rather than pretending to be live. **Where does the information come from?** Public sources found through live web search: pricing pages, changelogs, blogs, job postings, press coverage and public repositories. Every claim in the brief carries the URL behind it so you can read the original before acting on it. **Can it track competitors we cannot name yet?** Ask it to find them first. Given the problem you solve and the buyer you sell to, it will search for who else is selling into that space and return a list with evidence, which is a different task from watching a set you already know. **How is this different from a research analyst?** The research analyst answers a question once. The competitive analyst maintains a record over time, which means its value comes from the second and tenth brief rather than the first, and its skill file is written around consistency of format rather than depth on a single question. ## Related - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/use-cases/marketing/competitor-monitoring - https://www.polarishq.co/use-cases/product/competitive-tracking - https://www.polarishq.co/use-cases/executive/strategic-research - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/glossary/work-log --- --- title: "AI Customer Success Manager for Renewal Prep" description: "An AI customer success manager that prepares renewal briefs, drafts QBR material and turns support themes into account risks before the renewal call." url: https://www.polarishq.co/ai-workers/customer-success-manager section: AI workers updated: 2026-08-21 --- # Hire an AI customer success manager It reads everything you know about an account and writes the two paragraphs you wish you had before the renewal call. ## The short answer An AI customer success manager in Polaris is a worker you hire to prepare account work. It writes renewal briefs from the history you provide, drafts quarterly business review material, turns recurring support themes into named account risks, and keeps follow-ups as owned, dated tasks. The relationship, the call and the commercial conversation stay with a person. - **Typical output:** Renewal brief per account - **Core connections:** HubSpot, Stripe, Gmail - **Illustrative task:** 2.9 human-hours, $5.70 ## What this worker is great at Preparation, which is the part that gets skipped when one person carries forty accounts. - **The renewal brief** — What they bought, what they have complained about, what was promised and never delivered, and the two questions you should expect on the call. - **Naming the risk in writing** — Four support tickets about the same missing feature is a renewal risk, not a support statistic. It says so, on the account, with the tickets listed. - **QBR material** — The structure, the narrative and the sections you have to fill with real numbers, so nobody starts from an empty slide the night before. - **Follow-through** — Every promise made on a call becomes a dated task with an owner, which is where most churn quietly starts. - **Consistency across accounts** — The same brief shape for every customer, so a colleague covering your accounts next week can read them. ## Where the human stays in charge Everything commercial and everything relational belongs to a person. **The manager prepares** - Renewal briefs with risks named - QBR outlines and narrative - Follow-up tasks from a call - A themes summary across the customer base **You do** - Run the call and hold the relationship - Negotiate price, terms and renewal - Decide what to promise and when - Send everything that reaches the customer ## A realistic first week 1. **Monday: the accounts renewing this quarter** — One brief each. Acceptance criteria: risks named with evidence, open promises listed, no invented history. 2. **Tuesday: the health rules** — Write down what actually predicts churn for your product and put it in the skill file. Without it the worker uses generic signals that mean nothing for your business. 3. **Wednesday: a QBR pack** — For your largest account, structured and written except for the numbers only you can supply. 4. **Friday: the themes doc** — What customers asked for most this month, ranked, ready to hand to product. ## What it costs and what it cannot see Six renewal briefs in one session comes to roughly 2.9 human-equivalent hours under the published formula, or $5.70 at $2 per human-hour. Illustrative, computed from the formula, and every real job lands on the work log with the effort behind it. The limits matter more here than the price. It has no product usage data unless you paste it in, so a health score it produces is built from tickets and history rather than from behaviour. It has never spoken to the customer, so it cannot tell you that the champion sounded tired on the last call, which is often the only signal that mattered. It also does not send anything or negotiate anything. Renewal conversations, pricing concessions and difficult apologies are human work, and a brief that pretends otherwise costs you the account it was meant to save. ## Work this role picks up - [use-cases/customer-support/customer-feedback-loops](https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops) - [use-cases/customer-support/escalation-tracking](https://www.polarishq.co/use-cases/customer-support/escalation-tracking) - [use-cases/sales/pipeline-management](https://www.polarishq.co/use-cases/sales/pipeline-management) ## Questions people ask **Can an AI customer success manager calculate a health score?** It can apply a scoring rule you write into its skill file, using the signals you give it in the task. It has no independent view of product usage, so a score is only as trustworthy as the data pasted in and the rule behind it. **Does it email customers?** No. It drafts, and a person sends from their own account. Polaris workers deliver work as comments on tasks, which keeps every customer-facing message in front of a human first. **How does it know what a customer complained about?** From what you connect and what you provide. HubSpot and Gmail are in the connection catalog, and support history handed to it in the task is what most briefs are built from. It marks what it could not find rather than filling the gap. **Is this the same as the support specialist role?** No. The support specialist works ticket by ticket in the queue. This role works account by account across time, and its output is a brief about a relationship rather than a set of drafted replies. ## Related - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops - https://www.polarishq.co/use-cases/customer-support/escalation-tracking - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "AI Paralegal That Prepares, a Lawyer Decides" description: "An AI paralegal that builds contract checklists, tracks obligations and dates, and prepares document summaries for a qualified lawyer to review and decide on." url: https://www.polarishq.co/ai-workers/paralegal section: AI workers updated: 2026-08-21 --- # Hire an AI paralegal It organises the paperwork and surfaces the dates and clauses, and it stops well before the point where anything becomes advice. ## The short answer An AI paralegal in Polaris is a worker you hire to prepare and organise legal administration: summarising documents you provide, building clause checklists, tracking obligations, renewal dates and notice periods as tasks, and watching public sources for mentions. A qualified lawyer reviews everything and makes every decision. Polaris does not provide legal advice. - **Typical output:** Obligations tracker plus summary - **Core connections:** Google Drive, Gmail, web search - **Illustrative task:** 2.9 human-hours, $5.80 > **Read this before hiring one** > > Polaris does not provide legal advice and an AI paralegal is not a lawyer. Everything this worker produces is preparation for a qualified person to review, correct and decide on. Do not use it to interpret an obligation, assess a risk or determine whether a term is enforceable, and do not let its output reach a counterparty without legal review. ## What this worker is great at Finding the dates. A drawer of vendor agreements contains renewal dates, notice periods and auto-renew clauses that nobody has in a calendar, and missing one of them costs real money. Given the documents, the worker returns a tracker: one row per agreement, with the date, the notice window and the exact clause it came from, plus a task for each deadline with the reminder set before the window closes. Checklists are the second strength. The clauses your lawyer always looks for, written once into the skill file, applied to every incoming contract, with each item marked present, absent or unclear and the relevant text quoted. It is a reading aid that makes the review faster, not a substitute for the review. Public monitoring is the third. Trademark mentions and similar public filings and usages, searched on a cadence you set, delivered with sources and no interpretation attached. ## What to connect it to - **Google Drive** — Where the executed agreements and templates live. Read access is enough for the preparation work. - **Gmail** — Contracts and notices arrive by email, and the tracker is only current if it reflects what actually came in. - **Web search** — For public monitoring, and for finding the published source behind a reference. Never for determining what the law is. - **Slack** — So a request to review a document becomes a prefilled task rather than a message somebody forgets to act on. ## Where the human stays in charge This boundary is the whole design of the role. **The paralegal prepares** - A summary of what a document says - A clause checklist with quoted text - An obligations and dates tracker - Tasks for every deadline, dated and owned **A qualified lawyer decides** - What any clause means and what it exposes you to - Whether to accept, amend or reject a term - Every piece of advice given to the business - Anything that is signed, filed or sent to a counterparty ## What an agreement review pass costs Illustrative, computed from the published human-hours formula: nine vendor agreements summarised into one tracker. | Observed effort | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 8,100 characters of written output at 90 chars/min | 90 | | 5 acceptance criteria ticked at 8 min each | 40 | | 1 progress comment, 2 files produced | 30 | | Total: 2.9 human-hours at $2 | $5.80 | ## What an AI paralegal is not good at Anything that requires judgement about the law. Law is jurisdictional, it changes, and the consequences of being confidently wrong are borne by your company rather than by the worker. Where it is unsure it should mark the item unclear and stop, and the skill file should instruct it to do exactly that. It also cannot tell you what is missing from a contract that should be there. Absence is a lawyer's judgement, informed by what usually goes wrong in your industry, and a checklist only catches the omissions somebody already thought to list. It does not file, sign, serve or negotiate. It has no view of matters, dockets or case management systems, and it should never be the last reader of a document before it leaves the building. ## Questions people ask **Can an AI paralegal give legal advice?** No, and it should never be asked to. Polaris does not provide legal advice. The worker prepares summaries, checklists and trackers, and a qualified lawyer in the relevant jurisdiction reviews the material and makes every decision. **Can it review a contract before we sign?** It can prepare the review: summarise the terms, apply your clause checklist, quote the relevant text and list the dates and obligations. Whether to sign is a decision for your lawyer, working from the document rather than from the summary. **How does it handle a clause it does not understand?** It marks the item unclear and quotes the text rather than guessing, provided the skill file instructs it to. Writing that instruction explicitly is the single most important line in this worker's brief. **Is our contract data safe?** Connections are authorised once, verified live and stored server-side, and a browser cannot read the credentials back. Beyond that, treat documents you put into any tool according to your own confidentiality obligations, and check with your lawyer before uploading privileged material. **Does it replace a paralegal?** No. It removes some of the organising and summarising load, and every output still routes to a qualified person. Teams that use it well treat it as preparation that makes legal review faster, not as a way to review less. ## Related - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/use-cases/legal/contract-review-tracking - https://www.polarishq.co/use-cases/legal/vendor-agreement-management - https://www.polarishq.co/use-cases/legal/compliance-checklists - https://www.polarishq.co/use-cases/legal/policy-updates - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/glossary/human-in-the-loop --- --- title: "AI Financial Analyst for Variance and Board Packs" description: "An AI financial analyst that writes variance commentary, drafts board-pack narrative and documents scenario assumptions. A qualified person checks every number." url: https://www.polarishq.co/ai-workers/financial-analyst section: AI workers updated: 2026-08-21 --- # Hire an AI financial analyst You supply the figures; it writes the explanation of what moved, why it matters and which assumption the whole thing rests on. ## The short answer An AI financial analyst in Polaris is a worker you hire for the written half of finance work: variance commentary from figures you provide, board-pack narrative, scenario write-ups with every assumption stated, and documented reporting structures held steady period to period. A qualified person reviews the numbers and owns every conclusion. Polaris does not provide financial or investment advice. - **Typical output:** Variance commentary and narrative - **Core connections:** Stripe, Supabase, Google Drive - **Illustrative task:** 2.8 human-hours, $5.60 ## What this worker is great at Writing the commentary. Budget versus actual is a table anybody can produce and a paragraph almost nobody has time to write: which lines moved, by how much, which movements are timing rather than trend, and which single variance explains most of the gap. Hand it the figures and the commentary comes back in the same structure every month, which is what makes month nine readable next to month two. Board narrative is the second piece. The sections, the order, the sentence that says what the quarter was about, and clearly marked gaps where only the finance lead can supply a number or a judgement. It is faster to correct a structured draft than to face an empty document at eleven at night. Scenario write-ups are the third. Given a set of assumptions it produces the write-up with each assumption named and isolated, so the conversation becomes an argument about the assumptions rather than about the spreadsheet. ## What to connect it to Read-only credentials are the right default for a role that touches money. - **Stripe** — Revenue, refunds and invoices, restricted to read-only scopes. - **Supabase** — For product data that sits behind a metric, when the analysis has to reconcile usage with revenue. - **Google Drive** — Where the models and the board packs live and where the output has to land. - **Web search** — For benchmark conventions and how a ratio is usually defined, cited rather than assumed. ## Where the human stays in charge **The analyst prepares** - Variance commentary on figures you supply - Board-pack narrative with gaps marked - Scenario write-ups with assumptions isolated - A reporting structure that stays consistent **A qualified person decides** - Whether the numbers are right - What the company does about them - Everything that reaches a board or an investor - Any statement that could be read as advice ## What a monthly commentary costs Illustrative, computed from the published human-hours formula: one budget variance pack written in a single session. | Observed effort | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 3 live web searches at 12 min each | 36 | | 5,400 characters of finished prose at 90 chars/min | 60 | | 4 acceptance criteria ticked at 8 min each | 32 | | 1 progress comment, 1 doc, 1 PDF | 25 | | Total: 2.8 human-hours at $2 | $5.60 | > **What an AI financial analyst is not good at** > > It is not a substitute for a qualified finance professional, and Polaris does not provide financial or investment advice. It cannot audit a number, reconcile a ledger or catch a figure that was wrong before you pasted it in: garbage in produces well-written garbage out, which is more dangerous than the badly written kind. It does not build or run models in a spreadsheet, and it should never be the last reader of anything that goes to a board, a bank or an investor. ## Questions people ask **Can an AI financial analyst give financial advice?** No. Polaris does not provide financial or investment advice. The worker prepares written analysis from figures you supply, and a qualified person reviews the numbers and owns every conclusion drawn from them. **Does it connect to our accounting system?** The connection catalog covers Stripe, Supabase and Google Drive among others, and there is no general accounting integration. In practice you provide the figures in the task or through Drive, and the worker writes the analysis around them. **Can it build a financial model?** It writes the structure, the assumptions and the commentary; it does not build or calculate a spreadsheet model. The most useful division is a person owning the model and the worker owning everything written around it. **How does it handle assumptions?** Every assumption is named and isolated in the write-up rather than buried in a sentence, and where it had to choose one it says so. That is what makes the output arguable, which is the only property that matters in a scenario document. ## Related - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/use-cases/finance/monthly-close-checklist - https://www.polarishq.co/use-cases/executive/board-reporting - https://www.polarishq.co/use-cases/data/reporting-automation - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/integrations/supabase --- --- title: "AI Ops Coordinator for SOPs and Vendor Work" description: "An AI ops coordinator that writes SOPs people actually follow, prepares vendor comparisons with sourced pricing and keeps cross-team work moving." url: https://www.polarishq.co/ai-workers/ops-coordinator section: AI workers updated: 2026-08-21 --- # Hire an AI ops coordinator It writes down how your company does things, so the answer stops living in whichever colleague happens to be on holiday. ## The short answer An AI ops coordinator in Polaris is a worker you hire to document and move operational work. It writes standard operating procedures from what people describe, prepares vendor comparisons with sourced pricing, keeps cross-team dependencies visible as tasks, and maintains the checklists that recurring processes run on. People approve the process, choose the vendor and sign anything binding. - **Typical output:** SOP set plus a comparison table - **Core connections:** Slack, Notion, Google Drive - **Illustrative task:** 3.2 human-hours, $6.40 ## What this worker is great at Writing procedures that survive contact with a new hire. Somebody describes the process out loud in whatever order it comes out, and the worker returns numbered steps, the decision points marked, the exceptions listed at the end rather than tangled into the middle, and the questions it could not answer at the top. That last list is usually where the actual process problem is. Vendor work is the second strength. Four suppliers, the published prices, what each one includes at that price, the terms that differ, and a source for every figure. It produces the comparison table your finance lead asks for and rarely receives, and it marks the fields where the vendor publishes nothing so the gap is visible instead of blank. The third is keeping cross-team work honest. Operations is largely the work of noticing that team A is waiting on team B and nobody has said so out loud. The coordinator writes that down every week, with names on it. ## What to connect it to - **Slack** — Where the operational reality is discussed. The Inbox turns those messages into prefilled task suggestions rather than losing them in scrollback. - **Notion** — When the process wiki lives there and new procedures have to extend it rather than start a rival library. - **Google Drive** — For the procedures, comparison tables and templates the rest of the company already looks for there. - **Web search** — For vendor pricing and terms, cited page by page so the comparison is checkable. ## A realistic first week 1. **Monday: the process only one person knows** — Whatever breaks when a specific colleague is away. Dictate it into the task and get the numbered version back with the gaps listed. 2. **Tuesday: a vendor comparison** — Four suppliers for something you are about to buy. Acceptance criteria: a source URL for every price, and unpublished figures marked as unpublished rather than estimated. 3. **Thursday: the recurring checklists** — Turn the monthly operational routine into a task template with the acceptance criteria written in. 4. **Friday: the blocked list** — Who is waiting on whom, across every workstream, in one document with names and dates. ## Where the human stays in charge **The coordinator prepares** - Numbered procedures with gaps flagged - Vendor comparisons with sourced pricing - A cross-team blocked list, named and dated - Checklists for recurring operational work **You do** - Approve the process as the way it is done - Choose the vendor and negotiate terms - Place orders and sign contracts - Decide what the team stops doing ## What an AI ops coordinator is not good at It cannot buy anything. No orders, no bookings, no signatures, no commitments to a supplier, all of which is deliberate. It also cannot tell you whether the vendor is any good to deal with once the contract is signed, because that information is in your network and not on a pricing page. It has no view of the physical world. Stock counts, deliveries and whether the room is actually free are things a person tells it, and a procedure it writes will describe what was described to it rather than what happens on the floor. And it will faithfully document a bad process. Handed a workflow with four unnecessary approval steps, it produces an excellent write-up of four unnecessary approval steps. Ask it separately which steps look redundant and it will tell you, but it will not volunteer that judgement in the middle of a procedure. ## Work this role picks up - [use-cases/operations/process-documentation](https://www.polarishq.co/use-cases/operations/process-documentation) - [use-cases/operations/sop-maintenance](https://www.polarishq.co/use-cases/operations/sop-maintenance) - [use-cases/operations/vendor-procurement](https://www.polarishq.co/use-cases/operations/vendor-procurement) - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) ## Questions people ask **Can an AI ops coordinator place orders with suppliers?** No. It prepares the comparison and the recommendation with sources, and a person negotiates, orders and signs. Polaris workers deliver written work as comments and take no outbound action on your behalf. **How does it write an SOP for something undocumented?** Somebody describes it in the task, in any order, and the worker returns the ordered version with decision points marked and the unanswered questions listed at the top. Those questions are the fastest route to a procedure that actually holds. **Can it keep procedures up to date?** Assign it a recurring review task. It compares what the document says against what you tell it has changed and returns a list of statements that no longer match, which is a shorter and more useful output than a rewritten page. **What does a set of four SOPs cost?** Roughly 3.2 human-equivalent hours under the published formula, which is $6.40 at $2 per human-hour. The example is illustrative; your job's hours appear on the work log with the effort they were derived from and can be challenged there. ## Related - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/use-cases/operations/process-documentation - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/use-cases/operations/vendor-procurement - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/workstream --- --- title: "Alternatives to Notion, Jira, Linear, Slack and 12 more" description: "Honest alternative pages for 16 work tools, with pricing checked in August 2026, what each tool still does better, and where Polaris fits instead." url: https://www.polarishq.co/alternatives section: Alternatives updated: 2026-08-21 --- # Alternatives to the work tools you pay for by the seat Sixteen pages about tools teams leave, what leaving actually costs, and when leaving is the wrong call. ## The short answer An alternative to a per-seat work tool is any product that stores the same work under a different billing model. Polaris is one: project management, docs and team chat in a single free workspace, plus AI workers that run on cloud machines and deliver tasks as comments with files. The software costs nothing. Billing is roughly two dollars per human-hour of work delivered. - **Tools covered:** 16 - **Polaris software:** $0, no seats - **Billing unit:** ~$2 per human-hour delivered - **Pricing checked:** 21 August 2026 ## Why anyone searches for an alternative Almost nobody types a tool name plus the word alternative because the tool broke. They type it because the renewal came in, or because a fourth subscription arrived for something the other three already half did, or because a new hire pushed the bill past a number someone noticed. So these pages lead with the bill and the sprawl, not with feature checklists. Every one of them quotes the competitor's list price from a live check on 21 August 2026, names the version of the product it checked, and says plainly which situations the incumbent still wins. ## How these pages are built Same skeleton on all sixteen, so they can be read against each other. - **Real prices with a date** — List price per seat, per month, at the tier a normal team lands on, taken from the vendor's own pricing page where that page could be read. Where it could not, the page says so instead of guessing. - **A when-to-stay section that means it** — Jira's workflow engine, Linear's cycle discipline, Airtable's relational model, Confluence's space permissions and Basecamp's flat rate are all genuine reasons to stay put. Each page names its own. - **The migration answer, including the boring parts** — What exports cleanly, what does not, and which tools you do not have to leave at all. Slack, Notion, Linear and GitHub are in the Polaris connection catalog, so connecting beats migrating for most teams. - **One philosophical argument, repeated on purpose** — Trackers track work. Per-seat pricing charges for places to put work. Polaris charges for work that came back finished, itemised on a work log you can challenge line by line. ## The two billing models on the table **Per seat, per month** - Cost rises with headcount whether or not the new person opens the tool - A second tool doubles the count, a third triples it - AI arrives as another per-seat add-on stacked on the base plan - Nothing on the invoice tells you what got done **Free software, metered work** - Unlimited humans, tasks, workstreams and docs at no cost - Chief of Staff included in every org from the first sign-in - Billing starts only when an AI worker delivers something - Every hour is estimated by an open formula and logged job by job > **Say the honest part first** > > Polaris is in free public beta. There are no customers to point at, no logos and no case studies. What exists is a working product, four recorded demos including an unstaged machine session, and a pricing model that bills nothing until an AI worker hands back finished work. ## Every alternative page - [A Notion alternative for teams tired of maintaining the database](https://www.polarishq.co/alternatives/notion) — Notion is the best place to write the plan. The question is who executes it once the page is written. - [A Jira alternative that does some of the work, not just the tracking](https://www.polarishq.co/alternatives/jira) — Jira is the most configurable tracker ever built. Configuration is also the reason people go looking. - [A Trello alternative for teams who outgrew ten boards](https://www.polarishq.co/alternatives/trello) — Trello made work visible to everyone. Visible work is still work waiting for a person. - [A Linear alternative for the work Linear was never meant to hold](https://www.polarishq.co/alternatives/linear) — Linear is the best issue tracker most engineers have used. This page is not going to pretend otherwise. - [A Slack alternative where the message becomes a task, not scrollback](https://www.polarishq.co/alternatives/slack) — Slack is where the signal arrives. It is also where the signal goes to be forgotten by Thursday. - [An Asana alternative for teams pushed off the free plan](https://www.polarishq.co/alternatives/asana) — The third person on the team is now a billing event. That is why this search exists. - [A ClickUp alternative where AI is not a second subscription](https://www.polarishq.co/alternatives/clickup) — The everything app, plus an everything AI add-on, priced per seat on top of the seat you already bought. - [A monday.com alternative that is one product, not a product line](https://www.polarishq.co/alternatives/monday) — Four products, four price lists, four places your team's work now lives. - [A Basecamp alternative for teams who already refused per-seat pricing](https://www.polarishq.co/alternatives/basecamp) — Basecamp settled the pricing argument years ago. This page is about a different argument. - [A Confluence alternative for documentation that maintains itself](https://www.polarishq.co/alternatives/confluence) — The wiki is not out of date because the tool is bad. It is out of date because updating it is nobody's actual job. - [A Todoist alternative for the point where a list becomes a team](https://www.polarishq.co/alternatives/todoist) — Todoist is a superb personal task manager. Most people searching this are not looking for a better one. - [An Airtable alternative for teams billed per editor](https://www.polarishq.co/alternatives/airtable) — Airtable is a real database with a spreadsheet face. That is the strength, and it is why this comparison has limits. - [A Height alternative, and the difference between two kinds of autonomy](https://www.polarishq.co/alternatives/height) — Two products both say autonomous. They mean different things, and the difference is the whole decision. - [A Shortcut alternative for teams who just crossed the free seat limit](https://www.polarishq.co/alternatives/shortcut) — Shortcut sits between Jira's configuration and Linear's opinion. The sixth engineer is where the invoice starts. - [A Wrike alternative for teams buying seats in blocks](https://www.polarishq.co/alternatives/wrike) — Seat bands, minimum user counts and subscriptions sold in blocks of five. The pricing page is the product review. - [A Coda alternative for teams caught by the Superhuman rebrand](https://www.polarishq.co/alternatives/coda) — The product you bought is now part of somebody else's bundle. That is a reasonable moment to look around. ## Related reading - [compare](https://www.polarishq.co/compare) - [replace](https://www.polarishq.co/replace) - [cost](https://www.polarishq.co/cost) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [integrations](https://www.polarishq.co/integrations) ## Questions people ask **Is Polaris really free, or is there a seat charge hiding somewhere?** The software is free with no seat count and no tier. Unlimited humans, tasks, workstreams and docs, plus the Chief of Staff, are included in every organisation. Revenue comes from one place only: roughly two dollars per human-equivalent hour that an AI worker delivers. **What is the catch on two dollars per hour?** The catch, if it is one, is that hours are estimated rather than measured with a stopwatch, because a machine does not experience time the way a person does. The estimate uses an open formula built from observable effort such as searches run, prose produced and files returned, clamped between five minutes and eight hours per session. Every line is written to the worker's work log and you can challenge any of them. **How current is the competitor pricing on these pages?** All sixteen pages were checked on 21 August 2026 against the vendors' own pricing pages. Two vendors truncated their pages and one did not respond, and those pages say so explicitly rather than filling the gap from memory. Treat any published price as a starting point and confirm at the vendor before you sign. **Do I have to leave my current tools to use Polaris?** No, and for four of the sixteen the better answer is to stay. Slack, Notion, Linear and GitHub are in the Polaris connection catalog, so a worker can read and act in them while your team carries on working where it already works. Running Polaris alongside an existing tracker for one workstream is the normal way to start. **Which of these tools does Polaris genuinely not replace?** Airtable as a relational database, Confluence as a governed enterprise wiki, Jira as a configurable workflow engine and Slack as a real-time chat backbone with shared external channels all do things Polaris does not. Each of those pages says so in its own words. ## Related - https://www.polarishq.co/compare - https://www.polarishq.co/replace - https://www.polarishq.co/cost - https://www.polarishq.co/integrations - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/for --- --- title: "Notion alternative: free workspace with AI that works" description: "Notion charges $10 to $20 per seat per month plus AI credits, checked August 2026. Polaris is free, and its AI workers deliver tasks instead of drafting text." url: https://www.polarishq.co/alternatives/notion section: Alternatives updated: 2026-08-21 --- # A Notion alternative for teams tired of maintaining the database Notion is the best place to write the plan. The question is who executes it once the page is written. ## The short answer Polaris is a free alternative to Notion for teams that want docs, tasks and team chat in one workspace without per-seat billing. Notion lists Plus at ten dollars and Business at twenty dollars per seat per month, checked 21 August 2026, with AI agents billed separately through credits. Polaris charges nothing for software and roughly two dollars per human-hour an AI worker delivers. - **Notion Plus:** $10 / seat / month - **Notion Business:** $20 / seat / month - **Polaris software:** $0, unlimited seats - **Checked:** 21 Aug 2026 ## The actual reason teams look Notion databases start as a good idea and end as a second job. A team builds a properties schema, adds relations and rollups, writes a template, and then discovers that the schema only stays true if a person updates it. Six weeks later the status column is a museum of the last time anyone cared. The bill arrives on top of that. Checked on 21 August 2026, Notion lists Free at zero, Plus at ten dollars per seat per month, Business at twenty dollars per seat per month, and Enterprise on request, with up to twenty percent off for yearly billing. Notion also sells AI on credits: Custom Agents are free to try and then ten dollars per one thousand monthly Notion credits, and a Workers beta is listed as starting to consume credits on 15 October. So the direction of travel is the same on both products. The difference is what you are metered on. Notion meters AI on top of a seat you already bought. Polaris has no seat to buy and meters only finished work. ## Head to head Notion prices from notion.com/pricing, checked 21 August 2026. | | Notion | Polaris | | --- | --- | --- | | Software cost | $10/seat/mo Plus, $20/seat/mo Business | $0, unlimited humans, no tiers | | AI cost | Credits, from $10 per 1,000 monthly credits | ~$2 per human-hour of delivered work, itemised | | Docs | Block editor, databases, relations, rollups, formulas, public sites | Nested doc tree, block editor with markdown shortcuts, versioning, file review, comments | | Tasks | A database you configure yourself | Focus lane across today, this week and next 30 days, plus workstreams shared between list and board views | | Team chat | Comments and mentions on pages | Team chat in the product, plus an Inbox that turns Slack messages into prefilled task suggestions | | Who does the work | People, with AI drafting alongside them | People and AI workers on the same members table, assigned identically | | Where AI runs | Inside the Notion product | A cloud machine that wakes per task and keeps running after you close the laptop | | What AI hands back | Text in the page you are editing | A comment on the task with generated files attached, acceptance criteria ticked, closed by a human | ## When you should stay on Notion Four situations where Notion is the right tool and Polaris is not. - **The database is the product** — Relations, rollups, formulas and linked views let a non-engineer model something genuinely complex, and Notion is one of the few tools where that holds up past the prototype. Polaris does not have a user-defined database layer. - **You publish from your workspace** — Notion Sites turns an internal page into a public one with a URL and a theme. Teams running a public changelog, a careers page or a docs site from Notion would be giving up a working publishing pipeline. - **Your company runs on Notion templates** — A decade of community templates, and an internal library your team has already customised, is real accumulated work. Rebuilding it has a cost that no comparison table shows. - **Writing is the whole job** — For a research or editorial team where the output is the document, Notion's editor plus its AI drafting is a shorter path than briefing a worker and waiting for a delivery. ## Moving, or not moving Notion is in the Polaris connection catalog, so the honest first move is usually to connect rather than migrate. 1. **Connect Notion instead of exporting it** — Authorise Notion once from the connections catalog. The credential is verified live and stored server-side, and browsers cannot read it back. AI workers can then read and write in the Notion workspace you already have. 2. **Bring one workstream across, not the whole wiki** — Pick a single project with real deadlines. Recreate it as a Polaris workstream with lanes, and leave the reference documentation where it lives. Nothing forces a big-bang move. 3. **Know what does not survive an export** — Notion exports pages as Markdown or HTML and databases as CSV. Relations, rollups, formulas, synced blocks and permission settings do not come across as working objects in any tool. Plan for the schema to be rebuilt or dropped, not moved. 4. **Hire the worker that maintains what you used to maintain** — Describe the upkeep that keeps slipping in chat. The Chief of Staff runs a short interview, writes a SKILL.md you can read and edit, and the worker card lands ready for its first task in about a minute. > **The one-line difference** > > Notion is where the plan is written. Polaris is where the plan is written and where a machine wakes up to do part of it, then posts the result as a comment for a human to close. ## Go deeper - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [compare/notion-vs-confluence](https://www.polarishq.co/compare/notion-vs-confluence) - [compare/notion-vs-airtable](https://www.polarishq.co/compare/notion-vs-airtable) - [replace/notion-and-jira-and-slack](https://www.polarishq.co/replace/notion-and-jira-and-slack) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) ## Questions people ask **Can Polaris import a Notion workspace?** There is no one-click Notion importer. Notion exports pages to Markdown or HTML and databases to CSV, and those files can be pasted into Polaris docs, but relations, rollups and formulas do not survive any export in any tool. Most teams connect Notion through the connections catalog and leave the reference material where it is. **Is Polaris genuinely free, or free until it is not?** The software is free with unlimited humans, tasks, workstreams and docs, and there is no seat count anywhere in the product. Revenue comes entirely from delivered agent work at roughly two dollars per human-equivalent hour. Free-forever is only credible when the money is visibly coming from somewhere else, and here it is. **How does Polaris AI differ from Notion AI?** Notion AI works inside the page you are editing and produces text there. A Polaris worker is assigned a task like a colleague, runs on a cloud machine with real web search and tool access, ticks its own acceptance criteria, and posts the result as a comment with any files it generated. The machine never closes the task; a human does. **What happens to Notion's per-seat cost if I run both?** It stays exactly where it is. Notion lists Plus at ten dollars and Business at twenty dollars per seat per month as of 21 August 2026, and connecting Notion to Polaris does not change that. The saving only appears if you eventually stop paying for seats in one of the products. **Does Polaris have a block editor like Notion?** Yes. Docs in Polaris are a nested tree with a block editor, markdown shortcuts, to-dos and sub-pages, plus versioning, file review and comments. It is not a database engine, and it does not try to be one. ## Related - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/compare/notion-vs-airtable - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/replace/notion-and-jira-and-slack - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/alternatives/coda --- --- title: "Jira alternative for teams who want the work done too" description: "Jira lists Standard near $7.91 and Premium near $14.54 per user monthly, checked August 2026. Polaris is free software, billed only by delivered hour." url: https://www.polarishq.co/alternatives/jira section: Alternatives updated: 2026-08-21 --- # A Jira alternative that does some of the work, not just the tracking Jira is the most configurable tracker ever built. Configuration is also the reason people go looking. ## The short answer Polaris is a free alternative to Jira for teams that want issue tracking, docs and team chat in one product and want AI workers to take some of the tickets. Jira lists Free for up to ten users, Standard near seven dollars ninety-one and Premium near fourteen dollars fifty-four per user per month, checked 21 August 2026. Polaris software costs nothing. - **Jira Free:** up to 10 users - **Jira Standard:** ~$7.91 / user / month - **Jira Premium:** ~$14.54 / user / month - **Checked:** 21 Aug 2026 ## Nobody leaves Jira because it lacks features They leave because of the admin tax. Someone owns the workflow schemes. Someone owns the custom fields nobody fills in. Someone owns the board configuration that broke when a new project reused the wrong scheme. That person did not apply for the job, and their real job is not getting done. Then there is the price shape. Checked on 21 August 2026, Jira is free for up to ten users, with Standard at roughly seven dollars ninety-one and Premium at roughly fourteen dollars fifty-four per user per month at small team sizes, and Enterprise quoted on request. Atlassian's per-user rate slides down as the user count rises, which is worth modelling before comparing against anything. The eleventh person is where most teams first feel it, because that is the moment the free plan ends and a per-user line item begins. ## Head to head Jira prices are cloud list prices at small team sizes, checked 21 August 2026. Atlassian's own page was partially unreadable, so figures were cross-checked against multiple published 2026 pricing summaries. | | Jira | Polaris | | --- | --- | --- | | Software cost | Free to 10 users, then per user per month | $0 at any team size, no seat count | | Workflow model | Custom workflow schemes, statuses, transitions, conditions, validators, post-functions | Lanes shared between list and board view, plus a pinned Focus lane across three time horizons | | Query language | JQL, saved filters, dashboards built on them | No query language. Filtering and lanes only | | Docs | Sold separately as Confluence | Included: nested tree, block editor, versioning, review and comments | | Team chat | Not included | Included, plus an Inbox that turns Slack signals into prefilled task suggestions | | Assignees | Humans | Humans and AI workers in one members table, assigned the same way | | AI work | Assistive features priced into higher tiers | A cloud machine wakes per task, runs a tool loop with live web search, and returns files | | Who closes a ticket | A human | A human. The machine ticks acceptance criteria and delivers, it never marks work done | ## When you should stay on Jira These are not consolation prizes. If any of them describes you, do not move. - **Your workflow engine is load-bearing** — Conditions, validators, post-functions and multi-project schemes encode real process, often process an auditor has seen. Polaris has lanes and a Focus horizon. It does not have a workflow engine and will not pretend to. - **You need permission schemes and audit depth** — Issue-level security, project roles, permission schemes and Atlassian's admin and audit surface exist because large organisations require them. That requirement does not go away because a cheaper product appeared. - **You run Jira Service Management or a marketplace app you depend on** — The Atlassian app marketplace is deep, and a team whose intake, SLAs or compliance reporting runs on a marketplace app is buying that app as much as the tracker. - **Reporting is measured against Jira's own reports** — Sprint burndown, velocity, control charts and cumulative flow are the shared vocabulary of a lot of engineering organisations. Losing them mid-quarter is a real cost, not a rounding error. ## What changes on a Monday morning Same sprint, two products. **On Jira** - Twelve tickets in the backlog, all waiting for a person - Three of them are research, writing or chasing, not engineering - Standup spends four minutes on tickets nobody has started - Every new hire adds a per-user line to the invoice **On Polaris** - The same twelve tasks, with three assigned to AI workers - A cloud machine wakes for each one and posts progress as it goes - Deliverables arrive as comments with files attached, ready for review - The invoice moves only when something was actually delivered ## Migration, honestly There is no Jira importer in Polaris. Jira exports issues to CSV and has a well-documented REST API, so a bulk move of summary, description, status, assignee and labels is achievable with a script. Workflow schemes, permission schemes, custom field configurations, sprint history and marketplace app data are not portable to any product, Polaris included. The realistic pattern is narrower. Leave Jira running for the engineering work that depends on its schemes. Start one Polaris workstream for the work that keeps falling outside the tracker: research, writing, competitor checks, documentation upkeep, the recurring chase. Connect GitHub so an AI worker can see the code context. Judge it on whether that lane empties. ## Go deeper - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [compare/jira-vs-linear](https://www.polarishq.co/compare/jira-vs-linear) - [compare/jira-vs-asana](https://www.polarishq.co/compare/jira-vs-asana) - [replace/jira-and-confluence-and-slack](https://www.polarishq.co/replace/jira-and-confluence-and-slack) - [integrations/github](https://www.polarishq.co/integrations/github) - [use-cases/engineering/sprint-planning](https://www.polarishq.co/use-cases/engineering/sprint-planning) ## Questions people ask **Can Polaris import Jira issues?** Not with a one-click importer. Jira exports issues to CSV and exposes a REST API, which covers summary, description, status, assignee and labels well enough for a scripted move. Workflow schemes, permission schemes, custom field configuration and sprint history do not transfer to any tool, so treat them as rebuilt or retired rather than migrated. **Is Polaris a like-for-like Jira replacement for a 60-engineer org?** For most organisations that size, no. Jira's workflow engine, permission schemes and reporting vocabulary are usually embedded in how the org runs, and replacing them is a project rather than a switch. Polaris fits better as the place the non-ticket work lives, running alongside Jira, until the case for a wider move is obvious. **How much does Jira actually cost per year?** Checked on 21 August 2026, Jira Cloud is free to ten users, with Standard near seven dollars ninety-one and Premium near fourteen dollars fifty-four per user per month at small team sizes, and the per-user rate falling as the user count grows. A twenty-person team on Standard is therefore in the region of nineteen hundred dollars a year before Confluence or any marketplace app. **What does two dollars per hour buy on an engineering task?** It buys a session on a cloud machine that compiles the worker's instructions and skill files, runs a live tool loop with web search, ticks the acceptance criteria you wrote, and posts a comment with any files produced. Hours are human-equivalent and estimated by an open formula from observable effort, then logged so you can challenge any line. **Does Polaris connect to GitHub the way Jira does?** GitHub is in the Polaris connection catalog, authorised once and stored server-side, so an AI worker can be given repository access as part of its brief. It is not the same as Jira's smart-commit and branch-linking integration, and if that specific link is central to your workflow, that is a reason to keep Jira. ## Related - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/compare/jira-vs-asana - https://www.polarishq.co/replace/notion-and-jira - https://www.polarishq.co/replace/jira-and-confluence-and-slack - https://www.polarishq.co/integrations/github - https://www.polarishq.co/alternatives/confluence - https://www.polarishq.co/alternatives/linear --- --- title: "Trello alternative: boards plus teammates who do the cards" description: "Trello Free caps at 10 boards per Workspace and Standard is $5 per user monthly, checked August 2026. Polaris is free, with AI workers that clear cards." url: https://www.polarishq.co/alternatives/trello section: Alternatives updated: 2026-08-21 --- # A Trello alternative for teams who outgrew ten boards Trello made work visible to everyone. Visible work is still work waiting for a person. ## The short answer Polaris is a free alternative to Trello for small teams that hit the ten-board limit or want more than a board. Trello Free allows ten boards and ten collaborators per Workspace, with Standard at five dollars and Premium at ten dollars per user per month billed annually, checked 21 August 2026. Polaris adds docs, team chat and AI workers who deliver tasks. - **Trello Free:** 10 boards / Workspace - **Trello Standard:** $5 / user / month annual - **Trello Premium:** $10 / user / month annual - **Checked:** 21 Aug 2026 ## The wall people hit Trello is the easiest tool in this category to start. That is not faint praise; getting a whole non-technical team to adopt anything is the hard part, and Trello does it in an afternoon. The wall arrives later, and it is usually one of two. The first is structural. Checked on 21 August 2026, Trello Free allows up to ten boards and ten collaborators per Workspace, ten megabytes per file and two hundred and fifty Workspace command runs a month. Standard is five dollars per user per month billed annually or six billed monthly, Premium is ten annually or twelve fifty monthly, and Enterprise is seventeen dollars fifty per user per month billed annually, listed as two hundred and ten dollars per user per year. The second is that a board full of cards is an accurate picture of what has not happened yet. Trello never claimed otherwise. The question worth asking in 2026 is whether some of those cards can be picked up by something other than a person. ## Head to head Trello prices from trello.com/pricing, checked 21 August 2026. | | Trello | Polaris | | --- | --- | --- | | Free plan limits | 10 boards and 10 collaborators per Workspace, 10 MB files, 250 command runs a month | No limits on humans, tasks, workstreams or docs | | Paid cost | $5/user/mo Standard, $10 Premium, $17.50 Enterprise, billed annually | $0 for software at any size | | Views | Board, plus table, calendar, timeline and dashboard on paid tiers | List and board sharing the same lanes, plus a Focus lane across today, this week and next 30 days | | Automation | Butler rules, with monthly command run quotas by tier | AI workers assigned tasks like colleagues, no rule quota | | Docs | Card descriptions and attachments | A nested doc tree with block editor, versioning, file review and comments | | Chat | Card comments | Team chat in the product, plus a Slack-fed Inbox of prefilled task suggestions | | Who clears the card | A person, every time | A person or an AI worker on a cloud machine that returns files as a comment | | What you pay for | Places to put cards | Human-equivalent hours of work delivered, about $2 each, itemised on a work log | ## When you should stay on Trello Trello wins on the axis that matters most for some teams: nobody has to be taught. - **Your collaborators are not software people** — Volunteers, contractors, a client, a committee. Trello's board is understood on sight with no training, and that is a genuine advantage no feature list captures. - **Butler already runs your process** — Teams with mature Butler rules driving card movement, due dates and checklists have automation that works today, tuned by someone who understood the problem. - **You are one person with a personal board** — At a team of one, five dollars a month for Standard is not a problem worth solving, and Trello's Free plan may cover you outright. - **Power-Ups connect you to something specific** — If a Power-Up is bridging Trello to a system you depend on, that bridge is part of the product you bought. ## The board-count arithmetic What the ten-board ceiling costs when a team crosses it. Trello list prices checked 21 August 2026. - **$5** — Trello Standard, per user per month. Billed annually, $6 billed monthly - **$1,200** — Ten people on Trello Standard, per year. Before Premium views or any Power-Up - **$0** — Polaris software, ten people or a hundred. No seats, no board cap, no tier ## Moving from Trello Trello is one of the friendlier exports in this category. Each board exports as JSON or CSV containing lists, cards, descriptions, labels, due dates, checklists and comments, which maps cleanly onto Polaris workstreams, lanes and tasks. Attachments come across as links rather than files, so anything stored only on a card needs pulling down first. Butler rules do not transfer, and there is no equivalent object to import them into. The replacement is different in kind: instead of a rule that moves a card when a date passes, you brief a worker on what should happen and assign it the task. That is better for judgement work and worse for deterministic housekeeping, and it is worth knowing which of the two your rules are. ## Go deeper - [cost/trello-pricing](https://www.polarishq.co/cost/trello-pricing) - [compare/trello-vs-asana](https://www.polarishq.co/compare/trello-vs-asana) - [compare/trello-vs-monday](https://www.polarishq.co/compare/trello-vs-monday) - [replace/trello-and-slack](https://www.polarishq.co/replace/trello-and-slack) - [replace/notion-and-trello](https://www.polarishq.co/replace/notion-and-trello) - [for/small-business](https://www.polarishq.co/for/small-business) ## Questions people ask **Does Polaris have a board view like Trello?** Yes, and it shares its lanes with the list view, so a workstream is organised once and read either way. Everything drags. The extra piece is the Focus lane, which is pinned first in every view and holds the non-negotiables across today, this week and the next thirty days. **Can I import my Trello boards?** There is no one-click Trello importer, but Trello's JSON and CSV export contains lists, cards, descriptions, labels, due dates, checklists and comments, which lines up well with the Polaris task model. Attachments export as links, so download anything that lives only on a card before you move. **What replaces my Butler automations?** Nothing does, directly. Butler is deterministic rule automation and Polaris does not have a rules engine. What Polaris has instead is AI workers you brief and assign tasks to, which suits judgement work like research or drafting far better than it suits moving a card when a date passes. **Is Polaris free forever or is this a beta price?** The software is free with no seat count, and that is the model rather than a promotion. Polaris is in free public beta with no customers yet, and the revenue line is delivered agent work at roughly two dollars per human-equivalent hour, logged and challengeable job by job. **We are five people and Trello Free is fine. Why switch?** If you are inside the ten-board and ten-collaborator limits and nobody is waiting on work that a machine could do, there is no argument here worth making. The case starts when cards sit untouched because the person who owns them has no time, which is the specific problem an AI worker on a cloud machine addresses. ## Related - https://www.polarishq.co/cost/trello-pricing - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/compare/trello-vs-monday - https://www.polarishq.co/replace/trello-and-slack - https://www.polarishq.co/replace/notion-and-trello - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/for/small-business --- --- title: "Linear alternative, or the thing you run next to Linear" description: "Linear Free caps at 250 issues, Basic is $10 and Business $16 per user monthly on annual billing, checked August 2026. Polaris is free and connects to Linear." url: https://www.polarishq.co/alternatives/linear section: Alternatives updated: 2026-08-21 --- # A Linear alternative for the work Linear was never meant to hold Linear is the best issue tracker most engineers have used. This page is not going to pretend otherwise. ## The short answer Polaris is a free alternative to Linear for teams that want tasks, docs and team chat in one workspace and want AI workers to deliver some of the tasks. Linear lists Free with two hundred and fifty issues and two teams, Basic at ten dollars and Business at sixteen dollars per user per month billed annually, checked 21 August 2026. Linear is also a Polaris connection, so both can run together. - **Linear Free:** 250 issues, 2 teams - **Linear Basic:** $10 / user / month annual - **Linear Business:** $16 / user / month annual - **Checked:** 21 Aug 2026 ## Start with the concession, because it is large Linear is exceptional at what it does. Keyboard-first navigation that a fast engineer never leaves, an opinionated issue model that resists the configuration rot Jira invites, cycles that create real cadence, triage that actually gets used, and a git integration that closes issues without anyone thinking about it. Teams do not merely tolerate Linear. They defend it. So the honest framing for this page is not that Linear is worse. It is that Linear is a tracker with a per-seat price, checked on 21 August 2026 at ten dollars per user per month on Basic and sixteen on Business billed annually, with the Free plan capped at two hundred and fifty issues and two teams. Every issue in it is still waiting for a person. The interesting question for a Linear team is not which tracker to use. It is who picks up the issues that keep sliding across cycles because they are research, writing or chasing rather than code. ## Head to head Linear prices from linear.app/pricing, checked 21 August 2026. Enterprise is annual billing only. | | Linear | Polaris | | --- | --- | --- | | Software cost | $10/user/mo Basic, $16 Business, billed annually | $0, no seats, no tiers | | Free plan | 250 issues, 2 teams, 10 MB uploads | Unlimited humans, tasks, workstreams and docs | | Speed model | Keyboard-first, command menu, sub-second navigation | Drag-first, lanes shared between list and board, a pinned Focus lane | | Planning unit | Cycles, with automatic rollover and velocity | Three time horizons: today, this week, next 30 days | | Docs | Documents and project overviews | Nested doc tree, block editor, versioning, file review, comments | | Team chat | Not included | Included, plus an Inbox that turns Slack messages into prefilled task suggestions | | Non-engineering work | Possible, but the model is built around software issues | Workstreams for any container of work, including client and marketing work | | Assignees | Humans | Humans and AI workers in the same members table | | Delivery | A human moves the issue to done | An AI worker posts a comment with files and ticks acceptance criteria; a human closes it | ## When you should stay on Linear The strongest stay-case in this whole cluster. Read it before anything else here. - **Your engineers navigate by keyboard** — Linear's command menu and shortcut coverage remove a real tax from a working day. Polaris is drag-first and does not match that, and telling a Linear-fluent engineer to use a mouse more is a genuine downgrade. - **Cycles are how your team keeps its promises** — Cycle discipline, automatic rollover and the velocity that comes out of it are a planning system, not a view. Teams that run on cycles should keep running on cycles. - **Triage and the issue model are doing real work** — Linear's opinionated model is why Linear workspaces stay clean for years while other trackers accumulate custom fields nobody can delete. That opinion is the feature. - **Your git workflow depends on the integration** — Branch naming, pull request linking and automatic status transitions are tight enough that engineers stop noticing them. That is a high bar to replace. ## The realistic setup Most Linear teams who try Polaris end up here rather than migrating. **Stays in Linear** - Every engineering issue, cycle and project - Triage, branch links and pull request automation - The velocity and cycle reporting the team plans against - Whatever your engineers have shortcuts memorised for **Moves to Polaris** - Research, writing, competitor checks and documentation upkeep - The cross-functional work that has no home in an engineering tracker - Anything you would assign to an AI worker rather than a person - The Focus lane for the three things that must not slip this week ## Connect rather than migrate Linear is in the Polaris connection catalog. Authorise it once, org-wide, and the credential is verified live and held server-side where browsers cannot read it back. An AI worker can then be briefed with Linear access as part of its skill file, and work in the issues your team already keeps. If a full move is genuinely what you want, Linear has a documented GraphQL API and CSV export covering issues, labels, states, assignees and estimates. Cycles, triage rules and the git links do not have a counterpart in Polaris, so plan on losing them rather than moving them. ## Go deeper - [cost/linear-pricing](https://www.polarishq.co/cost/linear-pricing) - [compare/jira-vs-linear](https://www.polarishq.co/compare/jira-vs-linear) - [compare/linear-vs-shortcut](https://www.polarishq.co/compare/linear-vs-shortcut) - [compare/linear-vs-height](https://www.polarishq.co/compare/linear-vs-height) - [replace/linear-and-notion-and-github](https://www.polarishq.co/replace/linear-and-notion-and-github) - [integrations/linear](https://www.polarishq.co/integrations/linear) ## Questions people ask **Should a Linear team actually switch to Polaris?** Usually not wholesale. Linear's keyboard speed, cycle discipline and git integration are hard to give up and the engineering work is better off staying. The common pattern is to connect Linear to Polaris and use Polaris for the research, writing and cross-functional work that engineers keep pushing out of the tracker. **Does Polaris have keyboard shortcuts as good as Linear's?** No. Polaris is drag-first, with lanes shared between list and board views and a Focus lane pinned above them. If sub-second keyboard navigation is what your team optimises for, that is a real reason to keep Linear as the engineering tracker. **What does the Linear connection let an AI worker do?** Linear is one of fourteen connections in the Polaris catalog, authorised once and stored server-side. A worker briefed with Linear access can be given work that involves reading and acting in your Linear workspace, alongside the other tools in its skill file such as GitHub and web search. **How does Polaris pricing compare to Linear at twenty people?** Linear Business at sixteen dollars per user per month billed annually is three thousand eight hundred and forty dollars a year for twenty people, checked 21 August 2026. Polaris software for the same twenty people is zero, and the only bill is delivered agent work at roughly two dollars per human-equivalent hour. If your workers deliver nothing, you pay nothing. **Is Polaris ready for a production engineering team?** Polaris is in free public beta with no customers and no case studies, and it does not have cycles, a query language or a git-linked issue model. Treating it as a second surface next to Linear is an honest place to start; treating it as a drop-in replacement for a shipping engineering org is not. ## Related - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/compare/linear-vs-shortcut - https://www.polarishq.co/compare/linear-vs-height - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/replace/linear-and-slack - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent --- --- title: "Slack alternative that turns messages into owned tasks" description: "Slack Pro is $7.25 per user monthly on annual billing and Free keeps 90 days of history, checked August 2026. Polaris chat is free and turns Slack into tasks." url: https://www.polarishq.co/alternatives/slack section: Alternatives updated: 2026-08-21 --- # A Slack alternative where the message becomes a task, not scrollback Slack is where the signal arrives. It is also where the signal goes to be forgotten by Thursday. ## The short answer Polaris is a free alternative to Slack for teams that want team chat attached to the work rather than beside it. Slack lists Free with ninety days of message history, Pro at seven dollars twenty-five per active user per month billed annually, and Business plus at fifteen dollars, checked 21 August 2026. Polaris includes chat at no cost and converts Slack messages into prefilled task suggestions. - **Slack Free:** 90 days of history - **Slack Pro:** $7.25 / user / month annual - **Slack Business+:** $15 / user / month annual - **Checked:** 21 Aug 2026 ## The ninety-day problem, and the real one Checked on 21 August 2026, Slack's Free plan keeps ninety days of message history, Pro is eight dollars seventy-five per active user per month billed monthly or seven twenty-five billed annually, Business plus is eighteen billed monthly or fifteen annually, and Enterprise plus is quoted on request. The history limit is what makes teams start looking. The deeper problem is not retention. It is that a decision, a request and a complaint all arrive in the same stream, at the same size, with the same urgency signal, and only one of the three ever becomes something with an owner and a date. The rest is archaeology. Polaris does not attempt to be a better chat product than Slack. It attempts to make the chat stop being the last place a piece of work was seen. ## Head to head Slack prices from slack.com/pricing, checked 21 August 2026. | | Slack | Polaris | | --- | --- | --- | | Cost | Free with 90-day history, then $7.25 to $15 per active user per month annually | $0, unlimited humans, full history | | Real-time chat | Channels, threads, huddles, calls, presence | Team chat in the product, without huddles or voice | | External collaboration | Slack Connect shared channels with other companies | Not available. This is a genuine gap | | Message to task | Save for later, reminders, or an app that creates a ticket elsewhere | Inbox catches the signal and prefills a task suggestion with bucket, lane, labels and owner set | | Tasks and docs | Not included, lives in the tools you connect | Included: workstreams, Focus lane, nested docs with versioning and review | | AI in chat | Assistive features on paid tiers | The Chief of Staff is on your roster from day one and turns noise into approved tasks | | Who acts on a message | A person who has to notice it | A person, or an AI worker assigned the resulting task | | App directory | Thousands of third-party apps | Fourteen connections, authorised once, org-wide, stored server-side | ## How a Slack message becomes owned work Slack is in the Polaris connection catalog, so this works without leaving Slack. 1. **Connect Slack once** — Authorise Slack from the connections catalog. The credential is verified live at connection time and stored server-side, org-wide. Browsers cannot read it back. 2. **The signal lands in the Inbox** — A message your tools hear arrives in the Polaris Inbox as a prefilled task suggestion, with bucket, lane, labels and owner already set. Signals become suggestions and never silent tasks. 3. **One click makes it real** — Accept the suggestion and it becomes a task in a workstream with an owner and a date. Reject it and it goes away. Nothing is created behind your back. 4. **Assign it to whoever should do it** — Humans and AI workers sit in the same members table, so assignment is identical for both. If it goes to a worker, a cloud machine wakes for it and returns the result as a comment with any files it produced. ## When you should stay on Slack Slack does several things Polaris does not do at all. - **You work with other companies in shared channels** — Slack Connect is how a lot of agencies, vendors and partners actually communicate. Polaris has no equivalent, and losing shared channels would move those conversations to email. - **Huddles and calls are part of the day** — Ad-hoc audio, screen sharing and the low-ceremony huddle culture that grew around them are not in Polaris. Teams that live in huddles should keep Slack. - **Compliance, retention and eDiscovery are requirements** — Enterprise retention policies, legal holds, data export APIs and the surrounding administration exist because regulated organisations must have them. - **Your alerting runs through Slack apps** — Incident tooling, deploy bots, on-call rotations and monitoring alerts route through Slack's app directory. That plumbing is worth more than a subscription line. > **The recommendation most teams should take** > > Keep Slack. Connect it. Let the Inbox turn the requests buried in it into tasks with an owner and a date, and assign the ones that do not need a human to an AI worker. Replacing Slack outright is the wrong fight for almost everybody, and pretending otherwise would be the kind of claim you could check in a minute. ## Go deeper - [cost/slack-pricing](https://www.polarishq.co/cost/slack-pricing) - [compare/slack-vs-microsoft-teams](https://www.polarishq.co/compare/slack-vs-microsoft-teams) - [replace/jira-and-slack](https://www.polarishq.co/replace/jira-and-slack) - [replace/notion-and-slack](https://www.polarishq.co/replace/notion-and-slack) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) ## Questions people ask **Does Polaris replace Slack completely?** For most teams, no, and the page above says which parts are missing. Polaris has team chat but no huddles, no shared external channels and no third-party app directory. The stronger move is to connect Slack so the requests inside it become tasks with owners, and keep Slack for conversation. **Will connecting Slack create tasks without me knowing?** No. A Slack signal arrives in the Polaris Inbox as a suggestion with bucket, lane, labels and owner prefilled, and it stays a suggestion until someone accepts it. Signals become suggestions, never silent tasks. That rule exists because the alternative is a task list nobody trusts. **What does Slack cost a twenty-person team per year?** Slack Pro at seven dollars twenty-five per active user per month billed annually is one thousand seven hundred and forty dollars a year for twenty people, checked 21 August 2026. Business plus at fifteen dollars is three thousand six hundred. Polaris chat is included in free software with no seat count. **How is Slack authorised, and can anyone read the token?** Connections are authorised once and apply org-wide. Credentials are verified live at connection time and stored server-side, and the browser cannot read them back. AI workers use the connection through the runtime rather than through your session. **Does Polaris keep full message history on the free plan?** Yes. There is no retention tier in Polaris because there are no tiers at all. The software is free with unlimited humans, tasks, workstreams and docs, and the only billing line is roughly two dollars per human-equivalent hour of delivered agent work. ## Related - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/compare/slack-vs-microsoft-teams - https://www.polarishq.co/replace/jira-and-slack - https://www.polarishq.co/replace/notion-and-linear-and-slack - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/for/remote-teams --- --- title: "Asana alternative after the free plan dropped to 2 seats" description: "Asana Personal now caps at 2 users and Starter is $10.99 per user monthly on annual billing, checked August 2026. Polaris is free for unlimited people." url: https://www.polarishq.co/alternatives/asana section: Alternatives updated: 2026-08-21 --- # An Asana alternative for teams pushed off the free plan The third person on the team is now a billing event. That is why this search exists. ## The short answer Polaris is a free alternative to Asana for teams that no longer fit its two-user Personal plan. Asana lists Personal at zero for up to two users, Starter at ten dollars ninety-nine and Advanced at twenty-four dollars ninety-nine per user per month billed annually, checked 21 August 2026. Polaris has no seat count, includes docs and team chat, and adds AI workers who deliver tasks. - **Asana Personal:** 2 users - **Asana Starter:** $10.99 / user / month annual - **Asana Advanced:** $24.99 / user / month annual - **Checked:** 21 Aug 2026 ## What changed, and why the searches spiked Asana's free tier used to be one of the most generous in the category. It is now called Personal and, checked on 21 August 2026, caps at two users. Reporting on the change indicates it applies to accounts created after 12 November 2025, with older accounts keeping their previous allowance. Verify which side of that line your workspace falls on before you plan anything. Above it, Starter is ten dollars ninety-nine per user per month billed annually or thirteen forty-nine billed monthly, Advanced is twenty-four ninety-nine annually or thirty forty-nine monthly, and Enterprise is quoted on request. For a marketing or operations team of twelve, Starter is around fifteen hundred and eighty dollars a year for a shared task list. That is a defensible price for what Asana does. It is a harder price to defend when the list is full of items nobody has started, which is the condition most cross-functional teams are actually in. ## Head to head Asana prices from asana.com/pricing, checked 21 August 2026. | | Asana | Polaris | | --- | --- | --- | | Free tier | Personal, up to 2 users | Unlimited humans, tasks, workstreams and docs | | Paid cost | $10.99/user/mo Starter, $24.99 Advanced, billed annually | $0 software at any team size | | Process automation | Rules, forms, approvals, bundles | AI workers you brief and assign, with an editable SKILL.md each | | Portfolio view | Portfolios, Goals, Workload on higher tiers | Workstreams plus a pinned Focus lane across today, this week and next 30 days | | Docs | Project briefs and task descriptions | Nested doc tree with block editor, versioning, file review and comments | | Team chat | Not included | Included, plus a Slack-fed Inbox of prefilled task suggestions | | Assignees | One human owner per task | Humans and AI workers in the same members table, assigned identically | | What arrives when work is done | A person marks it complete | A comment with the deliverable and any files, acceptance criteria ticked, closed by a human | ## When you should stay on Asana Asana earned its position with cross-functional teams for reasons that still hold. - **Rules, forms and approvals are your intake process** — A well-built Asana intake form that routes requests into the right project with the right custom fields is a genuine operational asset. Polaris has no rules engine and no form builder. - **You manage capacity with Workload** — Resource levelling across a department is a specific job, and Workload with effort estimates does it. Polaris has no capacity planning surface. - **Goals and portfolios roll up to an executive** — If a leadership team reads Asana portfolios and goal progress every month, replacing that reporting is a change management project, not a tool swap. - **Adoption took a year and it finally stuck** — Asana is unusually good at getting non-technical departments to actually update their tasks. That habit has more value than the licence cost. ## What twelve people cost List prices checked 21 August 2026, before any add-on. - **$1,582** — Twelve people on Asana Starter, per year. $10.99 per user per month, billed annually - **$3,599** — Twelve people on Asana Advanced, per year. $24.99 per user per month, billed annually - **$0** — Polaris software, twelve people. Billing starts only when an AI worker delivers work ## Moving from Asana Asana exports projects to CSV and JSON, covering task names, descriptions, assignees, due dates, sections and custom field values. That maps onto Polaris workstreams, lanes and tasks without much argument. Attachments come as links, so pull anything down that exists only on a task. Rules, forms, bundles, portfolios and Workload have no destination. There is no equivalent object in Polaris to receive them, so they are retired rather than migrated. Before committing, list which of them are load-bearing. If the intake form is how requests reach your team at all, that alone is a reason to keep Asana running. ## Go deeper - [cost/asana-pricing](https://www.polarishq.co/cost/asana-pricing) - [compare/asana-vs-monday](https://www.polarishq.co/compare/asana-vs-monday) - [compare/trello-vs-asana](https://www.polarishq.co/compare/trello-vs-asana) - [compare/wrike-vs-asana](https://www.polarishq.co/compare/wrike-vs-asana) - [replace/asana-and-slack](https://www.polarishq.co/replace/asana-and-slack) - [use-cases/marketing/campaign-planning](https://www.polarishq.co/use-cases/marketing/campaign-planning) ## Questions people ask **Is the Asana free plan really limited to two users now?** Asana's pricing page listed Personal at two users when checked on 21 August 2026, and published reporting ties the change to accounts created after 12 November 2025 with older workspaces keeping their previous allowance. Check your own workspace rather than assuming, because the answer differs by account age. **Can Polaris do Asana-style rules and forms?** No. There is no rules engine and no form builder in Polaris. The replacement is different in kind: you hire an AI worker in about sixty seconds through a chat interview, its capabilities become a SKILL.md file you can read and edit, and you assign it tasks the same way you assign a colleague. **How do I move Asana projects into Polaris?** Export projects from Asana as CSV or JSON, which carries task names, descriptions, assignees, due dates, sections and custom field values, then rebuild them as Polaris workstreams and lanes. Attachments export as links, so download those first. Rules, forms, portfolios and Workload do not transfer. **What is the catch on two dollars per human-hour?** The hour is an estimate rather than a stopwatch reading, because a cloud machine does not experience time like a person. The estimate comes from an open formula using observable effort such as searches run, prose produced and files returned, clamped between five minutes and eight hours per session, and every line lands on the worker's work log where you can challenge it. **Does Polaris have anything like Asana Goals?** Not as a named feature. The nearest thing is the Focus lane, which is pinned first in every view and holds non-negotiables across today, this week and the next thirty days. It is a commitment device rather than a goal-tracking hierarchy, and it does not roll up into executive reporting. ## Related - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/compare/jira-vs-asana - https://www.polarishq.co/compare/basecamp-vs-asana - https://www.polarishq.co/replace/asana-and-notion - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/use-cases/marketing/campaign-planning --- --- title: "ClickUp alternative without the per-seat AI add-on" description: "ClickUp Business is $12 per user monthly and Everything AI adds $28 per user, checked August 2026. Polaris software is free and bills only delivered work." url: https://www.polarishq.co/alternatives/clickup section: Alternatives updated: 2026-08-21 --- # A ClickUp alternative where AI is not a second subscription The everything app, plus an everything AI add-on, priced per seat on top of the seat you already bought. ## The short answer Polaris is a free alternative to ClickUp for teams paying twice: once for seats and again for AI. ClickUp lists Unlimited at seven dollars and Business at twelve dollars per user per month billed annually, with Brain AI at nine dollars and Everything AI at twenty-eight dollars per user per month, checked 21 August 2026. Polaris software is free and bills roughly two dollars per delivered human-hour. - **ClickUp Business:** $12 / user / month annual - **Everything AI add-on:** +$28 / user / month annual - **Polaris software:** $0, AI included in the product - **Checked:** 21 Aug 2026 ## Two bills for one team Checked on 21 August 2026, ClickUp lists Free Forever at zero, Unlimited at seven dollars per user per month billed annually or ten billed monthly, Business at twelve annually or nineteen monthly, and Enterprise on request. AI is sold separately: Brain AI at nine dollars per user per month billed annually and Everything AI at twenty-eight. A ten-person team on Business with Everything AI is therefore four hundred dollars a month at list, before anyone has produced anything. The seat charge and the AI charge both scale with headcount rather than with output, which means the eleventh hire raises the AI bill whether or not that person ever uses AI. The other reason teams look elsewhere is configuration debt. ClickUp's range of views, custom fields, statuses and automations is real, and so is the state a workspace reaches eighteen months in when four people configured it differently and nobody remembers why. ## Head to head ClickUp prices from clickup.com/pricing, checked 21 August 2026. | | ClickUp | Polaris | | --- | --- | --- | | Software cost | $7/user/mo Unlimited, $12 Business, billed annually | $0 at any team size | | AI cost | Brain AI $9 or Everything AI $28 per user per month, annually | No AI seat. ~$2 per human-hour of work delivered | | What AI does | Assists inside the workspace: writing, summarising, answering | Runs on a cloud machine as an assigned teammate and returns finished work with files | | Views | List, board, calendar, Gantt, timeline, mind map, table and more | List and board sharing the same lanes, plus a pinned Focus lane over three horizons | | Docs and whiteboards | Docs, whiteboards, time tracking, goals | Docs with versioning, review and comments. No whiteboard, no time tracking | | Configuration burden | High. Statuses, custom fields, spaces, folders, automations | Low by design. Workstreams and lanes, dragged into place | | Assignees | Humans | Humans and AI workers in one members table | | Bill shape | Rises with every hire | Rises only when a worker delivers something you can inspect | ## When you should stay on ClickUp ClickUp gives you more surface area than Polaris does. Sometimes that is exactly what you need. - **You need views Polaris does not have** — Gantt charts, mind maps, workload views and whiteboards are all in ClickUp and none of them are in Polaris. Teams running dependency-heavy plans need a Gantt, and no argument about pricing changes that. - **Time tracking is part of how you bill** — ClickUp has native time tracking with estimates and reporting. Agencies billing clients by tracked hours would be removing a piece of their invoicing chain. - **The Free Forever plan already covers you** — ClickUp's free tier is one of the most capable in this category. A small team living inside it comfortably has no cost argument to answer. - **Custom fields carry your data model** — Teams that have modelled a real business process in ClickUp custom fields and automations have built something. Polaris has no custom field system to receive it. ## Where the money goes A ten-person team, list prices checked 21 August 2026. **ClickUp Business with Everything AI** - $12 per user per month for the workspace - $28 per user per month for the AI tier - $4,800 a year at ten people, before Enterprise features - The bill is the same whether AI produced anything or not **Polaris** - $0 for the workspace, at ten people or a hundred - Chief of Staff included in every org from day one - ~$2 per human-equivalent hour that a worker delivers - Nothing delivered, nothing billed, and every hour is on the work log ## Moving from ClickUp ClickUp exports tasks to CSV and has a documented API covering tasks, lists, custom fields, statuses and comments. The parts that map cleanly are the ones every tool shares: name, description, assignee, due date, status, comments. The parts that do not are the ones you spent the most time on. Automations, dashboards, whiteboards, Gantt dependencies, custom field schemas and time entries have no destination in Polaris. If the honest audit says half your ClickUp value is in those, stay. If the honest audit says your workspace is a task list with an expensive AI add-on bolted to it, the comparison is straightforward. ## Go deeper - [cost/clickup-pricing](https://www.polarishq.co/cost/clickup-pricing) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) - [compare/clickup-vs-monday](https://www.polarishq.co/compare/clickup-vs-monday) - [compare/clickup-vs-notion](https://www.polarishq.co/compare/clickup-vs-notion) - [replace/clickup-and-slack](https://www.polarishq.co/replace/clickup-and-slack) - [glossary/usage-based-pricing](https://www.polarishq.co/glossary/usage-based-pricing) ## Questions people ask **How much does ClickUp AI cost on top of a ClickUp plan?** Checked on 21 August 2026, ClickUp lists Brain AI at nine dollars per user per month and Everything AI at twenty-eight dollars per user per month, both billed annually, in addition to the workspace plan. A ten-person team on Business with Everything AI is around four hundred dollars a month at list price. **Does Polaris charge extra for AI?** There is no AI seat and no AI tier. The Chief of Staff is on your roster in every organisation from the first sign-in, and hiring more workers is free. The only charge is roughly two dollars per human-equivalent hour that a worker actually delivers, itemised on that worker's work log. **Does Polaris have Gantt charts and whiteboards?** No to both. Polaris has list and board views that share the same lanes, plus a Focus lane pinned across today, this week and the next thirty days. Teams that plan dependencies on a Gantt or run workshops on a whiteboard would be losing something real. **Can I export my ClickUp workspace?** ClickUp exports tasks to CSV and exposes an API covering tasks, lists, custom fields, statuses and comments, which is enough to move the task data. Automations, dashboards, whiteboards, Gantt dependencies and time entries do not have a counterpart in Polaris and should be treated as retired. **Is Polaris actually free, or free for now?** The software is free with unlimited humans, tasks, workstreams and docs, and no seat count exists in the product. Polaris is in free public beta with no customers and no revenue yet, and the intended revenue line is delivered agent work rather than seats. That is the whole model, stated plainly. ## Related - https://www.polarishq.co/cost/clickup-pricing - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/replace/clickup-and-slack - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/glossary/usage-based-pricing --- --- title: "monday.com alternative without four separate products" description: "monday.com sells Work Management, CRM, Service and Dev separately, from $9 per seat monthly, checked August 2026. Polaris is one free workspace with AI workers." url: https://www.polarishq.co/alternatives/monday section: Alternatives updated: 2026-08-21 --- # A monday.com alternative that is one product, not a product line Four products, four price lists, four places your team's work now lives. ## The short answer Polaris is a free alternative to monday.com for teams that do not want to buy several products from one vendor. monday.com lists Work Management Basic at nine dollars, Standard at twelve and Pro at nineteen per seat per month billed annually, with CRM, Service and Dev priced separately, checked 21 August 2026. Polaris is one workspace, free, with AI workers billed by delivered hour. - **monday Free:** up to 2 seats - **Work Management Standard:** $12 / seat / month annual - **Work Management Pro:** $19 / seat / month annual - **Checked:** 21 Aug 2026 ## The product line is the problem monday.com is no longer one thing. Checked on 21 August 2026, the pricing page lists Work Management at nine, twelve and nineteen dollars per seat per month billed annually across Basic, Standard and Pro, with a free tier capped at two seats. Alongside it sit monday CRM at twelve, seventeen and twenty-eight dollars per seat annually, monday Service from thirty-one, and monday Dev at nine, twelve and twenty dollars. For a company where sales, support and product all adopted monday, that is three subscriptions with three seat counts inside one vendor relationship. The colourful boards were supposed to end tool sprawl, and instead the sprawl moved inside the invoice. The second reason teams look is that boards, automations and dashboards describe work extremely well and perform none of it. That is not a criticism of monday specifically. It is true of every tracker in this cluster. ## Head to head monday.com prices from monday.com/pricing, checked 21 August 2026. Figures shown are Work Management billed annually. | | monday.com | Polaris | | --- | --- | --- | | Free tier | 2 seats | Unlimited humans, tasks, workstreams and docs | | Paid cost | $9 Basic, $12 Standard, $19 Pro per seat per month annually | $0 software at any team size | | Product shape | Work Management, CRM, Service and Dev sold separately | One product covering tasks, docs and team chat | | Automations | Recipe-based automations with monthly action quotas by tier | AI workers you brief and assign, each with an editable SKILL.md | | Dashboards | Widget dashboards across boards | Worker activity feed and insights, plus a work log per delivered job | | Docs | monday workdocs | Nested doc tree with block editor, versioning, file review and comments | | Team chat | Board updates and mentions | Team chat in the product, plus an Inbox of prefilled task suggestions from Slack | | Who does the work | People, prompted by automations | People and AI workers, assigned identically from the same members table | ## When you should stay on monday.com monday earns its price in specific places, and these are the real ones. - **Non-technical teams adopted it and use it daily** — monday's visual boards get finance, HR and operations updating status without a training programme. That behaviour is the hardest thing in this category to buy, and it is worth protecting. - **You run a real CRM or service desk on it** — monday CRM and monday Service are proper products with pipeline stages, ticketing and SLAs. Polaris has neither, and HubSpot is available only as a connection an AI worker can use. - **Dashboards go to people who never open a task** — Executive dashboards assembled from board widgets are often the only view leadership has. Rebuilding that reporting is a separate project with its own politics. - **Automation recipes are doing deterministic work** — Status changes, notifications and date-driven moves are exactly what recipe automation is good at. AI workers are the wrong tool for that job and Polaris has no rules engine. ## Moving from monday.com Boards export to Excel and CSV, carrying items, groups, column values, owners and dates, and there is a documented API for a scripted move. Items become Polaris tasks, groups become lanes, and a board becomes a workstream. That part is unremarkable. Automation recipes, dashboard widgets, board views and anything living in monday CRM or monday Service have no destination. Before starting, count how many of your seats are on Work Management versus the other products, because a team whose real dependency is monday CRM is not comparing like with like when it reads this page. > **The argument in one sentence** > > monday.com charges per seat, per product, for places to put work. Polaris charges nothing for the workspace and roughly two dollars per human-equivalent hour that an AI worker hands back finished, itemised on a log you can dispute line by line. ## Go deeper - [cost/monday-pricing](https://www.polarishq.co/cost/monday-pricing) - [compare/asana-vs-monday](https://www.polarishq.co/compare/asana-vs-monday) - [compare/clickup-vs-monday](https://www.polarishq.co/compare/clickup-vs-monday) - [compare/trello-vs-monday](https://www.polarishq.co/compare/trello-vs-monday) - [replace/monday-and-slack](https://www.polarishq.co/replace/monday-and-slack) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) ## Questions people ask **How much does monday.com cost for a team of fifteen?** On Work Management Standard at twelve dollars per seat per month billed annually, fifteen seats is two thousand one hundred and sixty dollars a year, checked 21 August 2026. Pro at nineteen dollars is three thousand four hundred and twenty. Teams also using monday CRM or monday Service pay for those seats separately. **Does Polaris have a CRM?** No. HubSpot is in the Polaris connection catalog, so an AI worker can be given access to your existing CRM as part of its brief, but there is no pipeline, deal or contact object inside Polaris. If monday CRM is where your revenue process lives, that is a reason to keep it. **Can I export my monday boards?** Boards export to Excel and CSV with items, groups, column values, owners and dates, and monday has a documented API for larger moves. Items map to Polaris tasks and groups to lanes. Automation recipes, dashboards and anything in the CRM or Service products do not transfer. **What replaces monday automation recipes?** Nothing does like for like. Polaris has no rules engine. It has AI workers who are hired through a short chat interview, get a readable and editable SKILL.md, and are assigned tasks the way a colleague is. That handles judgement work well and deterministic housekeeping poorly. **Is Polaris free because it is unfinished?** Polaris is in free public beta with no customers yet, and the software is free by design rather than by stage. There are no seats and no tiers anywhere in the pricing model, and the revenue comes from delivered agent work at roughly two dollars per human-equivalent hour. ## Related - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/compare/trello-vs-monday - https://www.polarishq.co/replace/monday-and-slack - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/cost/stack-cost-25-person-team --- --- title: "Basecamp alternative: flat rate, plus teammates who work" description: "Basecamp charges flat rates from $25 to $300 a month with no per-user fees, checked August 2026. Polaris is free software billed only on delivered agent hours." url: https://www.polarishq.co/alternatives/basecamp section: Alternatives updated: 2026-08-21 --- # A Basecamp alternative for teams who already refused per-seat pricing Basecamp settled the pricing argument years ago. This page is about a different argument. ## The short answer Polaris is a free alternative to Basecamp for teams that want the same flat-fee philosophy plus AI workers who deliver tasks. Basecamp lists flat plans with no per-user fees at twenty-five, fifty-nine, one hundred and three hundred dollars a month, checked 21 August 2026. Polaris charges nothing for the software and roughly two dollars per human-equivalent hour of delivered work. - **Basecamp Studio:** $59 / month, unlimited users - **Basecamp Pro:** $100 / month, 25 projects - **Polaris software:** $0, unlimited projects - **Checked:** 21 Aug 2026 ## Give Basecamp its due first Basecamp has argued against per-seat pricing for longer than most of the tools on this site have existed, and it did not just argue. Checked on 21 August 2026, the pricing page lists Free at zero for one project, five users and one gigabyte, Freelancer at twenty-five dollars a month for three active projects and twenty users, Studio at fifty-nine for ten active projects and unlimited users, Pro at one hundred for twenty-five projects, and the Unlimited Edition at three hundred a month billed annually for unlimited projects and a terabyte of storage. No per-user fees anywhere. Adding your eleventh person costs nothing. That is the same instinct behind Polaris charging zero for software, and it would be dishonest to write a page implying Basecamp overcharges. The argument is elsewhere. Basecamp caps active projects instead of people, which is a different squeeze on a growing agency, and more importantly it holds work without doing any of it. Every message, to-do and Hill Chart in Basecamp is waiting for a human. ## Head to head Basecamp prices from basecamp.com/pricing, checked 21 August 2026. | | Basecamp | Polaris | | --- | --- | --- | | Pricing model | Flat monthly fee, no per-user charge | Free software, metered only on delivered agent work | | What is capped | Active projects and storage by tier | Nothing. Unlimited humans, tasks, workstreams and docs | | Entry cost | $25/mo Freelancer, $59 Studio, $100 Pro, $300 Unlimited Edition | $0 | | Progress view | Hill Charts, a genuinely original way to show uncertainty | Focus lane across today, this week and next 30 days, plus lanes shared by list and board | | Client access | Clientside, built for sharing a project with a client | No client-facing surface | | Docs | Docs and files per project | Nested doc tree, block editor, versioning, file review, comments | | Chat | Campfire chat and Message Board per project | Team chat in the product, plus a Slack-fed Inbox of task suggestions | | Who does the work | People | People and AI workers assigned identically from one members table | ## When you should stay on Basecamp The most credible stay-case in this cluster, and worth taking seriously. - **Clientside is part of how you deliver** — Sharing a project surface with a client, with a clean boundary between internal and external, is a core agency workflow. Polaris has nothing equivalent, and losing it would push client communication back to email. - **Hill Charts tell your team something no status field does** — The distinction between figuring it out and doing it is real and almost every other tool ignores it. If your team reads Hill Charts, that is a habit worth keeping. - **The calm philosophy is why you chose it** — Basecamp is deliberately quiet, deliberately opinionated and deliberately not trying to be an AI product. Teams that value that will not want AI workers in the workspace, and that is a coherent position. - **Flat rate at a large headcount is hard to beat** — At sixty people, the Unlimited Edition at three hundred dollars a month is five dollars per person per month for the whole workspace. On price alone, that already beats most of this category. > **Where the two products actually differ** > > Basecamp solved the pricing problem: nobody should pay more because the team grew. Polaris takes the same position and adds a second one: you should not pay for software at all, and the thing worth paying for is work that came back finished with the files attached. ## Moving from Basecamp Basecamp offers a full account export as HTML with attachments, which is excellent for archiving and poor for importing anywhere. To-do lists and to-dos map onto Polaris lanes and tasks conceptually, but expect manual rebuilding rather than a clean pipe, and expect to keep the export as the historical record. Hill Charts, Campfire history, Clientside boundaries and automatic check-ins have no counterpart. For an agency, the honest test is whether client-facing work is the majority of what lives in Basecamp. If it is, keep Basecamp and use Polaris for the internal work you would otherwise hire a contractor for. ## Go deeper - [compare/basecamp-vs-asana](https://www.polarishq.co/compare/basecamp-vs-asana) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [for/agencies](https://www.polarishq.co/for/agencies) - [for/small-business](https://www.polarishq.co/for/small-business) - [replace](https://www.polarishq.co/replace) ## Questions people ask **Is Basecamp cheaper than Polaris?** No, but the comparison is closer than with any other tool here. Basecamp's flat plans run from twenty-five to three hundred dollars a month with no per-user fees, checked 21 August 2026. Polaris software is zero at any headcount, and the only charge is roughly two dollars per human-equivalent hour that an AI worker delivers. **Does Polaris have anything like Hill Charts?** No. The nearest concept is the Focus lane, which sorts non-negotiables across today, this week and the next thirty days and is pinned first in every view. That expresses time pressure rather than the uncertainty Hill Charts express, and they are not substitutes for each other. **Can I share a Polaris workstream with a client the way Basecamp does?** Not today. There is no Clientside equivalent and no external-guest surface in Polaris. Agencies that share project spaces with clients should keep Basecamp for that and consider Polaris for internal delivery work instead. **Why would a Basecamp team add Polaris at all?** Because Basecamp holds the work and someone still has to do it. The specific case is work you would otherwise hand to a freelancer: research, competitor checks, first drafts, documentation upkeep. An AI worker takes the task, runs on a cloud machine, and posts the result as a comment with any files it produced. **How do I export from Basecamp?** Basecamp provides a full account export in HTML including attachments. It is built for archiving rather than importing, so plan on rebuilding to-do lists as Polaris lanes manually and keeping the export as your historical record. ## Related - https://www.polarishq.co/compare/basecamp-vs-asana - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/for/agencies - https://www.polarishq.co/for/small-business - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/glossary/per-seat-pricing --- --- title: "Confluence alternative for wikis nobody updates" description: "Confluence is free to 10 users and roughly $5.42 to $10.44 per user monthly above, checked August 2026. Polaris is free and gives upkeep to an AI worker." url: https://www.polarishq.co/alternatives/confluence section: Alternatives updated: 2026-08-21 --- # A Confluence alternative for documentation that maintains itself The wiki is not out of date because the tool is bad. It is out of date because updating it is nobody's actual job. ## The short answer Polaris is a free alternative to Confluence for teams whose documentation rots faster than they can maintain it. Confluence is free for up to ten users and lists roughly five dollars forty-two on Standard and ten dollars forty-four on Premium per user per month at small team sizes, checked 21 August 2026. Polaris includes docs, tasks and chat free, and documentation upkeep can be assigned to an AI worker. - **Confluence Free:** up to 10 users, 2 GB - **Confluence Standard:** ~$5.42 / user / month - **Confluence Premium:** ~$10.44 / user / month - **Checked:** 21 Aug 2026 ## Why wikis decay, and what a price change has to do with it Every Confluence space starts with an onboarding page, a runbook and an architecture note written by someone who cared. Two quarters later the runbook references a service that was renamed, the architecture note describes a system that was replaced, and a new engineer reads all three and trusts none of them. Confluence did not cause that. What Confluence does is charge for it. Checked on 21 August 2026, Confluence Cloud is free for up to ten users with unlimited pages and two gigabytes of storage, with Standard at roughly five dollars forty-two and Premium at roughly ten dollars forty-four per user per month at small team sizes, and Enterprise quoted on request. Atlassian's own pricing page truncated when read, so those figures were cross-checked against multiple published 2026 pricing summaries rather than taken from a single source. The question this page answers is not which wiki has nicer page layouts. It is whether the upkeep can be given to something that will actually do it. ## Head to head Confluence Cloud figures at small team sizes, checked 21 August 2026. | | Confluence | Polaris | | --- | --- | --- | | Cost | Free to 10 users, then roughly $5.42 to $10.44 per user per month | $0, unlimited humans and docs | | Governance | Space permissions, page restrictions, audit logs, data residency | Org-wide access with server-side credential storage. No space permission model | | Page structure | Spaces, page trees, templates, macros, labels | Nested doc tree, block editor with markdown shortcuts, to-dos and sub-pages | | Versioning | Full page history with comparison | Versioned files with file review and comments | | Jira link | Deep, native, two-way | GitHub, Linear, Notion, Slack and ten more as connections a worker can use | | Tasks | Sold separately as Jira | Included: workstreams, lanes, and a pinned Focus lane over three time horizons | | Who keeps docs current | Whoever remembers | An AI worker you hired for it, assigned the task like a colleague | | What upkeep costs | Human hours you already paid a salary for | ~$2 per human-equivalent hour delivered, logged and challengeable | ## When you should stay on Confluence The governance case here is strong, and it is not a formality. - **Permissions are a compliance requirement** — Space permissions, page restrictions and inherited access controls exist because organisations must prove who could read what. Polaris does not have a space permission model, and no pricing argument covers that gap. - **Audit logs, data residency and admin controls are contractual** — Atlassian's admin surface, audit trail and residency options are why Confluence survives procurement in regulated industries. If your security review touches those, this is not a decision you get to make on price. - **Confluence and Jira are one system in practice** — Requirements pages linked to epics, release notes generated from issues, and decision records referenced from tickets form a loop that is genuinely useful. Breaking half of it is worse than paying for both. - **Macros and the marketplace are doing real work** — Teams depending on specific macros, diagramming apps or documentation add-ons are buying that catalogue as much as the wiki itself. ## Two ways to keep documentation true **The Confluence way** - Assign a documentation owner in the team meeting - Add a quarterly review reminder nobody actions - Discover the runbook is wrong during an incident - Pay per user per month for the pages either way **The Polaris way** - Hire a technical writer worker in about sixty seconds, in chat - Its capabilities become a SKILL.md file you can read and edit - Assign it the review task; a cloud machine wakes and does it - The updated doc arrives as a comment with files, and a human closes it ## Moving from Confluence Confluence exports spaces to HTML, XML or PDF, and individual pages to Word. The text and the page hierarchy survive. Macros, page restrictions, inherited permissions and anything a marketplace app rendered do not survive, because they are not content, they are behaviour. The pattern that works is selective. Move the documentation people actually read, which is usually a small fraction of the space, into a Polaris doc tree, and leave the archive in Confluence at the free tier if you are under ten users or on a reduced plan if you are not. Then assign the upkeep to a worker and see whether the docs stay true for a quarter. ## Go deeper - [cost/confluence-pricing](https://www.polarishq.co/cost/confluence-pricing) - [compare/notion-vs-confluence](https://www.polarishq.co/compare/notion-vs-confluence) - [replace/confluence-and-jira](https://www.polarishq.co/replace/confluence-and-jira) - [replace/confluence-and-notion](https://www.polarishq.co/replace/confluence-and-notion) - [ai-workers/technical-writer](https://www.polarishq.co/ai-workers/technical-writer) - [use-cases/engineering/technical-documentation](https://www.polarishq.co/use-cases/engineering/technical-documentation) ## Questions people ask **Can Polaris replace Confluence for a regulated company?** Probably not. Polaris has no space permission model, no page restrictions and no data residency options, and those are usually hard requirements in regulated environments rather than preferences. Confluence exists in those companies because it clears procurement, and clearing procurement is a feature. **How do I export Confluence content?** Confluence exports spaces as HTML, XML or PDF and individual pages as Word documents. Page text and hierarchy come through cleanly. Macros, restrictions, inherited permissions and marketplace app output do not, because those are behaviours rather than content. **What does an AI worker actually do to a document?** It is assigned a task like any teammate, then a cloud machine wakes up, compiles its instructions and skill files with the task context, runs a live tool loop that can include web search and connected tools, ticks the acceptance criteria you wrote, and posts the revised document as a comment with any files attached. A human reviews and closes it. **What does Confluence cost for twenty-five people?** At roughly five dollars forty-two per user per month on Standard, twenty-five users is around one thousand six hundred and twenty-six dollars a year, and Premium at roughly ten dollars forty-four is around three thousand one hundred and thirty-two, checked 21 August 2026. Atlassian rates slide with user count, so confirm at your seat band before budgeting. **Does Polaris keep document history?** Yes. Docs are versioned, with file review and comments, and the doc tree supports sub-pages, to-dos and markdown shortcuts in the block editor. It does not offer Confluence-style side-by-side version comparison across arbitrary revisions. ## Related - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/replace/confluence-and-jira - https://www.polarishq.co/replace/confluence-and-notion - https://www.polarishq.co/replace/jira-and-confluence-and-slack - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/jira --- --- title: "Todoist alternative when a personal list stops scaling" description: "Todoist Free caps at 5 projects, Pro is $5 and Business $8 per user monthly on annual billing, checked August 2026. Polaris is a free team workspace." url: https://www.polarishq.co/alternatives/todoist section: Alternatives updated: 2026-08-21 --- # A Todoist alternative for the point where a list becomes a team Todoist is a superb personal task manager. Most people searching this are not looking for a better one. ## The short answer Polaris is a free alternative to Todoist for people whose personal list has become a team's work. Todoist lists Free with five personal projects, Pro at five dollars per month billed annually or seven monthly, and Business at eight dollars per user per month annually or ten monthly, checked 21 August 2026. Polaris adds workstreams, docs, team chat and AI workers who deliver tasks. - **Todoist Free:** 5 personal projects - **Todoist Pro:** $5 / month annual, $7 monthly - **Todoist Business:** $8 / user / month annual - **Checked:** 21 Aug 2026 ## The honest version of this comparison Todoist costs about the price of a coffee. Checked on 21 August 2026, Pro is five dollars a month billed annually or seven billed monthly, and Business is eight dollars per user per month annually or ten monthly, with the Free tier capped at five personal projects. Published reporting notes both rose in December 2025. Nobody leaves Todoist over five dollars. People leave when the list stops describing their situation: when three other people need to see it, when tasks need a document attached, when the same six items get carried forward every Monday because the person who owns them has no capacity. So this page is not arguing that Polaris is a cheaper Todoist. It is arguing that the search behind it is usually about scale and delivery rather than price, and the answer changes depending on which one you meant. ## Head to head Todoist prices checked 21 August 2026. The vendor pricing page omitted its figures when read, so these were cross-checked against multiple published 2026 summaries. | | Todoist | Polaris | | --- | --- | --- | | Cost | Free with 5 projects, Pro $5/mo annual, Business $8/user/mo annual | $0, unlimited humans, tasks, workstreams and docs | | Capture speed | Natural language quick-add that is genuinely best in class | Chat capture plus an Inbox of prefilled task suggestions from connected tools | | Offline | Full offline support across desktop and mobile | Web app plus an installable mobile PWA. No offline-first guarantee | | Team model | Shared projects, comments, Business team workspace | Workstreams for any container of work, shared lanes across list and board | | Docs | Task comments and file attachments | Nested doc tree with block editor, versioning, file review and comments | | Recurrence | Rich recurring dates in plain language | Tasks with owners and dates. No natural-language recurrence engine | | Who does the task | You | You, a teammate, or an AI worker on a cloud machine that returns files | | Voice | Voice capture through mobile assistants | Mobile app with voice input and spoken replies, copilot at the centre | ## When you should stay on Todoist This is the page in this cluster where staying is most often the right answer. - **You are one person and the system works** — A personal task manager that you have used for years, with your own labels and filters, is a habit rather than a subscription. Five dollars a month against a working habit is not a decision worth making. - **Quick-add is why you use it** — Typing a task with a date, project, priority and label in one line of plain English is Todoist's signature and nothing here matches it. Losing that would slow down the part you do fifty times a day. - **You need it offline, on a plane, on a phone with no signal** — Todoist works offline properly across platforms. Polaris is a web app and an installable mobile PWA, and that is a different reliability profile. - **Recurring personal routines are the bulk of your list** — Every second Tuesday, every last working day of the month, every three days starting Friday. That recurrence engine is deep, and Polaris does not have one. > **Do not switch for the wrong reason** > > If the problem is that your list is long, a different app will not fix it. The case for Polaris starts when the problem is that items sit there because no human has capacity, because that is what assigning a task to an AI worker on a cloud machine actually addresses. ## If you do move Todoist exports projects to CSV covering content, description, priority, due date and section, and it has a well-documented API. Tasks and sections map onto Polaris tasks and lanes without much friction, and a project becomes a workstream. Filters, labels used as a personal system, karma and recurring date rules do not transfer. Rebuild the three filters you actually use and drop the rest, which is the same advice anyone would give about moving a personal system anywhere. ## Go deeper - [compare/todoist-vs-asana](https://www.polarishq.co/compare/todoist-vs-asana) - [for/solo-founders](https://www.polarishq.co/for/solo-founders) - [for/small-business](https://www.polarishq.co/for/small-business) - [glossary/focus-lane](https://www.polarishq.co/glossary/focus-lane) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [ai-workers/executive-assistant](https://www.polarishq.co/ai-workers/executive-assistant) ## Questions people ask **Is Polaris a good personal task manager?** It is built for teams rather than for one person, and the Focus lane across today, this week and the next thirty days is the closest thing to a personal system in it. Someone happy with Todoist's quick-add, filters and offline support would be trading a sharp personal tool for a broader team one. **How much does Todoist cost per year?** Checked on 21 August 2026, Pro is sixty dollars a year, equivalent to five dollars a month, or seven a month if billed monthly. Business is ninety-six dollars per user per year, equivalent to eight dollars per user per month, or ten monthly. Published reporting notes both increased in December 2025. **Can Polaris import my Todoist projects?** There is no one-click importer. Todoist exports projects to CSV with content, description, priority, due date and section, and those map onto Polaris tasks and lanes. Filters, labels used as a personal method, and recurring date rules do not come across. **What would an AI worker do with items from a personal list?** The useful ones are the tasks you keep deferring because they need an hour of research, a first draft or a set of comparisons. Assign one, a cloud machine wakes for it, and the result arrives as a comment with any files produced. Billing is roughly two dollars per human-equivalent hour, so a deferred hour costs about two dollars to clear. **Does Polaris work on a phone?** Yes. There is a Flutter mobile app targeting iOS and Android, live today as an installable PWA, with the copilot at the centre of the interface plus voice input and spoken replies. It does not offer Todoist's offline guarantees. ## Related - https://www.polarishq.co/compare/todoist-vs-asana - https://www.polarishq.co/for/solo-founders - https://www.polarishq.co/for/small-business - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/alternatives/trello --- --- title: "Airtable alternative when every editor is a billed seat" description: "Airtable Team is $20 per seat monthly and Business $45, with 1,000 records per base free, checked August 2026. Polaris is free and bills delivered work." url: https://www.polarishq.co/alternatives/airtable section: Alternatives updated: 2026-08-21 --- # An Airtable alternative for teams billed per editor Airtable is a real database with a spreadsheet face. That is the strength, and it is why this comparison has limits. ## The short answer Polaris is a free alternative to Airtable for teams using a base as a project tracker rather than as a database. Airtable lists Free with one thousand records per base and up to five editors, Team at twenty dollars per seat per month billed annually, and Business at forty-five, checked 21 August 2026. Polaris has no seat count and adds AI workers who deliver tasks. - **Airtable Free:** 1,000 records per base - **Airtable Team:** $20 / seat / month annual - **Airtable Business:** $45 / seat / month annual - **Checked:** 21 Aug 2026 ## Two different Airtable users are reading this The first built a relational model: linked records across tables, lookups, rollups, an interface for a team that never sees the underlying grid, and syncs pulling data in from elsewhere. That person should stop reading. Polaris has no relational modelling layer and this comparison does not apply to them. The second uses a base as a project tracker with a status column, an owner column and a date, and is being billed per editor for it. Checked on 21 August 2026, Airtable lists Free with one thousand records per base, up to five editors, one gigabyte of attachments and two weeks of revision history; Team at twenty dollars per seat per month billed annually or twenty-four monthly; Business at forty-five per seat annually; and Enterprise Scale on request. Everyone with edit access is billed. Viewers and form submitters are not. For that second reader, twenty dollars a seat for a task list with a nice grid view is a defensible thing to reconsider. ## Head to head Airtable prices from airtable.com/pricing plus published 2026 summaries for the free-tier limits, checked 21 August 2026. | | Airtable | Polaris | | --- | --- | --- | | Cost | Free to 1,000 records per base, Team $20/seat/mo, Business $45/seat/mo annually | $0, unlimited humans, tasks, workstreams and docs | | Who counts as billable | Every editor on any base | Nobody. There is no seat concept | | Data model | Real relational tables with links, lookups, rollups and field types | Tasks in lanes inside workstreams. No user-defined schema | | Interfaces | Interface Designer for building views for non-editors | List and board views sharing lanes, plus a pinned Focus lane | | Automations | Scripting, triggers and integrations inside the base | AI workers you brief and assign, each with an editable SKILL.md | | Docs | Long text fields and attachments | Nested doc tree, block editor, versioning, file review, comments | | Chat | Record comments | Team chat in the product, plus a Slack-fed Inbox of task suggestions | | Who acts on a row | A person who opens the base | A person, or an AI worker on a cloud machine that returns files as a comment | ## When you should stay on Airtable Airtable does something structurally different from everything else in this cluster. - **You need linked records and rollups** — Modelling inventory against suppliers, campaigns against assets, or clients against deliverables requires relations, and relations are what Airtable is. Polaris has no equivalent and will not gain one by shipping a feature. - **Interface Designer is how non-editors use your data** — Interfaces let people work with a base without touching the grid, which is often the difference between a system being adopted and being ignored. There is nothing like it in Polaris. - **Sync is pulling data in from other systems** — Airtable sync keeps a base current from external sources so the base is the shared view of truth. Losing that means rebuilding a data pipeline, not switching a tool. - **Scripting and the extension catalogue run your process** — Teams with scripts, automations and extensions doing daily work have built software. Treat replacing it as a software project with a real estimate. ## The narrow case where switching makes sense It comes down to one question: is your base a database or a to-do list wearing a grid? If you removed the links, lookups and rollups tomorrow and nothing broke, you are paying database prices for task tracking, and every editor on the team is a line on that invoice. Polaris covers that second case at zero, adds docs and team chat, and lets you assign the rows nobody has time for to an AI worker instead. What it will never do is hold your relational model, and any page telling you otherwise would be easy to falsify in about four minutes. ## Testing it without committing A one-week check that does not require moving anything. 1. **Pick the base that is really a tracker** — Choose the one with a status column, an owner and a date, and no links to other tables. That is the candidate. Leave every relational base exactly where it is. 2. **Export it and rebuild it as one workstream** — Airtable exports each table to CSV. Rows become tasks, the status column becomes lanes, owner and date carry over. Attachments export as links, so download anything that matters. 3. **Hire the worker for the rows that never move** — Describe the stuck work in chat. The Chief of Staff runs a short interview where every answer is one click, writes a SKILL.md you can edit, and the worker card lands ready for its first task in about a minute. 4. **Count the seats you would stop paying for** — At twenty dollars per seat per month on Team, five editors moving off one base is twelve hundred dollars a year. Compare that against what the workers actually delivered, which is itemised on their work logs. ## Go deeper - [compare/notion-vs-airtable](https://www.polarishq.co/compare/notion-vs-airtable) - [replace/airtable-and-slack](https://www.polarishq.co/replace/airtable-and-slack) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [use-cases/operations/process-documentation](https://www.polarishq.co/use-cases/operations/process-documentation) - [ai-workers/ops-coordinator](https://www.polarishq.co/ai-workers/ops-coordinator) ## Questions people ask **Can Polaris do what Airtable does?** No, and this is the clearest limit in the whole alternatives cluster. Airtable is a relational database with linked records, lookups, rollups and interfaces. Polaris has tasks in lanes inside workstreams with no user-defined schema. If your base depends on relations, Polaris is not a replacement for it. **Who does Airtable actually bill?** Every person with edit access to any base is a billable seat, checked 21 August 2026, at twenty dollars per seat per month on Team billed annually or twenty-four billed monthly, and forty-five on Business. Viewers and form submitters are free. That is why the bill grows faster than the number of people who feel like Airtable users. **What are the free plan limits?** Airtable's Free plan allows unlimited bases with one thousand records per base, up to five editors, one gigabyte of attachments and two weeks of revision history, checked 21 August 2026. The record ceiling is what forces most upgrades, rather than the seat count. **How would an AI worker help with data work?** The realistic tasks are research and enrichment rather than schema work: checking a list of companies, drafting the summaries a row needs, or producing a comparison document. The worker runs on a cloud machine with live web search, ticks the acceptance criteria, and posts the result as a comment with files attached. **Is there really no seat charge in Polaris?** None. The software is free with unlimited humans, tasks, workstreams and docs, and the Chief of Staff is included in every organisation. Revenue comes only from delivered agent work at roughly two dollars per human-equivalent hour, estimated by an open formula and logged job by job. ## Related - https://www.polarishq.co/compare/notion-vs-airtable - https://www.polarishq.co/replace/airtable-and-slack - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/coda - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/ai-workers/ops-coordinator - https://www.polarishq.co/use-cases/data/analysis-backlog --- --- title: "Height alternative: autonomous tracker vs delivered work" description: "Height automates the tracker itself. Polaris assigns tasks to AI workers on cloud machines that deliver files. Compared honestly, checked August 2026." url: https://www.polarishq.co/alternatives/height section: Alternatives updated: 2026-08-21 --- # A Height alternative, and the difference between two kinds of autonomy Two products both say autonomous. They mean different things, and the difference is the whole decision. ## The short answer Height and Polaris both apply AI to project management but at different points. Height applies autonomy to the tracker itself, keeping the backlog groomed, deduplicated and updated. Polaris applies it to the work, assigning tasks to AI workers that run on cloud machines and return deliverables as comments with files. Polaris software is free and bills roughly two dollars per human-equivalent hour delivered. - **Height autonomy:** maintains the tracker - **Polaris autonomy:** delivers the task - **Polaris software:** $0, no seats - **Checked:** 21 Aug 2026 ## The word autonomous is doing two jobs Height positions itself as an autonomous project management tool, and the autonomy is aimed at the overhead around tasks: grooming the backlog, spotting duplicates, updating statuses, keeping specs and threads tidy so a human spends less of the week administering the tracker. That is a real problem and a defensible thing to automate. Polaris aims autonomy somewhere else entirely. AI workers sit in the same members table as humans, with kind set to agent instead of human, and are assigned tasks the same way. When a task lands on one, a cloud machine wakes, compiles the worker's instructions and its skill files with the task context, runs a live tool loop including real web search, ticks its own acceptance criteria, and posts the result as a comment with any files it produced. A human closes it. One removes the admin around the work. The other does part of the work. A team choosing between them should decide which of those two costs them more hours per week, because that answer picks the tool. ## Head to head Height's pricing page did not respond across three attempts on 21 August 2026, so pricing below is described qualitatively rather than quoted. Confirm current figures at height.app before deciding. | | Height | Polaris | | --- | --- | --- | | Pricing model | Free tier for small teams plus per-member paid tiers, published on height.app | $0 software with no seat count, at any team size | | What AI does | Maintains the tracker: grooming, deduplication, status upkeep, spec tidying | Delivers the task: research, drafting, file production, checked against acceptance criteria | | Where AI runs | Inside the product, over your workspace data | On a cloud machine that wakes per task and keeps running after you close the laptop | | Assignment model | Tasks are assigned to people, assisted by AI | Humans and AI workers share one members table and are assigned identically | | What arrives | A cleaner backlog | A comment on the task with the deliverable and any generated files | | Configurability of the AI | Product behaviour, tuned by the vendor | An editable SKILL.md per worker that you can read and change | | Docs and chat | Task specs and threads | Nested doc tree with versioning and review, plus team chat and a Slack-fed Inbox | | Billing unit | Per member per month | ~$2 per human-equivalent hour delivered, itemised on a work log you can challenge | ## When you should stay on Height Two of these are strong enough that a Polaris page ought to say them out loud. - **Your pain is genuinely the tracker, not the work** — Teams drowning in duplicate issues, stale statuses and backlog grooming meetings have a problem Height is aimed directly at. Polaris does not automate backlog hygiene. - **You want AI that stays inside the tool** — Height's autonomy operates on your workspace without a machine going off to run tool loops elsewhere. For teams with a low tolerance for agents acting on external systems, that is a smaller surface to reason about. - **Predictable per-member billing is easier to approve** — A fixed per-member price is trivially budgetable. Usage-metered billing, even at two dollars an hour with an open formula, requires explaining a variable line to whoever signs off. - **You need a mature, shipped product with customers** — Height has been in market for years. Polaris is in free public beta with no customers and no case studies, and for some buyers that alone settles it. ## Same Tuesday, two products **With an autonomous tracker** - Duplicate issues merged before standup - Stale statuses updated without a nudge - The backlog is accurate and readable - Every task in it is still waiting for a person **With AI workers** - Three tasks assigned to workers instead of people - A cloud machine wakes for each and posts progress - Deliverables arrive as comments with files attached - You review, rate and close; the machine never marks work done ## A note on the pricing gap on this page Every other page in this cluster quotes a competitor's list price from their own pricing page, dated 21 August 2026. Height's pricing page did not load across three attempts on that date. Third-party summaries published in 2026 disagree with each other on the tier names and rates, so quoting them here would put a number on the page that a reader could disprove in one click. The tier structure is a free plan for small teams plus per-member paid tiers with higher limits. For the current figures, read height.app directly. This page will be updated when the source is readable. ## Go deeper - [compare/linear-vs-height](https://www.polarishq.co/compare/linear-vs-height) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [glossary/agentic-project-management](https://www.polarishq.co/glossary/agentic-project-management) - [cloud-claude-code/cloud-agents-vs-local-agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) - [alternatives/linear](https://www.polarishq.co/alternatives/linear) ## Questions people ask **What is the actual difference between Height and Polaris?** Height applies AI to the tracker, keeping the backlog groomed, deduplicated and current. Polaris applies AI to the work, assigning tasks to workers that run on cloud machines and return deliverables as comments with files. If your weekly cost is administering the board, Height is aimed at you. If it is tasks nobody has time to start, Polaris is. **Why does this page not quote Height's prices?** Height's pricing page did not respond across three attempts on 21 August 2026, and the third-party summaries available disagree on tier names and rates. Publishing a number nobody could verify would be worse than publishing none, so the tier structure is described instead. Check height.app for current figures. **Can a Polaris AI worker be edited or corrected?** Yes. A worker's capabilities are stored as a real SKILL.md file you can open, read and change, rather than a hidden prompt. Its tool access comes from a fixed catalog of fourteen connections, authorised once and stored server-side. Changing the skill file changes what the worker does on its next task. **How do I know an AI worker is not inflating its hours?** Every job writes to the worker's work log, and the human-equivalent hour is computed by an open formula from observable effort such as searches run, prose produced and files returned, clamped between five minutes and eight hours per session. You can challenge any line on the bill from the work log itself. **Is Polaris mature enough to compare against a shipped product?** It is a working product in free public beta with no customers, no revenue and no case studies, and the recorded demos on the site include an unstaged machine session delivering a task. Anyone comparing on track record should weigh that honestly rather than take a marketing page's word for it. ## Related - https://www.polarishq.co/compare/linear-vs-height - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/glossary/work-log --- --- title: "Shortcut alternative for software teams past five seats" description: "Shortcut is free to 5 users, then $8.50 per user monthly on Team and $12 on Business, checked August 2026. Polaris is free software with AI workers." url: https://www.polarishq.co/alternatives/shortcut section: Alternatives updated: 2026-08-21 --- # A Shortcut alternative for teams who just crossed the free seat limit Shortcut sits between Jira's configuration and Linear's opinion. The sixth engineer is where the invoice starts. ## The short answer Polaris is a free alternative to Shortcut for software teams that crossed its five-seat free limit. Shortcut lists Free for up to five users, Team at eight dollars fifty and Business at twelve dollars per user per month, checked 21 August 2026. Polaris has no seat count, includes docs and team chat, and lets tasks be assigned to AI workers that deliver on cloud machines. - **Shortcut Free:** up to 5 users - **Shortcut Team:** $8.50 / user / month - **Shortcut Business:** $12 / user / month - **Checked:** 21 Aug 2026 ## The sixth-engineer problem Shortcut is a well-judged product for software teams: stories, epics and iterations that map onto how a small engineering group actually plans, docs alongside the tracker, and a git integration that does the obvious thing. Checked on 21 August 2026, the pricing page lists Free for up to five users, Team at eight dollars fifty per user per month, Business at twelve, and Enterprise on request, with yearly billing advertised as saving up to twenty-five percent. Eight dollars fifty is not an outrageous number. The problem is what it is charging for. A tracker is a shared agreement about what is not finished, and the sixth engineer does not make the backlog move faster. They make it longer, and now they also cost eight dollars fifty a month. The teams who come looking are usually the ones where engineering capacity is the constraint and the backlog contains work that is not engineering: writing the release notes, checking what a competitor shipped, updating the docs that went stale in March. ## Head to head Shortcut prices from shortcut.com/pricing, checked 21 August 2026. | | Shortcut | Polaris | | --- | --- | --- | | Cost | Free to 5 users, then $8.50 Team or $12 Business per user per month | $0 at any team size, no seat count | | Planning model | Stories, epics, iterations and milestones | Workstreams with shared lanes, plus a Focus lane across three time horizons | | Docs | Shortcut Docs alongside the tracker | Nested doc tree, block editor, versioning, file review, comments | | Git integration | Branch and pull request linking with automatic state changes | GitHub as a connection an AI worker can be given, not a state-change link | | Team chat | Not included | Included, plus an Inbox that turns Slack messages into prefilled task suggestions | | Reporting | Burndown, cycle time and throughput reports | Worker activity feed, insights, and a per-job work log | | Assignees | Humans | Humans and AI workers in the same members table | | When work is finished | An engineer moves the story | A worker posts a comment with files and ticks acceptance criteria; a human closes it | ## When you should stay on Shortcut Shortcut is genuinely good at the job it chose, and these are real reasons to keep it. - **The story, epic and iteration model fits your team** — It is a middle ground that suits engineering groups who found Jira too configurable and Linear too opinionated. Polaris has lanes and time horizons, not iterations, and does not produce cycle time or throughput reports. - **Branch and pull request linking is part of the workflow** — Stories that change state when a branch merges remove a class of manual updates. Polaris connects to GitHub for AI worker access, which is a different thing entirely. - **You are five people or fewer** — Shortcut Free covers up to five users, so there is no bill to escape yet and no argument on this page that applies to you. - **Engineering reporting is how you run planning** — Burndown, cycle time and throughput are the numbers a lot of teams plan against. Polaris does not produce them and does not have an equivalent. ## What crossing five seats costs Shortcut list prices checked 21 August 2026. - **$1,020** — Ten engineers on Shortcut Team, per year. $8.50 per user per month, before annual discount - **$1,440** — Ten engineers on Shortcut Business, per year. $12 per user per month - **$0** — Polaris software, ten engineers or fifty. Billing begins only when a worker delivers something ## Moving from Shortcut, or running both Shortcut has a documented REST API and CSV export covering stories, epics, states, owners, estimates and labels, so a scripted move of the task data is straightforward. Iterations, git links, burndown history and workflow states do not have a destination in Polaris. The pattern that costs nothing to test is narrower than a migration. Keep Shortcut for engineering. Open one Polaris workstream for the non-engineering work that keeps landing in the engineering backlog, connect GitHub so a worker can see the repository context, and assign three of those tasks to AI workers. Judge it on whether that backlog stops growing. ## Go deeper - [compare/linear-vs-shortcut](https://www.polarishq.co/compare/linear-vs-shortcut) - [alternatives/linear](https://www.polarishq.co/alternatives/linear) - [alternatives/jira](https://www.polarishq.co/alternatives/jira) - [integrations/github](https://www.polarishq.co/integrations/github) - [for/software-teams](https://www.polarishq.co/for/software-teams) - [use-cases/engineering/technical-documentation](https://www.polarishq.co/use-cases/engineering/technical-documentation) ## Questions people ask **How much is Shortcut per user?** Checked on 21 August 2026, Shortcut lists Free for up to five users, Team at eight dollars fifty per user per month, Business at twelve, and Enterprise on request, with yearly billing advertised as saving up to twenty-five percent. Ten engineers on Team is roughly one thousand and twenty dollars a year at monthly rates. **Does Polaris have iterations or sprints?** No. Polaris organises work into workstreams with lanes shared between list and board views, plus a Focus lane pinned across today, this week and the next thirty days. Teams that plan in fixed-length iterations with velocity tracking would be losing their planning unit. **Can I move Shortcut stories into Polaris?** Shortcut exposes a documented REST API and CSV export with stories, epics, states, owners, estimates and labels, which covers the task data. Iterations, git links, workflow states and burndown history do not transfer, so treat them as retired rather than migrated. **What kind of engineering work can an AI worker actually take?** The realistic set is the work around the code rather than the code review itself: release notes, documentation updates, competitor checks, research write-ups, comparison documents. The worker runs on a cloud machine with live web search and connected tools, and delivers as a comment with files for a human to review and close. **Is there a seat charge in Polaris at any team size?** No. The software is free with unlimited humans, tasks, workstreams and docs, and there are no tiers. The only charge is roughly two dollars per human-equivalent hour that a worker delivers, estimated by an open formula and written line by line to that worker's work log. ## Related - https://www.polarishq.co/compare/linear-vs-shortcut - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/integrations/github - https://www.polarishq.co/for/software-teams - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/use-cases/engineering/sprint-planning --- --- title: "Wrike alternative without seat bands and blocks of five" description: "Wrike Team is $10 per user monthly for 2 to 15 users and Business $25, with seats sold in groups, checked August 2026. Polaris is free with no seat count." url: https://www.polarishq.co/alternatives/wrike section: Alternatives updated: 2026-08-21 --- # A Wrike alternative for teams buying seats in blocks Seat bands, minimum user counts and subscriptions sold in blocks of five. The pricing page is the product review. ## The short answer Polaris is a free alternative to Wrike for teams tired of seat bands and block purchasing. Wrike lists Free at zero, Team at ten dollars per user per month for two to fifteen users, and Business at twenty-five dollars per user per month for five to two hundred, with seats sold in groups, checked 21 August 2026. Polaris has no seat count and bills only delivered agent work. - **Wrike Team:** $10 / user / month, 2 to 15 users - **Wrike Business:** $25 / user / month, 5 to 200 - **Polaris software:** $0, no bands, no minimums - **Checked:** 21 Aug 2026 ## Read the seat rules before the feature list Checked on 21 August 2026, Wrike lists Free at zero per user per month, Team at ten dollars per user per month for two to fifteen users, Business at twenty-five dollars per user per month for five to two hundred users, and Pinnacle and Apex on request. Pricing is quoted per month and billed annually per user. The detail that shapes a budget is further down the page: up to thirty seats, subscriptions are sold in groups of five; from thirty to one hundred, in groups of ten; above one hundred, in groups of twenty-five. Hiring one person can therefore mean buying five seats, and the sixteenth user pushes a Team account into a different plan entirely. None of that makes Wrike a bad product. It makes the pricing a procurement exercise, and it is why teams whose headcount moves in ones and twos start looking at something without bands. ## Head to head Wrike prices from wrike.com/price, checked 21 August 2026. | | Wrike | Polaris | | --- | --- | --- | | Cost | $10/user/mo Team for 2 to 15 users, $25/user/mo Business for 5 to 200 | $0 at any team size | | Seat rules | Sold in groups of 5, 10 or 25 depending on account size | No seats. Unlimited humans | | Request intake | Custom request forms that route work into the right project | Inbox suggestions from connected tools. No form builder | | Creative review | Proofing and approvals on assets, including video | File review and comments on versioned files. No visual annotation on assets | | Resource management | Workload charts, effort allocation, time tracking | Not available. Polaris has no capacity or time tracking surface | | Docs | Task descriptions and attached files | Nested doc tree with block editor, versioning, review and comments | | Team chat | Task comments and @mentions | Team chat in the product, plus a Slack-fed Inbox of prefilled task suggestions | | Who produces the deliverable | An agency or in-house team member | A person, or an AI worker on a cloud machine that returns generated files | ## When you should stay on Wrike Wrike is built for marketing operations and professional services, and it shows in the places Polaris is empty. - **Proofing and approvals are how creative gets signed off** — Annotating a video frame or a layout, routing it through approvers and keeping the version trail is a specific workflow Wrike does well. Polaris has file review and comments, which is not the same thing. - **Request forms are your intake** — Custom forms that route an incoming request into the right project with the right fields are often how a marketing team stops taking briefs over chat. There is no form builder in Polaris. - **You plan capacity and bill time** — Workload views, effort allocation and time tracking underpin professional services billing. Polaris has none of them and cannot substitute for them. - **Enterprise reporting and admin cleared your security review** — Wrike is bought by organisations with procurement processes, and passing one is expensive. That work is already done and does not transfer. ## The narrower question worth asking Count how many of your Wrike seats belong to people who open it once a week to update a status. Under the group purchasing rules, those users cost the same as the ones running the department. On Business at twenty-five dollars per user per month, ten of them is three thousand dollars a year for a status field. That is the gap Polaris fits into, and only that gap. The software is free at any headcount and the only bill is roughly two dollars per human-equivalent hour a worker delivers, itemised on a work log you can challenge line by line. If the creative approval workflow is the reason Wrike exists in your company, none of this applies and you should keep it. > **The philosophical difference** > > Wrike sells you seats in blocks of five so that work has a place to go. Polaris gives the workspace away and charges when work comes back finished, with the hour estimated by a formula you can read and dispute. ## Go deeper - [compare/wrike-vs-asana](https://www.polarishq.co/compare/wrike-vs-asana) - [alternatives/asana](https://www.polarishq.co/alternatives/asana) - [alternatives/monday](https://www.polarishq.co/alternatives/monday) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [for/agencies](https://www.polarishq.co/for/agencies) - [use-cases/marketing/campaign-planning](https://www.polarishq.co/use-cases/marketing/campaign-planning) ## Questions people ask **Why does Wrike sell seats in groups?** Wrike's pricing page states that for accounts up to thirty seats, subscriptions are sold in groups of five; from thirty to one hundred, in groups of ten; and above one hundred, in groups of twenty-five, checked 21 August 2026. The practical effect is that adding one person can mean buying a block. **What does Wrike cost for a team of fifteen?** On Team at ten dollars per user per month, fifteen users is one thousand eight hundred dollars a year, checked 21 August 2026. Team covers two to fifteen users, so the sixteenth person moves the account to Business at twenty-five dollars per user per month, which is four thousand eight hundred a year for the same sixteen people. **Does Polaris have proofing or approval workflows?** No. Polaris has versioned files with file review and comments, which covers document review but not visual annotation on creative assets or multi-stage approval routing. Teams whose sign-off process depends on proofing should keep Wrike for that work. **Can I export my Wrike data?** Wrike offers Excel and CSV export of tasks and folders and has a documented API. Tasks, owners, dates and folder structure map onto Polaris tasks, workstreams and lanes. Request forms, proofing history, workload allocations and time entries do not transfer. **How is a two dollar hour calculated?** It is a human-equivalent hour estimated from observable effort: a base pickup time, searches at roughly twelve minutes each, finished prose at around ninety characters a minute, and fixed overheads for checklist items, comments addressed and files produced, clamped between five minutes and eight hours per session. Every job writes the calculation to the worker's work log. ## Related - https://www.polarishq.co/compare/wrike-vs-asana - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/for/agencies - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/use-cases/marketing/campaign-planning --- --- title: "Coda alternative now that Coda is Superhuman Docs" description: "Coda became Superhuman Docs in July 2026 and now sells inside a suite from $12 per member monthly, checked August 2026. Polaris is a free workspace." url: https://www.polarishq.co/alternatives/coda section: Alternatives updated: 2026-08-21 --- # A Coda alternative for teams caught by the Superhuman rebrand The product you bought is now part of somebody else's bundle. That is a reasonable moment to look around. ## The short answer Polaris is a free alternative to Coda, which became Superhuman Docs in July 2026 and is now sold inside the Superhuman suite alongside Grammarly, Mail and Go. Superhuman lists Free at zero, Pro at twelve dollars and Business at thirty-three dollars per member per month billed annually, checked 21 August 2026. Polaris software is free with no seat count. - **Coda became:** Superhuman Docs, July 2026 - **Superhuman Pro:** $12 / member / month annual - **Superhuman Business:** $33 / member / month annual - **Checked:** 21 Aug 2026 ## What changed, and why it matters more than a name Coda is now Superhuman Docs. Published documentation from both Coda and Superhuman dates the change to 8 July 2026, placing the product inside the Superhuman suite alongside Grammarly, Superhuman Mail and Superhuman Go, following Grammarly's own rebrand to Superhuman. The pricing shape changed with it. Checked on 21 August 2026, superhuman.com lists Free at zero including Grammarly, Docs and Go; Pro at twelve dollars per member per month billed annually or fifteen billed monthly; Business at thirty-three annually or forty monthly, adding Mail; and Enterprise on request. Coda's own historical model billed only Doc Makers, at published rates around ten dollars per Doc Maker per month annually on Pro and thirty on Team, with editors and viewers free. That is the real story for anyone searching this. Coda's distinctive billing idea, charging for the few people who build docs rather than the many who read them, has been folded into a per-member suite price covering products you may not have wanted to buy. ## Head to head Superhuman suite prices from superhuman.com/plans, checked 21 August 2026. Coda legacy Doc Maker rates from published 2026 summaries. | | Coda / Superhuman Docs | Polaris | | --- | --- | --- | | Product identity | Coda since 8 July 2026 is Superhuman Docs, inside the Superhuman suite | One product: tasks, docs and team chat, with AI workers | | Cost | Suite Free, Pro $12 and Business $33 per member per month annually | $0 for software at any team size | | Billing unit | Per member, across the bundled products | Per human-equivalent hour of delivered work, roughly $2, itemised | | Legacy model | Doc Maker billing, editors and viewers free | No billing unit tied to people at all | | Document power | Tables in docs, formula language, Packs, buttons and automations | Block editor with markdown shortcuts, to-dos, sub-pages, versioning and review | | Task management | Tables configured as trackers | Workstreams with shared lanes plus a pinned Focus lane over three horizons | | Team chat | Comments in docs | Team chat in the product, plus a Slack-fed Inbox of prefilled task suggestions | | What AI does | Assists inside the doc, bundled with the suite | Runs on a cloud machine as an assigned teammate and returns files as a comment | ## When you should stay on Coda The document engine underneath the new name is still one of the strongest in this category. - **The formula language is doing real work** — Coda formulas, buttons and automations let a non-engineer build something close to an application inside a document. Polaris has none of that, and rebuilding a mature Coda doc would mean writing software. - **Packs connect the doc to your systems** — A Pack pulling live data from another product into a table is infrastructure. Losing it means losing the data flow, not just the layout. - **You already want the Superhuman bundle** — Teams buying Grammarly and Superhuman Mail anyway get Docs inside the same per-member price, which changes the arithmetic considerably in Superhuman's favour. - **Most of your people only read** — If you are still on legacy Doc Maker billing, a small number of builders and a large number of free readers is a genuinely cheap arrangement that per-seat products cannot match. ## Working out whether the rebrand changes your bill Four things to check before deciding anything. 1. **Find out which billing model you are on** — Legacy Coda workspaces were billed per Doc Maker, with editors and viewers free. Superhuman suite plans are billed per member. Those produce very different totals for the same team, so confirm which one your invoice reflects. 2. **Count the products you actually use** — The Business tier at thirty-three dollars per member per month bundles Mail, Grammarly, Docs and Go. If your team uses one of the four, work out the effective price of that one. 3. **Separate the docs from the trackers** — Coda docs built on formulas and Packs should stay. Coda tables being used as a project tracker with a status column are the part worth comparing, and that part is free in Polaris. 4. **Test the delivery question, not the editor question** — Open one Polaris workstream, hire a worker through the chat interview, and assign it a task you have been deferring. What comes back is a comment with files and a work log line at roughly two dollars per human-equivalent hour. ## Moving your documents Coda exports docs to Markdown, HTML, PDF and CSV per table, and has a documented API. Prose and table contents come out. Formulas, buttons, automations, Pack connections and cross-doc references do not, because they are behaviour rather than content. The practical division is the same one that applies to Notion. Reference documentation moves cleanly into a Polaris doc tree. Anything approaching an application should stay where the formula engine lives, and no honest comparison page would suggest otherwise. ## Go deeper - [compare/notion-vs-coda](https://www.polarishq.co/compare/notion-vs-coda) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [alternatives/airtable](https://www.polarishq.co/alternatives/airtable) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) ## Questions people ask **Is Coda still called Coda?** Published documentation from Coda and Superhuman dates the change to 8 July 2026, when Coda became Superhuman Docs within the Superhuman suite alongside Grammarly, Mail and Go. Existing workspaces continued to work through the transition. Check the vendor's help centre for the current state of your own account. **What does Superhuman Docs cost now?** Checked on 21 August 2026, superhuman.com lists Free at zero covering Grammarly, Docs and Go; Pro at twelve dollars per member per month billed annually or fifteen monthly; Business at thirty-three annually or forty monthly, adding Mail; and Enterprise on request. Legacy Coda Doc Maker billing was a separate and different model. **Does Polaris have anything like Coda formulas or Packs?** No. Polaris docs are a nested tree with a block editor, markdown shortcuts, to-dos, sub-pages, versioning, file review and comments. There is no formula language, no button objects and no Pack system. Documents that behave like applications should stay in Coda. **Can I export a Coda doc?** Coda exports to Markdown, HTML, PDF and CSV per table, and has a documented API. Prose and table contents come across. Formulas, buttons, automations, Pack connections and cross-doc references do not transfer to any tool, including Polaris. **Why is Polaris free if everything else here charges per seat?** Because the seat is not the thing worth charging for. The software is free with unlimited humans, tasks, workstreams and docs, and revenue comes from AI workers delivering work at roughly two dollars per human-equivalent hour. Polaris is in free public beta with no customers yet, and that is the model rather than an introductory offer. ## Related - https://www.polarishq.co/compare/notion-vs-coda - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/airtable - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/glossary/per-seat-pricing --- --- title: "Compare work tools: 16 honest head-to-head breakdowns" description: "Sixteen head-to-head comparisons of Jira, Linear, Notion, Asana, monday.com, ClickUp and Slack, with list prices checked in August 2026 and a stated winner." url: https://www.polarishq.co/compare section: Comparisons updated: 2026-08-21 --- # Head-to-head comparisons of the tools teams choose between Sixteen pages that pick a side, name the team each tool suits, and show the arithmetic behind both bills. ## The short answer Sixteen head-to-head comparisons of the work tools teams actually choose between: Jira against Linear, Notion against Confluence, Asana against monday.com, and thirteen more. Each page names a winner for a specific kind of team, quotes both vendors' list prices from a check on 21 August 2026, and says where the losing tool is still the better call. - **Comparisons:** 16 - **Prices checked:** 21 August 2026 - **Pages that duck the verdict:** 0 ## Why most comparison pages are useless The standard vendor comparison ends in a shrug. Both tools are excellent, it depends on your needs, book a demo. That is a page written by someone who did not want to be quoted, and it helps nobody who has two browser tabs open and a decision due on Friday. These pages pick a side. Not because one tool is better in the abstract, but because the reader is a particular kind of team with a particular constraint, and for that team one of the two is clearly the right call. Where the answer flips at a certain team size or a certain kind of work, the page says where the line is. ## How each page is built Same skeleton on all sixteen so you can read them against each other. - **The verdict is the first thing you read** — One paragraph naming which of the two wins and for whom. If you only read the top of the page you still get the answer. - **Prices with a date on them** — List price per seat per month at the tier a normal team lands on, taken from the vendor's own pricing page on 21 August 2026. Where the page would not render, the source is named and the number is labelled as tracked rather than confirmed. - **The differences that survive a trial** — Seat minimums, billing blocks, export formats, what breaks in a migration, and which free tier quietly shrank. Feature-checkbox parity is not what decides these choices. - **A third option, at the end, clearly labelled** — One short section on Polaris with the single argument that applies to someone comparing those two tools. Not a pitch, and not at the top of the page, because you did not come here for it. > **One page changed its own conclusion** > > Height shut down on 24 September 2025, six months after its founder announced it. The Linear against Height page is still worth reading, but it is now a page about where an autonomous project tool's users went, not a page about which one to buy. ## What the price check turned up Three numbers from the August 2026 sweep that changed how several of these pages read. - **2 seats** — Asana's free plan cap. For workspaces created after 12 November 2025. Older accounts keep the legacy allowance. - **3 seats** — monday.com paid minimum. Sold in blocks, so a four-person team buys five seats. - **1,000** — Airtable records per base, free plan. Reduced from 1,200 in February 2026. ## Every comparison - [Jira vs Linear](https://www.polarishq.co/compare/jira-vs-linear) — One tool is opinionated and fast. The other is configurable and permanent. That is the whole decision. - [Notion vs Confluence](https://www.polarishq.co/compare/notion-vs-confluence) — A documentation tool people enjoy against a documentation tool people comply with. - [Asana vs monday.com](https://www.polarishq.co/compare/asana-vs-monday) — One models work as a plan with dependencies. The other models it as a spreadsheet that can run automations. - [Trello vs Asana](https://www.polarishq.co/compare/trello-vs-asana) — For a small team on a free plan this is no longer close. For a team with dependencies it never was. - [Linear vs Shortcut](https://www.polarishq.co/compare/linear-vs-shortcut) — Two trackers built for engineers, disagreeing about how much structure a team should be allowed to have. - [ClickUp vs monday.com](https://www.polarishq.co/compare/clickup-vs-monday) — Two platforms that both claim to replace your whole stack, with very different ideas about who does the configuring. - [Notion vs Coda](https://www.polarishq.co/compare/notion-vs-coda) — The product comparison still holds. The company comparison changed underneath it this summer. - [Slack vs Microsoft Teams](https://www.polarishq.co/compare/slack-vs-microsoft-teams) — This is not really a chat comparison. It is a question about which suite you already pay for. - [Jira vs Asana](https://www.polarishq.co/compare/jira-vs-asana) — Most companies that argue about this end up running both. The useful question is which one owns the boundary. - [Notion vs Airtable](https://www.polarishq.co/compare/notion-vs-airtable) — One is a document tool that grew tables. The other is a database that grew documents. It shows. - [Basecamp vs Asana](https://www.polarishq.co/compare/basecamp-vs-asana) — The cheapest tool here is obvious once you count heads. Whether it can hold your work is the real question. - [ClickUp vs Notion](https://www.polarishq.co/compare/clickup-vs-notion) — Both promise one tool for everything. They disagree about which half of everything comes first. - [Linear vs Height](https://www.polarishq.co/compare/linear-vs-height) — One of these two products no longer exists. The idea behind it is the interesting part. - [Trello vs monday.com](https://www.polarishq.co/compare/trello-vs-monday) — Both put cards in columns. Only one of them expects the columns to do arithmetic. - [Wrike vs Asana](https://www.polarishq.co/compare/wrike-vs-asana) — Two work platforms priced almost identically, aimed at two different departments. - [Todoist vs Asana](https://www.polarishq.co/compare/todoist-vs-asana) — The question is not which is better. It is whether anyone besides the assignee needs to see the shape of the work. ## If you already know which tool you are leaving - [alternatives/jira](https://www.polarishq.co/alternatives/jira) - [alternatives/linear](https://www.polarishq.co/alternatives/linear) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [alternatives/asana](https://www.polarishq.co/alternatives/asana) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) ## Questions people ask **How often are the prices on these pages rechecked?** Every list price in this cluster was read on 21 August 2026 and the date is printed on each page. Vendor pricing moves several times a year, so treat any figure older than a quarter as an estimate and confirm on the vendor's own page before you sign anything. **Do these pages compare enterprise pricing?** No. Enterprise tiers at Atlassian, Asana, monday.com, ClickUp, Airtable and Wrike are all quoted by sales and vary with volume, contract length and region. These pages compare published list prices at the tier a team of five to fifty normally lands on. **Why does Polaris appear on pages about other people's products?** Because it is a third option that a reader comparing two trackers has usually not considered, and hiding it would be dishonest about who wrote the page. It appears once, near the end, under a heading that says exactly what it is. **Which comparison should I read if I am picking a first tool from scratch?** Start with Trello against Asana if the team is under ten people, or Jira against Linear if the team ships software. Both pages cover the question of whether you need a configurable tracker at all, which is usually the real decision hiding under the tool choice. ## Related - https://www.polarishq.co/alternatives - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/slack-vs-microsoft-teams - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/glossary/tool-sprawl --- --- title: "Jira vs Linear: which issue tracker wins, and for whom" description: "Linear wins for one product team shipping weekly. Jira wins when work crosses departments and needs an audit trail. Migration notes and prices, August 2026." url: https://www.polarishq.co/compare/jira-vs-linear section: Comparisons updated: 2026-08-21 --- # Jira vs Linear One tool is opinionated and fast. The other is configurable and permanent. That is the whole decision. ## The short answer Linear suits a single software team under roughly fifty engineers that ships continuously and wants opinionated defaults. Jira suits organisations where work crosses departments, workflows need configuring, and access has to be audited. On list prices checked 21 August 2026, Linear Basic runs $10 per user per month billed yearly; Jira Standard is tracked at $7.91 per user per month. - **Linear Basic:** $10 / user / mo, yearly - **Jira Standard:** $7.91 / user / mo - **Free tiers:** Linear 250 issues · Jira 10 users - **Checked:** 21 August 2026 ## The verdict Pick Linear if you are one product organisation, everyone in the tracker is building the same thing, and you would rather accept somebody else's opinion about how issues, cycles and projects should work than design your own. The speed difference is real and it compounds. Engineers file issues in Linear that they would have skipped in Jira, which is a better argument for it than any feature list. Pick Jira the moment the tracker has to serve people who are not engineers. Support wants a different workflow. Compliance wants a field that cannot be edited after transition. Finance wants a project that nobody outside finance can read. Jira does all of that with configuration; Linear mostly does not, on purpose. The honest failure mode for each is well documented. Jira's is a workflow scheme that grew for six years and now nobody will touch. Linear's is the day a second, differently shaped team joins and discovers that the opinionated defaults were opinions about somebody else's process. ## Head to head List prices read on 21 August 2026. Linear's from linear.app/pricing; Jira's from a pricing tracker updated 31 July 2026, because Atlassian's own pricing page would not render in full for this check. | What decides it | Jira | Linear | | --- | --- | --- | | Entry paid price | Standard tracked at $7.91 per user per month | Basic $10 per user per month, billed yearly | | Next tier up | Premium tracked at $14.54 per user per month | Business $16 per user per month, billed yearly | | Free tier | Up to 10 users, one site, 2 GB storage | 250 issues, 2 teams, unlimited members | | Workflow model | Configurable schemes, custom statuses, transition rules, field-level permissions | Fixed issue states with light customisation, cycles and projects as the primary structure | | Speed | Web app that most teams describe as slow at scale | Keyboard-first, offline-capable client that is the main reason people switch | | Reporting | Deep, including sprint burndown, control charts, cross-project dashboards, JQL | Insights on Business tier; simpler and less configurable | | Third-party apps | Atlassian Marketplace, thousands of apps, Confluence and Bitbucket integration | A short, hand-picked list, with strong GitHub, Slack and Figma integrations | | Learning curve | Days for a user, weeks for an admin | An afternoon, mostly keyboard shortcuts | | Best fit | Multi-department organisations, regulated work, anything needing an audit trail | One product team shipping continuously | ## What each one is genuinely better at Not marketing positions. The things a team notices in month three. **Jira** - Workflow configuration that survives an auditor asking who moved what, when - Permission schemes fine enough to hide a project from most of the company - JQL, which is a real query language and the reason Jira admins stay Jira admins - Portfolio and cross-project reporting at organisation scale - An app marketplace that has already solved your weird requirement **Linear** - Latency low enough that filing an issue costs nothing, so more issues get filed - Cycles that end whether or not the work is finished, which forces honest scoping - Triage as a first-class inbox rather than a backlog nobody opens - Defaults good enough that no one has to own tracker configuration as a job - A free tier with unlimited members, capped on issues rather than people ## Migrating from Jira to Linear, and the part that breaks Linear ships an in-product migration assistant that imports from Jira, Asana, Shortcut and GitHub without touching a CLI, plus a CSV route and an open-source importer package. Throughput is high enough that the copy itself is not the problem: teams report thousands of issues moving in hours, not days. The problem is shape. Jira epics generally land as Linear projects, and Jira sub-tasks generally land as standalone Linear issues, so deeply nested hierarchies arrive flatter than they left. Original creation and modification dates do not carry over, which matters if you were relying on them for anything more than nostalgia. Linear also supports a two-way sync so you can run both while the team decides, which is the sane way to trial it. Going the other direction, Linear to Jira, is materially harder and mostly a CSV exercise. Budget for that asymmetry before you switch, not after. ## Four questions that settle it faster than a trial - **Does anyone outside engineering need to live in this tool?** — If yes, Jira. Linear can be shared with guests on the Business tier, but it is not built to be the company's cross-functional work surface. - **Is there a compliance requirement with the word immutable in it?** — If yes, Jira. Field-level permissions and transition rules exist for exactly this and Linear has no equivalent. - **Does your team already have a Jira admin?** — If no, and nobody wants the job, Jira's flexibility becomes a liability within a year. That is the most common reason teams leave it. - **Is issue-filing friction currently losing you information?** — If engineers are keeping bugs in their heads or a Slack thread because filing is annoying, Linear fixes that on day one. ## A third option worth knowing about Both of these products are trackers. They are very good at holding the record of work and neither of them does any of it. If what actually hurts is not where the tickets live but that there are more tickets than people, that is a different problem and a different kind of tool. Polaris is an all-in-one workspace where AI workers sit on the roster next to humans. Humans and agents are rows in the same members table, so assigning a task to an AI worker uses the same flow as assigning it to a person. A cloud machine wakes for the task, runs a live tool loop with web search, ticks its own acceptance criteria, and posts the result as a comment on the task with any files it produced. The agent never marks work done; a human closes it. The software is free with no seats, and billing is roughly two dollars per human-equivalent hour delivered, itemised on a work log you can challenge line by line. It is in free public beta, with no customers to point at yet. Relevant to this comparison specifically: Linear, GitHub and Slack are all in the Polaris connections catalog, so a team that has just settled on Linear does not have to unsettle anything to give an AI worker access to it. ## Where the third option sits on the same axes | | Jira | Linear | Polaris | | --- | --- | --- | --- | | Price of the software | $7.91 per user per month, Standard | $10 per user per month, Basic | $0, unlimited humans, no tiers | | What you are billed for | Seats, whether or not they log in | Seats | Human-equivalent hours of work an AI worker delivered | | Who does the work | Your team | Your team | Your team, plus AI workers on the same roster | | Maturity | Since 2002 | Since 2019 | Free public beta, no customers yet | ## Questions people ask **Can you migrate Jira issues into Linear?** Yes. Linear has an in-product migration assistant for Jira, plus a CSV import and an open-source importer package. Epics generally become Linear projects and sub-tasks generally become standalone issues, so nested hierarchies flatten, and original created and modified dates do not transfer. **Is Linear cheaper than Jira?** No, not at list price. On 21 August 2026 Linear Basic was $10 per user per month billed yearly against Jira Standard tracked at $7.91 per user per month. Linear's free tier is more generous on people and stricter on volume, capping at 250 issues with unlimited members, while Jira's free tier caps at 10 users. **Does Linear have sprints?** Linear calls them cycles, and they behave differently from Jira sprints in one important way: a cycle ends on its date and unfinished work rolls forward automatically. There is no ceremony for closing one, which teams either love or find too loose depending on how they run planning. **Which one handles a support team better?** Jira, or more precisely Jira Service Management, which is a separate product with its own pricing. Linear added Zendesk and Intercom integrations on the Business tier and can route customer issues into Triage, but it is not a ticketing system for external customers. **Can a non-engineering team use Linear?** Some do, and Linear Asks on the Business tier turns Slack requests into issues for that purpose. The constraint is not capability but shape: Linear's model assumes cycles, projects and a single flow, so a team with a genuinely different process ends up bending its work to fit. ## Related - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/compare/jira-vs-asana - https://www.polarishq.co/compare/linear-vs-shortcut - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/replace/jira-and-confluence-and-slack --- --- title: "Notion vs Confluence: which wiki should your team use?" description: "Confluence wins for big orgs needing space permissions and Jira linking. Notion wins for teams under 100 who will actually write. Prices checked August 2026." url: https://www.polarishq.co/compare/notion-vs-confluence section: Comparisons updated: 2026-08-21 --- # Notion vs Confluence A documentation tool people enjoy against a documentation tool people comply with. ## The short answer Notion wins for teams under about a hundred people who need documents that colleagues will voluntarily open and edit. Confluence wins for larger organisations already running Jira, where page permissions, space administration and required approvals matter more than writing experience. On 21 August 2026 Notion Plus listed at $10 per seat per month; Confluence Standard was tracked at $5.42 per user per month. - **Notion Plus:** $10 / seat / mo - **Confluence Standard:** $5.42 / user / mo - **Free tiers:** Notion unlimited solo · Confluence 10 users - **Checked:** 21 August 2026 ## The verdict If your problem is that nobody writes anything down, choose Notion. Its block editor, databases and page nesting make documentation feel closer to building than to filing, and that difference shows up in how many pages exist a year later. Small and mid-sized teams consistently produce more documentation in Notion than in Confluence, and the reason is ergonomics rather than features. If your problem is that documentation exists but cannot be governed, choose Confluence. Space-level permissions, page restrictions, approval workflows through Marketplace apps, and native two-way linking with Jira issues are things Notion either approximates or does not do. In a company where legal, security or a regulator reads the wiki, that governance layer is the product. The cost comparison is less flattering to Notion than it looks. Confluence Standard was tracked at $5.42 per user per month on 21 August 2026 against Notion Plus at $10, and Confluence's per-user rate falls as you add users while Notion's does not. ## Head to head Notion figures from notion.com/pricing on 21 August 2026. Confluence figures from a pricing tracker of the same date, because Atlassian's pricing page would not render in full for this check; Atlassian's own calculator varies the per-user rate with user count. | What decides it | Notion | Confluence | | --- | --- | --- | | Entry paid price | Plus $10 per seat per month, up to 20% off yearly | Standard tracked at $5.42 per user per month | | Next tier up | Business $20 per seat per month | Premium tracked at $10.44 per user per month | | Free tier | Unlimited pages solo; a 1,000-block cap applies once a second member joins; 10 guests | Up to 10 users | | Editing model | Blocks, drag anywhere, markdown shortcuts, databases inside pages | Structured pages inside spaces, page tree, templates, macros | | Permissions | Page and workspace level; teamspaces on Business | Space and page level, with restrictions that inherit down the tree | | Ticket linking | Two-way sync with Jira and other trackers via integrations | Native, bidirectional, including requirement and release pages | | AI pricing | Included on Business and Enterprise; Custom Agents billed at $10 per 1,000 Notion credits | Rovo credits bundled by tier, with add-on packs | | Learning curve | An hour to use, a week to design a good workspace | Twenty minutes to use, and mostly unchanged since | | Best fit | Startups and teams under about a hundred people | Organisations already committed to Atlassian | ## The things nobody mentions until you have committed - **Notion's free tier changes when a second person arrives** — A solo free workspace has unlimited blocks. Add one colleague and a 1,000-block cap applies, which a real team hits faster than expected. Guests are capped at ten. - **Confluence's price per user falls with volume** — Unlike most per-seat tools, Atlassian's cloud rate declines as you add users. A 200-person deployment does not cost ten times a 20-person one, which changes the arithmetic at scale. - **Notion search is a known weak point at scale** — Once a workspace passes a few thousand pages, finding the right one is harder than in Confluence, whose search was built for large spaces first. - **Confluence exports are workable, Notion exports are lossy** — Notion exports to markdown and HTML, but databases, relations and rollups do not survive the trip intact. If you might leave, structure your important content as pages rather than as database views. ## Who should pick which A twelve-person startup with no Jira instance should use Notion and not think about it again for two years. The block editor is worth the money, the databases will quietly replace three spreadsheets, and nobody will need training. A three-hundred-person company running Jira should use Confluence, and the deciding factor is not the writing experience. It is that requirements pages, release notes and Jira issues stay linked automatically, and that a security team can restrict a space without filing a ticket with whoever owns the workspace. The genuinely hard case is the fifty-to-hundred-person company that grew up on Notion and now has compliance requirements. Migrating is real work and usually not worth it. Running Confluence for the twenty pages that need governance and keeping Notion for everything else is ugly, and it is what most teams in that position actually do. ## A third option worth knowing about Both of these are places to write. Neither writes anything. The documentation problem most teams describe is not that the wiki is bad, it is that the runbook is eleven months stale and nobody has a spare afternoon to reconcile it with reality. Polaris is an all-in-one workspace with a nested document tree and a block editor, versioned files, file review and comments, plus AI workers you can assign work to. The relevant part here is that a document task can be given to a worker rather than to a person who has other things to do. A cloud machine picks it up, does live web research if the task needs it, ticks the acceptance criteria you set, and returns the result as a comment on the task with the files attached. A human reads it and closes it. The software is free with no per-seat charge; billing is roughly two dollars per human-equivalent hour delivered, logged job by job so you can dispute any line. Notion is in the Polaris connections catalog, so a team that keeps Notion as its wiki can still hand a worker access to it. Polaris is in free public beta and has no customers yet, which is worth knowing before you plan a migration around it. ## Questions people ask **Is Confluence cheaper than Notion?** At list price on 21 August 2026, yes. Confluence Standard was tracked at $5.42 per user per month against Notion Plus at $10 per seat per month. Confluence's per-user rate also declines as the user count rises, while Notion's does not, so the gap widens with team size. **Can you import Confluence pages into Notion?** Yes. Notion has a Confluence importer that handles page trees and most formatting. Macros do not survive, attachments need checking, and page restrictions do not transfer at all, so anything that was locked down in Confluence arrives open in Notion. **Does Notion work as a company wiki for a large organisation?** It can, but two limits bite above roughly a hundred people: search quality across thousands of pages, and permission granularity compared with Confluence spaces. Teamspaces on the Business plan help with the second. Neither is fatal, and both are the reasons large companies keep Confluence. **Does Confluence include AI?** Atlassian bundles Rovo credits into Confluence tiers, with additional credit packs available. Notion takes a different shape: AI features sit on the Business and Enterprise plans, and Custom Agents bill separately at $10 per 1,000 Notion credits. Both were checked on 21 August 2026 and both vendors change AI packaging frequently. **Which one is better for public documentation?** Notion publishes a page to the web in one click and is widely used for public help centres and changelogs. Confluence can publish spaces publicly too, but the output looks like Confluence, which is fine internally and rarely what a marketing team wants on a customer-facing domain. ## Related - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/confluence - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/compare/notion-vs-coda - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/replace/confluence-and-notion - https://www.polarishq.co/replace/confluence-and-jira --- --- title: "Asana vs monday.com: which one fits your team's work?" description: "Asana wins when work is a chain of dependent tasks with dates. monday.com wins when work is records in a table. Prices and seat minimums checked August 2026." url: https://www.polarishq.co/compare/asana-vs-monday section: Comparisons updated: 2026-08-21 --- # Asana vs monday.com One models work as a plan with dependencies. The other models it as a spreadsheet that can run automations. ## The short answer Asana suits work that behaves like a plan: dependent tasks, owners, dates that shift together, and goals rolling up to portfolios. monday.com suits work that behaves like a table of records with status columns and automations, such as client pipelines and operations queues. Checked 21 August 2026, monday.com Basic listed at $9 per seat per month with a three-seat minimum; Asana Starter at $10.99 annually. - **monday.com Basic:** $9 / seat / mo, annual - **Asana Starter:** $10.99 / user / mo, annual - **Free tiers:** monday 2 seats · Asana 2 users - **Checked:** 21 August 2026 ## The verdict Choose Asana if your work has a shape. A product launch, a hiring loop, a campaign with sixty tasks where moving one date moves four others. Asana's dependency model, timeline view, workload balancing and goals-to-portfolio rollup exist for exactly that, and monday.com's equivalents are shallower. Choose monday.com if your work is a list of similar things that move through states. Client accounts, applicants, inventory, support escalations, sponsorship deals. Boards behave like a database with typed columns, automations fire on column changes, and non-technical people build usable workflows without help. That is a genuinely different product experience, not a marketing distinction. The pricing looks close and is not. monday.com's Basic tier at $9 per seat per month is cheaper per seat than Asana Starter at $10.99 annually, but monday.com enforces a three-seat minimum and sells seats in blocks, so a four-person team buys five seats and a six-person team buys ten. Asana has no minimum but its free plan now caps at two users for workspaces created after 12 November 2025, so a small team pays sooner than it used to. ## Head to head Read from asana.com/pricing and monday.com/pricing on 21 August 2026. | What decides it | Asana | monday.com | | --- | --- | --- | | Entry paid price | Starter $10.99 per user per month annually, $13.49 monthly | Basic $9 per seat per month annually | | Next tier up | Advanced $24.99 annually, $30.49 monthly | Standard $12, then Pro $19 per seat per month annually | | Seat minimum | None | Three seats on every paid plan, sold in blocks of 3, 5, 10, 15, 20 | | Free tier | Personal, up to 2 users for workspaces created after 12 Nov 2025 | Up to 2 seats, 3 boards, 3 docs | | Core metaphor | Tasks in projects, with dependencies and due dates | Items in boards, with typed columns and status | | Automations | Rules on Starter and above | Counted per tier: 250 per month on Standard, 25,000 on Pro | | Reporting | Goals, portfolios, workload, universal reporting on Advanced | Dashboards built from board widgets | | Learning curve | Familiar to anyone who has used a task tool | Fast to start, and easy to build something that later needs rebuilding | | Best fit | Cross-functional programmes and campaigns | Operational pipelines and record-keeping with status | ## Where each one is genuinely stronger **Asana** - Task dependencies that actually shift downstream dates on the timeline - Workload view, which shows who is over-committed before the week starts - Goals that roll into portfolios, so an executive view exists without a separate tool - A task can live in several projects at once without being duplicated - No seat minimum, so a three-person team pays for three people **monday.com** - Column types rich enough that boards replace operational spreadsheets outright - Automation building that a non-technical operations lead can do unaided - Unlimited free viewers on paid plans, so stakeholders can watch without a seat - Visual customisation that gets teams to adopt it rather than tolerate it - One platform sold alongside CRM and dev products if you want to expand later > **Run the seat arithmetic before the feature comparison** > > At six people, monday.com Standard bills ten seats at $12, which is $120 a month, because seats are sold in blocks. Asana Starter for the same six people is $65.94 a month at the annual rate. The cheaper per-seat price is not the cheaper bill. ## Three questions that settle it - **Does a date change on one task need to move other tasks?** — If yes, Asana. Dependency handling is the clearest functional gap between the two and no amount of monday.com automation closes it cleanly. - **Are you replacing a spreadsheet somebody maintains by hand?** — If yes, monday.com. Typed columns plus automations is exactly that job, and Asana's task model fights you when the unit of work is a record rather than an action. - **How many people, exactly?** — Under five, the monday.com seat minimum and block sizes usually make Asana cheaper despite the higher headline rate. Over twenty, run both quotes properly. ## A third option worth knowing about Both products bill for places to put work, and both have raised the cost of the free tier for small teams in the last year. If the thing making you compare them is the bill rather than the feature set, there is a different model worth a look. Polaris is an all-in-one workspace where the software costs nothing. Unlimited humans, tasks, workstreams and documents, no seats, no tiers, no minimum. What you pay for is work delivered by AI workers on your roster, at roughly two dollars per human-equivalent hour. Hours come from an open formula based on observable effort and are logged job by job on the worker's work log, so any line on the bill can be challenged from the evidence that produced it. Nothing delivered means nothing billed. The mechanism is ordinary to describe: you assign a task to an AI worker exactly as you would to a colleague, a cloud machine wakes up for it and keeps running after you close your laptop, and the result arrives as a comment on the task with any files attached. The worker ticks its own acceptance criteria; a human closes the task. Polaris is in free public beta with no customers yet, so treat it as something to test alongside your current tool rather than a migration to schedule. ## Questions people ask **Does monday.com really require a minimum of three seats?** Yes. Every paid monday.com plan starts at three seats and seats are sold in blocks, so a four-person team buys five and a six-person team buys ten. A solo user or a pair cannot buy a paid plan sized to them, which is the single most common complaint about the pricing. **How many users does Asana's free plan allow?** Two, for workspaces created after 12 November 2025. Accounts created before that date keep a legacy allowance of up to ten seats. The change means new small teams reach a paid plan much faster than Asana's older reputation suggests. **Can you import an Asana project into monday.com, or the reverse?** Both vendors offer CSV import and monday.com publishes an Asana importer. Tasks, owners and dates transfer reliably. Dependencies, custom field types and automation rules do not, so plan on rebuilding the logic rather than moving it. **Which is better for a marketing team?** Asana, in most cases, because campaign work is a dependency chain with dates and Asana models that natively. Marketing teams that mainly track a queue of similar requests, such as a design intake pipeline, are usually happier on monday.com boards. **Do both charge extra for AI features?** monday.com includes basic AI tools from the Basic tier upward and gates more capable features higher. Asana bills AI usage as credits on top of seats. Both vendors repackage AI often, so confirm the current arrangement on their own pricing pages before budgeting. ## Related - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/compare/wrike-vs-asana - https://www.polarishq.co/compare/trello-vs-monday - https://www.polarishq.co/cost/stack-cost-25-person-team --- --- title: "Trello vs Asana: the free-tier gap decides it in 2026" description: "Trello's free plan allows 10 collaborators per workspace. Asana's now caps at 2 users. For small teams that settles it. Prices checked August 2026." url: https://www.polarishq.co/compare/trello-vs-asana section: Comparisons updated: 2026-08-21 --- # Trello vs Asana For a small team on a free plan this is no longer close. For a team with dependencies it never was. ## The short answer Trello wins for small teams that want boards and will stay on a free plan: its free tier allows up to ten collaborators per workspace, while Asana's free Personal plan caps at two users for workspaces created after 12 November 2025. Asana wins once work has dependencies, shifting dates or reporting needs, which Trello handles only through paid views and Power-Ups. - **Trello free:** 10 collaborators per workspace - **Asana free:** 2 users, post-Nov 2025 accounts - **Trello Standard:** $5 / user / mo, annual - **Asana Starter:** $10.99 / user / mo, annual ## The verdict In August 2026 the free tiers decide this for most small teams, and they decide it for Trello. Trello's free plan supports up to ten collaborators per workspace with ten boards. Asana's free Personal plan supports two users for any workspace created after 12 November 2025, with older accounts grandfathered at a higher limit. A five-person team that wants to pay nothing has one option here, and it is not Asana. Once money is on the table the comparison inverts. Asana Starter at $10.99 per user per month annually buys dependency chains, timeline view, custom fields, rules and reporting. Trello Standard at $5 buys unlimited boards and better checklists; you need Premium at $10 to get the calendar, timeline, table and dashboard views that Asana includes at its first paid tier. The real dividing line is whether your work has a critical path. Trello is a board of cards and it is excellent at that. It has no first-class dependency model, so the moment moving one date is supposed to move three others, you are either doing it by hand or paying for a Power-Up that half solves it. ## Head to head Read from trello.com/pricing and asana.com/pricing on 21 August 2026. | What decides it | Trello | Asana | | --- | --- | --- | | Free tier | Up to 10 collaborators per workspace, 10 boards | Up to 2 users for workspaces created after 12 Nov 2025 | | Entry paid price | Standard $5 per user per month annually, $6 monthly | Starter $10.99 annually, $13.49 monthly | | Next tier up | Premium $10 annually, $12.50 monthly | Advanced $24.99 annually, $30.49 monthly | | Enterprise | $17.50 per user per month, billed annually | Quoted by sales | | Views beyond the board | Calendar, timeline, table, dashboard and map, on Premium | List, board, timeline and calendar from Starter | | Dependencies | Not native; approximated with Power-Ups | Native, with dates that shift downstream work | | Automation | Butler, included on all tiers with rising limits | Rules from Starter upward | | Learning curve | Minutes, and no training needed for anyone | An hour, and worth it if the plan is complex | | Best fit | Small teams, simple pipelines, personal organisation | Cross-functional projects with real sequencing | ## What each one is actually for - **Trello is a shared surface, not a planning tool** — Cards in columns, dragged by whoever is looking. That constraint is why adoption is near-total in teams that use it, and why it stops scaling around the point where two people disagree about what a column means. - **Asana is a plan with owners and dates** — Every task has an assignee and a due date by default, and the product nags about both. Teams that need accountability visible without a meeting get it here and not in Trello. - **Both belong to companies with adjacent products** — Trello is Atlassian, which matters if Jira and Confluence are already in the building. Asana stands alone, which some buyers prefer. - **Neither is a document tool** — Trello card descriptions and Asana project briefs are notes, not documentation. Teams using either usually pay for a wiki as well, which is where the second subscription starts. ## Who should pick which A five-person agency tracking client work in columns should use Trello, stay on the free plan, and spend the saved money on something else. Ten collaborators is enough headroom for a while. A twenty-person company running a product launch should use Asana. The dependency chain is the reason. Trello Premium plus Power-Ups can be bent into that shape and the result is worse than Asana at a similar price. The awkward middle is a team of eight that started on Trello and now has three boards nobody can reconcile. That is not a tool problem yet. It becomes one when someone asks which board holds the truth, and the answer takes more than ten seconds. ## A third option worth knowing about Both free tiers got smaller in the last two years, and both companies charge per person for a place to put work. If that is the pattern that brought you to this page, the alternative to a cheaper tracker is a different billing model. Polaris is an all-in-one workspace with lanes shared between list and board view, so you organise once and look at the same work either way, plus documents and team chat in the same product. The software is free: unlimited humans, unlimited tasks, unlimited workstreams, no seat caps and no tiers. Billing only starts when an AI worker on your roster delivers something, at roughly two dollars per human-equivalent hour, itemised on a work log you can dispute line by line. For a small team specifically, the argument is narrow and easy to check: there is no five-person or ten-person threshold at which the software starts charging you, because it never charges for the software. Polaris is in free public beta and has no customers yet. ## What a six-person team pays Annual list rates, August 2026, software only. | | Trello | Asana | Polaris | | --- | --- | --- | --- | | Free plan covers six people | Yes, up to 10 collaborators | No, capped at 2 users | Yes, no cap | | Entry paid plan, six people, per month | $30 on Standard | $65.94 on Starter | $0 | | Timeline and calendar views | Premium, $60 per month | Included on Starter | Lanes shared between list and board view | | What triggers a bill | Adding an eleventh collaborator | Adding a third user | An AI worker delivering finished work | ## Questions people ask **How many people can use Trello for free?** Up to ten collaborators per workspace, with a limit of ten boards per workspace. That was the position on trello.com/pricing on 21 August 2026. Unlimited cards and lists are included, and Butler automation runs on the free tier with lower limits than the paid plans. **Why did Asana reduce its free plan to two users?** Asana cut the free Personal cap to two collaborators for workspaces created after 12 November 2025; accounts opened before that date keep a legacy allowance of up to ten. Asana has not published a detailed rationale, and the practical effect is that new small teams reach a paid plan far sooner. **Can you migrate Trello boards into Asana?** Yes. Asana has a Trello importer that maps boards to projects, lists to sections and cards to tasks. Checklists become subtasks, attachments transfer, and Power-Up data does not come across at all, so anything a Power-Up was storing needs exporting separately first. **Does Trello support task dependencies?** Not natively. Several Power-Ups add dependency tracking and some work well, but they store the relationship outside Trello's core data model, so it does not appear in exports and does not drive date shifting the way Asana's dependencies do. **Is Trello still maintained now that Atlassian owns Jira and Confluence?** Yes. Trello continues to ship, added AI features on paid tiers, and prices independently of Jira. It remains positioned at a different buyer, so there is no sign of it being folded into Jira, though Atlassian has never guaranteed that either. ## Related - https://www.polarishq.co/alternatives/trello - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/cost/trello-pricing - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/compare/trello-vs-monday - https://www.polarishq.co/compare/todoist-vs-asana - https://www.polarishq.co/compare/basecamp-vs-asana - https://www.polarishq.co/replace/notion-and-trello --- --- title: "Linear vs Shortcut: speed against structure for dev teams" description: "Shortcut keeps epics and custom workflows at $8.50 per user. Linear trades that structure for speed. Free tiers differ sharply. Prices checked August 2026." url: https://www.polarishq.co/compare/linear-vs-shortcut section: Comparisons updated: 2026-08-21 --- # Linear vs Shortcut Two trackers built for engineers, disagreeing about how much structure a team should be allowed to have. ## The short answer Shortcut suits engineering teams that want epics, stories, milestones and configurable workflows without Jira's administrative weight, at $8.50 per user per month billed annually as checked on 21 August 2026. Linear suits teams that would rather have fewer options and a faster client, at $10 per user per month billed yearly. The free tiers differ: Shortcut allows ten users, Linear allows unlimited members but only 250 issues. - **Shortcut Team:** $8.50 / user / mo, annual - **Linear Basic:** $10 / user / mo, yearly - **Free tiers:** Shortcut 10 users · Linear 250 issues - **Checked:** 21 August 2026 ## The verdict Shortcut is the better choice for a team that liked Jira's structure and hated Jira's administration. It keeps the epic, story and task hierarchy, keeps milestones above epics, keeps iterations, and lets you build up to five custom workflows on the Team plan without hiring anyone to own the configuration. It is also cheaper, at $8.50 per user per month annually against Linear's $10. Linear is the better choice for a team that wants the tracker to disappear. The client is fast enough that filing an issue is not a decision, Triage turns incoming reports into a queue somebody actually clears, and the defaults are opinionated enough that nobody spends a Thursday arguing about statuses. That is worth $1.50 a head to a lot of teams and worth nothing to others. The free tiers are shaped so differently that they can decide it outright. Shortcut's free plan takes ten users with no cap on issues, which suits a small team with a large backlog. Linear's free plan takes unlimited members but stops at 250 issues and two teams, which suits a larger group evaluating it on a slice of work. ## Head to head Read from shortcut.com/pricing and linear.app/pricing on 21 August 2026. Both vendors quote annual rates; monthly billing costs more at each tier. | What decides it | Linear | Shortcut | | --- | --- | --- | | Entry paid price | Basic $10 per user per month, billed yearly | Team $8.50 per user per month, billed annually | | Next tier up | Business $16 per user per month | Business $12 per user per month | | Free tier | 250 issues, 2 teams, unlimited members | Up to 10 users, unlimited issues | | Hierarchy | Projects and issues, with sub-issues | Milestones, epics, stories and tasks | | Workflow customisation | Light; states are largely fixed by design | Up to 5 custom workflows on Team, unlimited on Business | | Iterations | Cycles, which end on date and roll work forward | Iterations with explicit start and end, closer to sprints | | Reporting | Insights on Business | Advanced reports on Team, objectives and key results on Business | | Speed | The reason most switchers cite | Fast, but not the product's headline claim | | Best fit | Teams that want one opinionated flow | Teams that need several flows without an admin | ## Where each one wins on merit **Linear** - Triage as a real inbox, so bug reports do not rot in a backlog - A desktop client that stays usable when the network does not - Cycles that force honest scoping by ending whether or not work is done - An agent platform available even on the free tier - Design consistency tight enough that new hires need no walkthrough **Shortcut** - Epics and milestones, so a quarter of work has a container above the sprint - Custom workflows per team without a configuration owner - Docs built into the same product, which removes one subscription - A free tier that does not cap issue volume - A lower per-seat price at both paid tiers ## Two questions that decide it - **Does more than one team need a different set of states?** — If yes, Shortcut. This is the clearest functional gap. Linear's fixed states are a feature until the second team arrives with a genuinely different process. - **Is anyone currently not filing issues because filing is annoying?** — If yes, Linear. Lost information is more expensive than a workflow compromise, and Linear's whole design argument is about that friction. ## A third option worth knowing about Both of these are engineering trackers, and both have added agent features in the last two years. If part of what you are evaluating is how AI fits into your team's work rather than into a single developer's editor, the shape of that question is worth separating from the tracker question. Polaris is an all-in-one workspace where AI workers are members of the team rather than a feature of the tracker. Humans and agents live in the same members table, so an AI worker is assigned a task through the same flow as a colleague. A cloud machine wakes for that task, runs a live tool loop with real web search, ticks the acceptance criteria on the task as it goes, and posts the finished work as a comment with any files it produced. The machine never marks the task done; a person closes it and rates it. Capabilities are stored as a SKILL.md file you can open and edit rather than a hidden prompt, which is the part engineers tend to care about. The software is free with no seats; billing is roughly two dollars per human-equivalent hour delivered, from an open formula, logged per job. Linear, GitHub and Slack are all in the connections catalog. Polaris is in free public beta with no customers yet. ## Questions people ask **Can you import Shortcut data into Linear?** Yes. Linear's migration assistant supports Shortcut alongside Jira, Asana and GitHub, and there is a CSV route as well. Epics generally map to Linear projects. Custom workflow states have no direct equivalent, so anything relying on a hand-built state machine needs remapping by hand before the import. **Which is cheaper, Linear or Shortcut?** Shortcut, at both paid tiers as checked on 21 August 2026: $8.50 against $10 at entry, and $12 against $16 at the next level, all billed annually. For a twenty-person team the Business-tier gap is about $960 a year. **Does Shortcut have something like Linear's Triage?** Not as a distinct product surface. Shortcut handles incoming work through workflow states and integrations rather than a dedicated triage inbox. Teams that route a high volume of external bug reports usually find Linear's model less effortful. **Do both integrate with GitHub the same way?** Both support branch naming conventions, pull request linking and automatic state transitions on merge, and both are reliable at it. Linear's version is slightly more automatic out of the box; Shortcut's is more configurable. Neither is a reason to choose one over the other. **Is Shortcut a good Jira replacement?** For an engineering-only Jira instance, usually yes: the hierarchy is familiar, workflows are configurable, and the administrative burden is far lower. For a Jira instance shared with support, legal or finance, no. Shortcut is built for software teams and does not try to be a general work platform. ## Related - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/alternatives/shortcut - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/compare/linear-vs-height - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/glossary/skill-file --- --- title: "ClickUp vs monday.com: price, AI add-ons and real cost" description: "ClickUp is cheaper per seat and highly configurable. monday.com is easier to keep tidy. AI add-ons change both bills. Seat rules checked August 2026." url: https://www.polarishq.co/compare/clickup-vs-monday section: Comparisons updated: 2026-08-21 --- # ClickUp vs monday.com Two platforms that both claim to replace your whole stack, with very different ideas about who does the configuring. ## The short answer ClickUp suits teams with someone willing to configure it: $7 per user per month billed yearly for Unlimited, with more views, hierarchy levels and customisation than monday.com. monday.com suits operations teams that need boards a non-technical owner can maintain, from $9 per seat per month with a three-seat minimum. Both figures were read on 21 August 2026, and both vendors sell AI as a separate charge. - **ClickUp Unlimited:** $7 / user / mo, yearly - **monday.com Basic:** $9 / seat / mo, annual - **ClickUp Brain add-on:** $9 / user / mo - **Checked:** 21 August 2026 ## The verdict ClickUp is the better buy if somebody on the team enjoys building systems. It is cheaper at every equivalent tier, has more views, a deeper hierarchy, and a customisation surface that will do almost anything you ask of it. The cost is that a ClickUp workspace reflects whoever set it up, and an unowned one degrades into a maze within a year. monday.com is the better buy if nobody wants that job. Boards are simpler, the visual model is easier for a non-technical operations lead to keep tidy, and automations are built through a sentence-shaped builder that most people can read. You pay for that in seat minimums, block-based seat purchasing, and a lower ceiling on what the tool can be bent into. Neither headline price is the price. ClickUp Business is $12 per user per month billed yearly, and ClickUp Brain is a separate $9 per user per month, so an AI-using team is at $21 rather than $12. monday.com bundles basic AI from the entry tier and gates the useful parts higher. Compare the tier you will actually run, with AI included, or the comparison is fiction. ## Head to head Read from clickup.com/pricing and monday.com/pricing on 21 August 2026. | What decides it | ClickUp | monday.com | | --- | --- | --- | | Entry paid price | Unlimited $7 per user per month yearly, $10 monthly | Basic $9 per seat per month annually | | Next tier up | Business $12 yearly, $19 monthly | Standard $12, Pro $19 per seat per month annually | | Seat minimum | None | Three seats, sold in blocks of 3, 5, 10, 15, 20 | | Free tier | Free Forever, unlimited members with feature limits | Up to 2 seats, 3 boards, 3 docs | | AI pricing | Brain add-on $9 per user per month; Everything AI $28 | Basic AI tools included from Basic; more capable features on higher tiers | | Hierarchy | Workspace, space, folder, list, task, subtask | Workspace, board, group, item, subitem | | Views | List, board, calendar, Gantt, timeline, table, mind map, whiteboard and more | Table, kanban, calendar, timeline and Gantt from the relevant tiers | | Who keeps it tidy | Whoever built it, and they have to stay | The operations lead, without engineering help | | Best fit | Teams that want one tool to do everything and will configure it | Operational pipelines run by non-technical owners | ## What twelve people actually pay per month List rates, annual billing, August 2026. monday.com seats round up to the next block. - **$84** — ClickUp Unlimited, 12 users. $7 per user, no minimum, no AI. - **$180** — monday.com Standard, 15 seats. Twelve people round up to a 15-seat block at $12. - **$252** — ClickUp Business with Brain, 12 users. $12 plus the $9 AI add-on, per user. ## The honest strengths **ClickUp** - Cheapest credible all-in-one at the entry tier - Docs, whiteboards, goals and time tracking inside the same subscription - Custom fields and automations deep enough to model unusual processes - A free tier with unlimited members rather than a two-seat cap - Granular permissions on the Business tier without an enterprise conversation **monday.com** - A board model simple enough that adoption is not a project - Automation building that reads like English and needs no training - Unlimited free viewers, so stakeholders watch without consuming seats - Consistent performance, where ClickUp has a long-standing reputation for slowness - Adjacent CRM and dev products if the company later wants one vendor > **The seat block is the hidden line item** > > monday.com sells paid seats in blocks of 3, 5, 10, 15 and 20. A team of six buys ten seats. A team of eleven buys fifteen. On Standard at $12 that is $48 a month for people who do not exist, which is more than most teams' entire ClickUp bill at the entry tier. ## A third option worth knowing about Both of these products sell themselves as the one tool that replaces the rest of your stack, and both then price AI as a separate per-seat line on top. That is the pattern worth naming: the software is billed by how many people might use it, and the AI is billed by how many people might use that too, regardless of what either produces. Polaris inverts both. The software costs nothing, permanently, for unlimited humans, tasks, workstreams and documents, with no seats and no tiers. The only bill is for work an AI worker delivered, at roughly two dollars per human-equivalent hour. Hours are estimated by an open formula from observable effort, clamped between five minutes and eight hours per session, and written to a work log job by job, so every line on the invoice points at the evidence that produced it. Nothing delivered means nothing billed. Assigning work to an AI worker uses the same flow as assigning it to a colleague, because humans and agents are the same kind of record in the system. A cloud machine wakes for the task and keeps working after your laptop is shut, then posts the result as a comment on the task with any files attached. Polaris is in free public beta and has no customers yet. ## Questions people ask **Is ClickUp cheaper than monday.com?** At list price, yes at every comparable tier: $7 against $9 at entry and $12 against $12 with more included at the next level, all annual, as checked on 21 August 2026. The gap widens because monday.com enforces a three-seat minimum and sells seats in blocks, while ClickUp bills the exact number of users. **How much does ClickUp Brain cost on top of a plan?** ClickUp Brain was $9 per user per month on 21 August 2026, added to whichever plan you are on, with an Everything AI tier at $28 per user per month. Super Credits can also be bought separately. Budget the plan and the AI add-on together, because the add-on roughly doubles a Business-tier bill. **Can you migrate from monday.com to ClickUp?** ClickUp publishes an importer for monday.com and both tools export CSV. Items, owners, dates and simple columns transfer. Automations, board-level integrations and formula columns do not, so the rebuild effort sits in the logic rather than the data. **Which one is slower?** ClickUp has the longer-standing reputation for latency, particularly in large workspaces with many custom fields and automations. It has improved substantially since the 3.0 rewrite. If speed is the deciding factor for your team, trial both with a realistic data volume rather than trusting either vendor's benchmark. **Do either of them replace a wiki?** ClickUp Docs is closer to a real wiki than monday.com Docs, with nesting and a full editor, and some teams do drop a separate documentation tool because of it. monday.com Docs is lighter and usually supplements rather than replaces Notion or Confluence. ## Related - https://www.polarishq.co/alternatives/clickup - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/cost/clickup-pricing - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/cost/per-seat-vs-usage-pricing --- --- title: "Notion vs Coda in 2026: Coda is now Superhuman Docs" description: "Coda became Superhuman Docs in July 2026 after the Grammarly deal. What that means for the Notion comparison, plus Maker billing maths. Checked August 2026." url: https://www.polarishq.co/compare/notion-vs-coda section: Comparisons updated: 2026-08-21 --- # Notion vs Coda The product comparison still holds. The company comparison changed underneath it this summer. ## The short answer Notion is the safer default for most teams in 2026 because Coda is being absorbed into the Superhuman suite: Coda became Superhuman Docs in July 2026 and coda.io/pricing now redirects to superhuman.com. Coda's per-Maker billing remains genuinely cheaper where a few people build documents and many only read them, since editors and viewers are free on paid workspaces. - **Coda now:** Superhuman Docs, since July 2026 - **Coda Pro:** ~$10 / Doc Maker / mo, annual - **Notion Plus:** $10 / seat / mo - **Checked:** 21 August 2026 > **The thing that changed** > > Grammarly acquired Coda, then renamed itself Superhuman, and on 8 July 2026 Coda became Superhuman Docs inside that suite. Checked on 21 August 2026, coda.io/pricing returns a redirect to superhuman.com. Existing Coda customers are described as grandfathered on their current plans. Any comparison written before mid-2026 is comparing a company that no longer exists under that name. ## The verdict If you are choosing a documentation and lightweight-database tool today and you want to make the decision once, choose Notion. Not because the product is better on every axis, but because you are also choosing a roadmap, and Coda's is now a component of a suite built around writing assistance, email and an AI assistant. That may end well. It is still a variable you did not have to accept. If you have a specific shape of team, Coda is genuinely better and always was. Its per-Maker billing charges only for the people who build documents; editors and viewers cost nothing on a paid workspace. For a forty-person company where five people build the tooling and thirty-five read it, that is a materially smaller bill than Notion's per-seat model, which charges for everyone. On the editor itself the two have converged. Coda's tables, formulas and Packs are still more capable as a spreadsheet-in-a-document; Notion's databases are simpler and there are far more templates, importers and third-party integrations for it. If you want a document that behaves like an application, Coda still wins that argument on the merits. ## Head to head Notion figures read from notion.com/pricing on 21 August 2026. Coda figures are tracked rather than read from a live vendor page, because coda.io/pricing now redirects into the Superhuman site; treat them as indicative and confirm before buying. | What decides it | Notion | Coda / Superhuman Docs | | --- | --- | --- | | Billing unit | Every member is a paid seat | Only Doc Makers are billed; editors and viewers are free | | Entry paid price | Plus $10 per seat per month, up to 20% off yearly | Pro tracked at $10 per Doc Maker per month annually, $12 monthly | | Next tier up | Business $20 per seat per month | Team tracked at $30 per Doc Maker per month annually, $36 monthly | | Free tier | Unlimited pages solo; 1,000-block cap once a second member joins; 10 guests | Free plan with limits on doc size and objects | | Tables and formulas | Databases with relations and rollups | Closer to a spreadsheet, with a fuller formula language | | Extensibility | Integrations, an API, and a very large template library | Packs, which run logic inside a doc | | Company status | Independent, well capitalised | Part of the Superhuman suite since July 2026 | | Best fit | Most teams that want one wiki and to stop thinking about it | Teams where few people build and many people read | ## The Maker maths, worked A forty-person company where five people build documents. Annual list rates, August 2026. - **$400** — Notion Plus, 40 seats, per month. $10 per seat, everyone billed. - **~$50** — Coda Pro, 5 Doc Makers, per month. Editors and viewers free on paid workspaces. - **8x** — The gap at this shape of team. It closes fast as the number of builders rises. ## What to weigh besides price - **Roadmap risk is now asymmetric** — Notion's roadmap is Notion's. Coda's is set inside a suite with three other products competing for attention. Existing customers keep their plans, which is a reasonable commitment, but it is not the same as independence. - **Notion exports are lossy in a specific way** — Pages export to markdown and HTML cleanly. Databases, relations and rollups do not survive intact. If leaving is a live consideration, keep the content that matters as pages. - **Coda Packs are powerful and are also lock-in** — A doc that runs on Packs is a small application. It does not port anywhere, in either direction. That is the price of the capability, and it is worth paying knowingly. - **The AI story is unsettled on both sides** — Notion gates AI to Business and Enterprise and bills Custom Agents at $10 per 1,000 credits. Superhuman is repackaging Coda's AI into a suite bundle. Neither arrangement is stable enough to budget from for more than a quarter. ## A third option worth knowing about Both of these tools are ways to build a document that does something. The limit they share is that the document still waits for a person to open it. A Coda Pack pulls data in; nobody reads the result and writes the recommendation. Polaris is an all-in-one workspace with a nested document tree, a block editor with markdown shortcuts, versioned files and comments, plus AI workers on the same roster as your colleagues. A research or writing task assigned to a worker is picked up by a cloud machine, which runs a live tool loop with real web search, ticks the acceptance criteria on the task, and posts the finished piece as a comment with any generated files attached. A person reads it and closes it; the machine never closes anything itself. On billing, Polaris takes neither the per-seat nor the per-Maker route. The software is free for unlimited people; the bill is roughly two dollars per human-equivalent hour of delivered work, logged job by job so you can challenge any line from the work log. Polaris is in free public beta and has no customers yet, which is the relevant caveat for anyone who just watched their document tool get acquired. ## Questions people ask **Is Coda shutting down?** No. Coda became Superhuman Docs on 8 July 2026 as part of the Superhuman suite, following Grammarly's acquisition of Coda and Grammarly's own rename to Superhuman. Existing Coda customers are described as grandfathered on their current plans unless they connect their workspace to the Superhuman bundle. **What is a Doc Maker in Coda's pricing?** A Doc Maker is anyone who creates or structurally edits a document in a paid workspace. Only Doc Makers are billed. People who edit content within an existing doc, comment or view it are free, which is why Coda is much cheaper than Notion for teams with few builders and many readers. **Can you migrate a Coda doc to Notion?** Partially. Notion imports Coda documents and handles text, headings and simple tables. Formulas, Packs, buttons and automations do not transfer, and a doc built as an application will need rebuilding rather than importing. Export your tables as CSV first so you have the data independent of the layout. **Which is better for a spreadsheet-heavy team?** Coda. Its formula language is closer to a real spreadsheet, its tables handle larger datasets more gracefully, and Packs let a doc pull live data from other systems. Notion databases are easier to learn and less capable once the logic gets complicated. **Does the acquisition change Coda's pricing?** Not for existing customers according to the transition notices, which say current plans continue as they are. New pricing now sits under the Superhuman brand rather than at coda.io, so the figures on this page are tracked rather than confirmed from a live Coda pricing page, and you should verify before committing. ## Related - https://www.polarishq.co/alternatives/coda - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/compare/notion-vs-airtable - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/replace/confluence-and-notion - https://www.polarishq.co/glossary/all-in-one-workspace --- --- title: "Slack vs Microsoft Teams: the price gap is $1.60 a head" description: "Microsoft prices Teams at about $1.60 per user inside Business Basic. Slack Pro is $7.25. Why teams still pay it, and when they should not. Checked August 2026." url: https://www.polarishq.co/compare/slack-vs-microsoft-teams section: Comparisons updated: 2026-08-21 --- # Slack vs Microsoft Teams This is not really a chat comparison. It is a question about which suite you already pay for. ## The short answer Microsoft Teams wins on cost for any company already buying Microsoft 365: Business Basic listed at $7.00 per user per month with Teams and $5.40 without on 21 August 2026, pricing Teams at roughly $1.60 a head. Slack Pro listed at $7.25 per user per month billed annually. Slack's case is its integration surface, external Slack Connect channels and daily working experience, not price. - **Teams inside M365:** ~$1.60 / user / mo - **Teams Essentials:** $4.00 / user / mo, annual - **Slack Pro:** $7.25 / user / mo, annual - **Checked:** 21 August 2026 ## The verdict If your company already pays for Microsoft 365, use Teams and stop the evaluation. Microsoft's own price list makes the argument: on 21 August 2026, Microsoft 365 Business Basic listed at $7.00 per user per month with Teams included and $5.40 per user per month without it. The chat product is being sold to you for about $1.60 a head. Slack Pro at $7.25 is more than four times that for the same seat, and it does not come with Word, Excel, Outlook or a terabyte of storage. If you are not on Microsoft 365, the comparison opens up. Teams Essentials at $4.00 per user per month is chat, calling and meetings without the Office apps, which is a real product and cheaper than Slack. Slack Pro at $7.25 buys a materially better daily experience, a much larger third-party app directory, and Slack Connect channels with other companies, which is the feature that most often makes the decision for agencies, investors and anyone whose work crosses org boundaries. Slack's free tier deserves a warning rather than a recommendation. Ninety days of message history means the answer somebody gave in March is gone by July. Teams' free tier keeps history. For a small team that intends to stay free, that is the whole comparison. ## Head to head Read from slack.com/pricing and Microsoft's Teams plan comparison on 21 August 2026. All prices are annual-billing rates. | What decides it | Slack | Microsoft Teams | | --- | --- | --- | | Free tier | 90 days of message history | Free version with retained history and meeting limits | | Entry paid price | Pro $7.25 per user per month annually, $8.75 monthly | Teams Essentials $4.00 per user per month annually | | Inside a suite | Standalone | Business Basic $7.00 with Teams, $5.40 without, so Teams is about $1.60 | | Next tier up | Business+ $15 annually, $18 monthly | Business Standard with Copilot $23.50 annually | | Third-party apps | Very large directory, and the default integration target for most SaaS | Growing, and strongest inside the Microsoft estate | | Cross-company channels | Slack Connect, widely used | External access and shared channels, more administratively involved | | Meetings and calling | Huddles and calls, adequate rather than a selling point | Full meeting platform, phone system available, the stronger product | | Compliance and eDiscovery | Available on higher tiers | Purview and the Microsoft compliance stack, hard to beat | | Best fit | Companies that live in chat and integrate many tools | Companies already on Microsoft 365, or that need serious meetings | ## Why teams still pay for Slack The reasons are real, and none of them are price. **Slack's genuine advantages** - Channels with external partners that behave like internal channels - The default place any SaaS vendor ships an integration first - Search that people actually use, on paid tiers - Threads that hold a conversation instead of scattering it - Workflow Builder for small automations without engineering time **Microsoft Teams' genuine advantages** - Effectively bundled into a licence most companies already hold - Meetings, webinars and telephony at a standard Slack does not match - Files that live in SharePoint and OneDrive with the permissions already set - Compliance, retention and eDiscovery through the Microsoft stack - Free tier that does not delete your history after ninety days ## Questions that decide it faster than a trial - **Do you already pay for Microsoft 365?** — If yes, the burden of proof is on Slack to justify a second bill. Sometimes it clears that bar. Usually it does not, and the honest answer is that people prefer Slack rather than that Slack is required. - **How much of your work happens with people outside the company?** — If a lot, Slack Connect is the strongest single argument in this comparison. Teams can do external access, and it is more administratively involved on both sides. - **Do you run webinars or use a phone system?** — If yes, Teams. The meeting and telephony gap is wide and Slack has never seriously contested it. - **Is anyone relying on the free tier for a record?** — If yes, not Slack. Ninety days of history is fine for banter and unsuitable for decisions. ## A third option worth knowing about Whichever you pick, the same thing happens: decisions get made in a channel and then have to be manually copied into a tracker by whoever remembers. Most of them are not. That loss is the actual cost of team chat and neither vendor treats it as their problem. Polaris includes team chat alongside tasks and documents, and its Inbox is built around exactly that gap. A Slack message arrives as a prefilled task suggestion with the workstream, lane, labels and owner already filled in, waiting for one click. Signals become suggestions rather than silent tasks, so nothing is created behind your back and nothing is lost in scrollback either. Slack is in the Polaris connections catalog, so this does not require leaving it. The software is free with no seats; billing is roughly two dollars per human-equivalent hour that an AI worker on your roster delivers, itemised on a work log you can challenge. Polaris is in free public beta with no customers yet. ## Questions people ask **Is Microsoft Teams free?** There is a free version of Teams with limits on meeting length and participants, and it retains message history rather than expiring it. Paid Teams starts at Teams Essentials, listed at $4.00 per user per month billed annually on 21 August 2026, and is also included in Microsoft 365 Business plans. **Why does Microsoft sell Business Basic with and without Teams?** Microsoft unbundled Teams from its Microsoft 365 suites following regulatory pressure in Europe, and now lists both versions. On 21 August 2026 the difference was $7.00 with Teams against $5.40 without, which is the closest thing to a published standalone price for Teams inside a suite. **Can you migrate Slack history into Microsoft Teams?** Yes, with third-party migration tools, and it is more work than it sounds. Channels, direct messages and files can be moved. Threading structure, reactions, custom emoji and app-generated messages transfer inconsistently, and message permalinks in old documents will break. **How long does Slack keep messages on the free plan?** Ninety days. Older messages and files are not deleted but become inaccessible until you upgrade, at which point history returns. Teams that treat Slack as a searchable record need at least the Pro plan, listed at $7.25 per user per month annually as of 21 August 2026. **Which one has better AI features?** Microsoft has the deeper integration through Copilot, which reads across mail, files, meetings and chat, and prices it accordingly: Business Standard with Copilot listed at $23.50 per user per month. Slack AI is narrower, focused on channel summaries and search. Both are additional cost rather than included capability. ## Related - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/replace/jira-and-slack - https://www.polarishq.co/replace/linear-and-slack - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/glossary/tool-sprawl --- --- title: "Jira vs Asana: engineering work against everything else" description: "Jira is built for engineering workflow and audit. Asana is built for cross-functional plans with dates. Which to run, and what to do when you need both." url: https://www.polarishq.co/compare/jira-vs-asana section: Comparisons updated: 2026-08-21 --- # Jira vs Asana Most companies that argue about this end up running both. The useful question is which one owns the boundary. ## The short answer Jira suits engineering work: configurable workflows, sprints, release versions, and an audit trail an assessor will accept. Asana suits cross-functional programmes where owners, dates and dependencies matter more than workflow rules. Checked 21 August 2026, Jira Standard was tracked at $7.91 per user per month against Asana Starter at $10.99 billed annually, and Jira's free tier takes ten users to Asana's two. - **Jira Standard:** $7.91 / user / mo - **Asana Starter:** $10.99 / user / mo, annual - **Free tiers:** Jira 10 users · Asana 2 users - **Checked:** 21 August 2026 ## The verdict Jira wins for anything an engineering team owns end to end. Sprints, versions, release notes, bug workflows with states that non-engineers cannot skip, and a history that survives an auditor's questions. Asana can hold engineering work, and teams that try it usually discover within a quarter that they are rebuilding sprints out of custom fields. Asana wins for everything a company does that is not shipping software. A product launch that spans marketing, legal, support and sales. A hiring loop. An office move. The unit is a task with an owner and a date, dependencies shift downstream work, and the whole thing is legible to somebody who has never heard the word backlog. The uncomfortable answer is that a company past about forty people usually needs both, and the failure mode is not paying twice. It is that the same piece of work exists in both, diverges, and two teams make plans from different truths. Decide which system owns the boundary object, usually the epic or the launch, and make the other one link to it rather than copy it. ## Head to head Asana figures read from asana.com/pricing on 21 August 2026. Jira figures from a pricing tracker updated 31 July 2026, because Atlassian's own pricing page would not render in full for this check. | What decides it | Jira | Asana | | --- | --- | --- | | Entry paid price | Standard tracked at $7.91 per user per month | Starter $10.99 annually, $13.49 monthly | | Next tier up | Premium tracked at $14.54 per user per month | Advanced $24.99 annually, $30.49 monthly | | Free tier | Up to 10 users, one site, 2 GB storage | Personal, up to 2 users for workspaces created after 12 Nov 2025 | | Unit of work | Issue, inside a project with a workflow scheme | Task, inside a project, optionally in several projects at once | | Sprints and releases | Native, with burndown, velocity and version management | Not native; approximated with sections and custom fields | | Dependencies | Issue links, which do not shift dates | Native dependencies that shift downstream dates on the timeline | | Permissions | Scheme-based, down to field level | Project and team level, private projects on paid tiers | | Reporting | JQL, dashboards, control charts, portfolio tooling | Goals, portfolios, workload, universal reporting on Advanced | | Who it reads well to | Engineers and anyone who has been trained | Anyone, on day one | ## The strengths that matter in month three **Jira** - A workflow that prevents a state change rather than asking politely - Version and release management tied to the issues that shipped in it - JQL for the questions no dashboard anticipated - Native Confluence and Bitbucket linking if you are already in that estate - A free tier that covers ten people, which now beats Asana's outright **Asana** - Dependencies that actually move dates instead of merely recording a relationship - Workload view, which surfaces over-commitment before the week starts - One task visible in several projects without duplication or sync - Goals that roll up to a portfolio view an executive will read - Onboarding measured in minutes rather than in a training session > **The free-tier comparison flipped** > > Jira's free plan has covered ten users for years. Asana's free Personal plan now caps at two users for workspaces created after 12 November 2025. For a small team evaluating both at zero cost, the tool with the reputation for being heavyweight is the one that will let your whole team in. ## If you have to run both, do these four things Every company that lands here rediscovers the same rules the hard way. - **Name the boundary object** — One concept, usually the launch or the epic, exists in one system and is linked from the other. Never duplicated, never synced two ways. - **Pick the system of record per kind of work, not per team** — Bugs live in Jira even when marketing files them. Campaigns live in Asana even when engineering has tasks in them. - **Ban the status-copy meeting** — If someone's job is to read one tool and update the other every Monday, the integration is not working and the meeting is hiding it. - **Count the seats twice** — A person in both tools costs $7.91 plus $10.99 a month at the entry tiers. At forty overlapping people that is roughly $9,000 a year for the privilege of two truths. ## A third option worth knowing about The seat arithmetic above is the reason this page ends here rather than with a link to an integration. Two trackers, both billing per person, both holding a record of work that neither of them performs. Polaris puts tasks, documents and team chat in one product and charges nothing for the software: unlimited humans, unlimited tasks, unlimited workstreams, no seats and no tiers. The bill is for work delivered by AI workers on the roster, at roughly two dollars per human-equivalent hour, estimated by an open formula and logged job by job so any line can be challenged from the evidence. The part that is relevant to a company running two trackers: humans and AI workers are the same kind of record, so a cross-functional programme and an engineering task use the same assignment flow, and an AI worker can be given the task instead of a person. A cloud machine wakes for it and keeps running after the laptop closes, then delivers the result as a comment on the task with files attached. Polaris is in free public beta and has no customers yet. ## Questions people ask **Can Asana replace Jira for a software team?** For a small team shipping a simple product, sometimes. For anything with sprints, releases and bug workflows, no. Asana has no native sprint or version model, so teams rebuild it out of sections and custom fields and end up with a worse Jira that costs more per seat. **Can Jira replace Asana for a marketing team?** Technically yes, and Atlassian markets Jira for business teams. In practice the workflow-scheme model and the vocabulary are a poor fit for campaign work, and adoption suffers. Jira Work Management exists for this and is a better answer than a general Jira project. **How do Jira and Asana integrate with each other?** Both offer a native two-way integration that links an Asana task to a Jira issue and syncs status. It works for a small number of linked items and gets fragile at volume. Treat it as a way to show status across the boundary, not as a way to keep two full backlogs identical. **Which is cheaper for a thirty-person company?** Jira, at list price. Thirty seats of Jira Standard at $7.91 is about $237 a month; thirty seats of Asana Starter at $10.99 annually is about $330. Atlassian's per-user rate also declines as user counts rise, which widens the gap further at scale. **Which one handles resource planning better?** Asana, through the Workload view on higher tiers, which shows capacity per person across projects. Jira needs a Marketplace app or Atlassian's portfolio tooling for the equivalent, which is more capable and considerably more setup. ## Related - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/wrike-vs-asana - https://www.polarishq.co/replace/jira-and-confluence-and-slack --- --- title: "Notion vs Airtable: pages with tables, or a real database" description: "Airtable is a database that costs $20 per seat. Notion is a wiki with tables at $10. Record limits, sync and automations decide it. Prices checked August 2026." url: https://www.polarishq.co/compare/notion-vs-airtable section: Comparisons updated: 2026-08-21 --- # Notion vs Airtable One is a document tool that grew tables. The other is a database that grew documents. It shows. ## The short answer Notion suits teams whose primary unit is a page, with tables as a convenience, at $10 per seat per month for Plus. Airtable suits teams whose primary unit is a record and who need thousands of them, with interfaces, syncs and automations built for that, at $20 per seat per month for Team billed annually. Both prices were checked on 21 August 2026. - **Notion Plus:** $10 / seat / mo - **Airtable Team:** $20 / seat / mo, annual - **Airtable free:** 1,000 records per base - **Checked:** 21 August 2026 ## The verdict Choose Notion if the thing you are building is mostly documents with some structured data attached. A handbook with an owner field. A meeting notes database. A lightweight CRM for forty accounts. Notion is half the price, the writing experience is far better, and its databases are enough for most of what small teams call a database. Choose Airtable if the thing you are building is data that documents happen to describe. A content pipeline with a thousand assets. An inventory. A grants tracker with rollups feeding a dashboard. Airtable's field types, linked records, interfaces, sync from external sources and automation limits are all built for volume, and Notion's are not. The clearest practical test is record count. Airtable publishes explicit caps per base by plan: 1,000 on Free as of February 2026, 50,000 on Team, 125,000 on Business. Notion publishes no hard cap and instead degrades, with large databases becoming slow to filter and unpleasant to work in well before you reach anything like those numbers. Published limits you can plan around are worth more than an unpublished ceiling you discover on a Thursday. ## Head to head Read from notion.com/pricing and airtable.com/pricing on 21 August 2026. Airtable's headline figures are annual-billing rates; monthly billing runs higher. | What decides it | Notion | Airtable | | --- | --- | --- | | Entry paid price | Plus $10 per seat per month, up to 20% off yearly | Team $20 per seat per month annually, around $24 monthly | | Next tier up | Business $20 per seat per month | Business $45 per seat per month annually, around $54 monthly | | Free tier | Unlimited pages solo; 1,000-block cap once a second member joins | 1,000 records per base, reduced from 1,200 in February 2026 | | Record ceiling | None published; performance degrades first | 50,000 per base on Team, 125,000 on Business | | Field types | Enough for most uses, weaker on rollups and lookups at depth | Extensive, including linked records, rollups, formulas and sync | | Views | Table, board, calendar, gallery, list, timeline | Grid, kanban, calendar, gallery, gantt, timeline and Interfaces | | Writing | The main reason people choose it | Long-form text is a field, not a document | | External data | Integrations and API | Sync sources that pull external tables in and keep them current | | Best fit | Wiki-first teams with some structured data | Data-first teams that need volume and interfaces | ## What twenty seats cost per year List rates, annual billing where offered, August 2026. - **$2,400** — Notion Plus, 20 seats. At the $10 monthly list rate; yearly billing discounts up to 20%. - **$4,800** — Airtable Team, 20 seats. At the $20 annual-billing rate. - **$10,800** — Airtable Business, 20 seats. At the $45 annual-billing rate, which is where interfaces and admin controls get serious. ## Four things that catch teams out - **Notion's free tier changes when the second person joins** — Solo free workspaces have unlimited blocks. Add one colleague and a 1,000-block cap applies. Guests are capped at ten. Plan for the paid tier if this is a team tool. - **Airtable cut the free record limit in February 2026** — From 1,200 to 1,000 records per base. Small, and it moved several hobby projects onto a paid plan, which is presumably the point. - **Notion database exports lose their structure** — Relations and rollups do not survive a markdown or CSV export intact. Airtable's CSV export is clean because the underlying model is a table. - **Airtable Interfaces are the real product for many buyers** — They turn a base into an app a non-technical colleague can use without seeing the data. Notion has no direct equivalent, and this is often what the extra $10 a seat is actually buying. ## Who should pick which A fifteen-person startup wanting one place for docs, meeting notes, a roadmap and a light CRM should use Notion. Airtable would cost twice as much and the writing experience would push people back into Google Docs, which defeats the purpose. A content operation moving four hundred assets a month through commissioning, editing, legal review and publication should use Airtable. Notion will hold it for a while and then get slow, and the slowness will arrive exactly when the volume matters most. Plenty of companies run both, and that is a defensible choice rather than a failure: Notion for the handbook and the meeting notes, Airtable for the two pipelines that genuinely need a database. What is not defensible is paying for Airtable seats for people who only ever read an interface. ## A third option worth knowing about Both of these tools are very good at holding structured work and neither of them fills a row in. The content pipeline still needs someone to do the competitive scan, write the brief and attach the file. That labour is what the seats are ultimately in service of, and neither vendor bills for it because neither vendor does it. Polaris is an all-in-one workspace where AI workers sit on the roster next to people. You assign a task to a worker the same way you assign it to a colleague; a cloud machine wakes for it, runs a live tool loop with real web search, ticks the acceptance criteria on the task, and posts the finished work as a comment with any files it produced, including documents and PDFs. The machine never marks the task complete. A person reads the work, closes it and rates it. The software is free with no per-seat charge, and the bill is roughly two dollars per human-equivalent hour delivered, from an open formula, itemised on a work log you can challenge line by line. Polaris is in free public beta with no customers yet, so it belongs on a shortlist to test rather than in a migration plan. ## Questions people ask **How many records can an Airtable base hold?** As checked on 21 August 2026: 1,000 records per base on the Free plan, 50,000 on Team and 125,000 on Business, with higher limits on Enterprise Scale. The free limit was reduced from 1,200 to 1,000 in February 2026. Attachment storage is capped separately per plan. **Does Notion have a record limit?** Notion publishes no hard record cap for databases on paid plans. The practical limit is performance: filtering and sorting large databases becomes slow well before any theoretical ceiling, and there is no published number to plan against, which is a genuine disadvantage compared with Airtable. **Can you sync data between Notion and Airtable?** Not natively in a way that keeps both current. Third-party tools such as Zapier and Make can push records in either direction, and Airtable can sync from external sources into a base. Two-way sync between the two is fragile enough that most teams pick one as the system of record. **Is Airtable worth twice the price of Notion?** If you need interfaces, published record ceilings, sync sources or serious automations, yes without much argument. If you are using it as a nicer spreadsheet with a few hundred rows, no. The honest test is whether anyone would notice the missing field types within a month. **Which one is better for a CRM?** Airtable, once the pipeline passes a few hundred accounts or anyone needs a dashboard. Below that, a Notion database with a board view does the job at half the seat price and everyone already knows how to use it. Neither is a substitute for a real CRM once sales process and email tracking matter. ## Related - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/airtable - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/compare/notion-vs-coda - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/replace/airtable-and-slack - https://www.polarishq.co/glossary/all-in-one-workspace --- --- title: "Basecamp vs Asana: flat fee or per seat, and what you lose" description: "Basecamp Studio is $59 a month for unlimited users. Asana Starter is $10.99 each. Break-even is six people, and the trade is dependencies. Checked Aug 2026." url: https://www.polarishq.co/compare/basecamp-vs-asana section: Comparisons updated: 2026-08-21 --- # Basecamp vs Asana The cheapest tool here is obvious once you count heads. Whether it can hold your work is the real question. ## The short answer Basecamp charges a flat monthly fee with unlimited users on its Studio plan and above, at $59 per month for up to ten active projects as checked on 21 August 2026. Asana charges $10.99 per user per month for Starter, billed annually, and models dependencies, timelines and workload that Basecamp deliberately omits. The cost lines cross at roughly six people. - **Basecamp Studio:** $59 / mo flat, unlimited users - **Asana Starter:** $10.99 / user / mo, annual - **Break-even:** About 6 people - **Checked:** 21 August 2026 ## The verdict For a team of six or more that runs a moderate number of projects and does not need dependency chains, Basecamp is the better deal by a wide margin and the gap widens with every hire. Twenty people on Basecamp Studio cost $59 a month. Twenty people on Asana Starter cost $219.80 a month at the annual rate. That is not a rounding difference, and Basecamp's flat fee is the whole reason the company still exists in a per-seat market. For a team whose work has a critical path, Asana is worth the per-seat premium and Basecamp cannot substitute. Basecamp has no dependency model, no timeline that shifts dates, no workload view and no portfolio rollup. This is a stated design position rather than a gap they intend to close, and arguing about it misses the point of the product. The constraint people miss when they price Basecamp is projects, not people. Studio at $59 allows ten active projects. Pro at $100 allows twenty-five. An agency with forty live client engagements is on the $300 Unlimited plan, which is still cheaper than Asana at that headcount but is not the $59 headline. ## Head to head Read from basecamp.com/pricing and asana.com/pricing on 21 August 2026. | What decides it | Basecamp | Asana | | --- | --- | --- | | Billing model | Flat monthly fee, unlimited users from Studio upward | Per user per month | | Entry paid price | Freelancer $25 per month, 3 projects, up to 20 users | Starter $10.99 per user per month annually | | Mid tier | Studio $59 per month, 10 projects, unlimited users | Advanced $24.99 per user per month annually | | Top self-serve | Unlimited $300 per month, annual billing, unlimited projects | Enterprise, quoted by sales | | Free tier | 1 project, 1 GB, up to 5 users | Personal, up to 2 users for workspaces created after 12 Nov 2025 | | Real limit | Number of active projects | Number of people | | Dependencies | None, by design | Native, with dates that shift downstream work | | Reporting | Deliberately minimal | Goals, portfolios, workload, universal reporting on Advanced | | What comes bundled | Message boards, chat, docs, file storage, schedules, check-ins | Task management, with docs and chat handled elsewhere | ## Where the lines cross Monthly list cost at three team sizes, August 2026, assuming Basecamp's project limits are not exceeded. - **$59 vs $65.94** — Six people. Basecamp Studio against Asana Starter at the annual rate. Roughly the break-even. - **$59 vs $219.80** — Twenty people. Basecamp is now less than a third of the cost. - **$100 vs $549.50** — Fifty people. Basecamp Pro, 25 projects, against Asana Starter. ## What you are actually trading **What Basecamp gives you** - A bill that does not change when you hire - Message boards, chat, docs, schedules and file storage in the same fee - Automatic check-in questions, which quietly replace a standing meeting - Clients as guests without paying for their seats - A product with almost nothing to configure, so nothing to maintain **What Asana gives you** - Dependencies that move dates instead of merely noting a relationship - Timeline and workload views that show a plan and its capacity - One task visible in multiple projects with no duplication - Goals rolling into portfolios for an executive view - Rules and custom fields to encode a process rather than describe it ## Two teams, two right answers - **A twelve-person design studio with eight client projects** — Basecamp Studio at $59. Client guests are free, the message board replaces most of the email, and nothing in the work needs a critical path. Asana would cost $131.88 a month for a plan whose best features go unused. - **A twelve-person team running a regulated product launch** — Asana. Sixty tasks across four functions where a slipped legal review has to move five downstream dates is precisely what Basecamp declines to model, and rebuilding it in to-do lists produces a worse plan nobody trusts. ## A third option worth knowing about Basecamp's flat fee is a deliberate argument about pricing, and it is the closest thing in this market to the position Polaris takes. Basecamp says you should not be charged more for adding a colleague. The next step from there is asking why you are charged for the software at all, given that the software is not what does the work. In Polaris the software is free, permanently: unlimited humans, unlimited tasks, unlimited workstreams, unlimited documents, no seats, no tiers and no project cap. Tasks, documents and team chat are in the same product, and lanes are shared between list and board view so you organise once. The only bill is for work AI workers on your roster deliver, at roughly two dollars per human-equivalent hour, estimated by an open formula from observable effort and written to a work log job by job. If nothing is delivered, nothing is billed. For an agency in particular the difference from both tools is concrete: an hour of delivered work has a line on an invoice you can trace back to the job that produced it, rather than a monthly subscription that has no relationship to output. Polaris is in free public beta and has no customers yet. ## Questions people ask **Does Basecamp really have no per-user fees?** From the Studio plan upward, correct: $59 a month covers unlimited team members. The Freelancer plan at $25 caps at 20 users. What Basecamp limits instead is active projects, at 3, 10 and 25 on Freelancer, Studio and Pro, with the $300 Unlimited plan removing that cap. **Can Basecamp handle task dependencies?** No, and this is intentional. Basecamp has to-do lists with assignees and due dates but no dependency model, no critical path and no timeline that shifts dates. Teams that need those either work around them with manual date management or use a different tool. **Can you import Asana projects into Basecamp?** Not with a native importer. Both tools export CSV, and third-party migration services handle the mapping, but Basecamp's data model is simpler, so custom fields, dependencies, subtask hierarchies and rules have no destination. Expect to rebuild rather than migrate. **At what team size does Basecamp become cheaper than Asana?** Around six people at the entry tiers. Six Asana Starter seats at $10.99 annually is $65.94 a month against Basecamp Studio at $59. Below six, Asana Starter or Basecamp Freelancer at $25 will usually be cheaper depending on how many projects you run. **Does Basecamp include chat and documents?** Yes. Message boards, real-time chat, documents, file storage, schedules and automatic check-in questions are all included in the flat fee. That bundling is a large part of the value, because an Asana team typically pays separately for Slack and a wiki on top of its seats. ## Related - https://www.polarishq.co/alternatives/basecamp - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/compare/todoist-vs-asana - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team --- --- title: "ClickUp vs Notion: tasks first or documents first?" description: "Both claim to replace your stack. ClickUp is a task platform with docs attached, Notion a doc platform with tasks attached. Limits checked August 2026." url: https://www.polarishq.co/compare/clickup-vs-notion section: Comparisons updated: 2026-08-21 --- # ClickUp vs Notion Both promise one tool for everything. They disagree about which half of everything comes first. ## The short answer ClickUp suits teams whose centre of gravity is execution: assignees, due dates, sprints, automations and time tracking, from $7 per user per month billed yearly. Notion suits teams whose centre of gravity is written knowledge, with task databases as a secondary use, at $10 per seat per month for Plus. Both prices were checked on 21 August 2026, and each is weaker at the other's job. - **ClickUp Unlimited:** $7 / user / mo, yearly - **Notion Plus:** $10 / seat / mo - **Free tiers:** ClickUp unlimited members · Notion 1,000 blocks with 2+ - **Checked:** 21 August 2026 ## The verdict Pick ClickUp if the daily question in your team is what am I doing today. It is cheaper at the entry tier, it has real task management with dependencies, recurring work, automations, time tracking and sprint tooling, and ClickUp Docs is a competent wiki that stops many teams needing a second subscription. Pick Notion if the daily question is where is the thing we decided. Notion's editor is better by a distance, its page nesting makes a real knowledge base, and its databases are enough for a roadmap and a task list if execution rigour is not what your team is missing. The failure that follows from getting this backwards is predictable in each direction. Teams that pick Notion for execution end up with a task database nobody updates, because Notion never nags anyone and has no notion of a sprint. Teams that pick ClickUp for knowledge end up with documentation people avoid writing, because the editor is fine and fine is not enough to make anyone want to write. ## Head to head Read from clickup.com/pricing and notion.com/pricing on 21 August 2026. | What decides it | ClickUp | Notion | | --- | --- | --- | | Entry paid price | Unlimited $7 per user per month yearly, $10 monthly | Plus $10 per seat per month, up to 20% off yearly | | Next tier up | Business $12 yearly, $19 monthly | Business $20 per seat per month | | Free tier | Free Forever, unlimited members with feature limits | Unlimited pages solo; 1,000-block cap once a second member joins | | Task management | Dependencies, recurring tasks, sprints, time tracking, workload | A database with a status property and no execution mechanics | | Writing | ClickUp Docs, competent and improving | The reason people choose Notion | | Automations | Included from the entry tier with rising limits | Light database automations on paid tiers | | AI pricing | Brain add-on $9 per user per month, Everything AI $28 | Included on Business and Enterprise; Custom Agents $10 per 1,000 credits | | Configurability | Very high, and it needs an owner | Moderate, and it needs a librarian | | Best fit | Teams that need work to move | Teams that need knowledge to persist | ## The honest gaps Each product's weakness is the other's headline. **What ClickUp does that Notion does not** - Sprints, velocity and burndown for a team that runs iterations - Time tracking inside the task rather than in a separate tool - Workload and capacity views across a team - Automations that fire on status change without a third-party service - A free tier with unlimited members and no block cap **What Notion does that ClickUp does not** - An editor good enough that people write documentation voluntarily - Page nesting deep enough to hold a real company handbook - One-click publishing of a page to the public web - Far more templates and third-party integrations than ClickUp has - Databases that are pleasant to read rather than merely functional > **Compare the tier you will really run** > > ClickUp Business at $12 plus ClickUp Brain at $9 is $21 per user per month. Notion Business at $20 includes AI features but bills Custom Agents separately at $10 per 1,000 credits. Neither headline price survives contact with a team that wants AI, so price the combination or the comparison is meaningless. ## Who should pick which A twenty-person software team with a Notion wiki they like and a task problem should add a tracker rather than move the wiki. ClickUp can hold both, but the migration cost of a well-loved Notion workspace usually exceeds the benefit of consolidation. A twenty-person agency starting from nothing should pick ClickUp, use its Docs for the handful of documents that matter, and revisit in a year. One subscription with adequate documentation beats two subscriptions with excellent documentation when nobody has written the documentation yet. A team that has both already and wants to cut one should count what would break. In practice Notion loses less when it goes, because the tasks in it were probably not being updated anyway. ## A third option worth knowing about Both of these products are sold as the tool that ends tool sprawl, and both then price AI as an extra per-seat line on top of the seats you already bought. The result is a bill that scales with how many people might use the software, and no line anywhere that corresponds to work getting done. Polaris puts tasks, documents and team chat in one workspace and charges nothing for it. Unlimited humans, unlimited tasks, unlimited workstreams, unlimited documents, no seats and no tiers, with a Chief of Staff agent included in every organisation from the first sign-in. The bill is for delivered work: roughly two dollars per human-equivalent hour that an AI worker on your roster completes, estimated by an open formula from observable effort and logged job by job, so any line on the invoice can be challenged from the record that produced it. The mechanism is worth stating plainly because it is the difference. Assign a task to an AI worker and a real cloud machine wakes up for it, runs a live tool loop with web search, ticks the acceptance criteria on the task as it goes, and posts the finished work as a comment with any generated files attached. It never marks the task done. A person closes it. Polaris is in free public beta and has no customers yet. ## Questions people ask **Can ClickUp replace Notion as a wiki?** For most teams, yes at a functional level. ClickUp Docs supports nesting, rich formatting, templates and page relationships. The gap is qualitative: writing in ClickUp Docs is workmanlike rather than enjoyable, and documentation quality tends to track how much people want to write. **Can Notion replace ClickUp for project management?** For lightweight work, yes. For a team running sprints, tracking time or needing automations that fire on status change, no. Notion databases record state; they do not drive a process, and nothing in Notion will chase an overdue task. **Which is cheaper?** ClickUp at the entry tier: $7 per user per month billed yearly against Notion Plus at $10, as checked on 21 August 2026. Add ClickUp Brain at $9 per user per month and ClickUp becomes the more expensive option, so the answer depends entirely on whether you want the AI features. **Can you import Notion pages into ClickUp?** Yes. ClickUp publishes a Notion importer that brings across pages and databases, mapping databases to Lists. Formatting mostly survives; relations, rollups and synced blocks do not translate cleanly, so complex database structures need checking page by page after the import. **Do teams really run just one of these?** Many run both, and it is often the right call. The pattern that fails is using each for the other's strength: a Notion task database nobody updates, or a ClickUp doc tree nobody reads. Pick which tool owns execution and which owns knowledge, then be strict about it. ## Related - https://www.polarishq.co/alternatives/clickup - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/cost/clickup-pricing - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/compare/notion-vs-coda - https://www.polarishq.co/replace/notion-and-jira --- --- title: "Linear vs Height: Height shut down in September 2025" description: "Height shut down on 24 September 2025, so this is no longer a live choice. What Height did that Linear does not, and where its idea went. Checked August 2026." url: https://www.polarishq.co/compare/linear-vs-height section: Comparisons updated: 2026-08-21 --- # Linear vs Height One of these two products no longer exists. The idea behind it is the interesting part. ## The short answer Height shut down on 24 September 2025, six months after its founder announced the closure in March 2025, so Linear is the only live option of the two. Height's distinctive feature was autonomous project management: an embedded reasoning engine that triaged incoming bugs, de-duplicated the backlog and updated specs. Linear has agent features but does not replicate that model. - **Height status:** Shut down 24 Sept 2025 - **Announced:** March 2025, six months' notice - **Linear Basic:** $10 / user / mo, yearly - **Checked:** 21 August 2026 > **Height is gone** > > Height's founder Michaël Villar announced the shutdown in March 2025 and the service went dark on 24 September 2025, with six months given for teams to export and move. If you arrived here to choose between the two, the choice has been made for you. The rest of this page is about what Height was trying to do and where that idea now lives. ## What Height was, and why people liked it Height was founded in 2018 by Michaël Villar, previously at Stripe, and raised around $18 million including a Series A led by Redpoint. For most of its life it was a well-designed, fast issue tracker in the same competitive set as Linear, with a chat-first interface that made task discussion feel less like filing and more like talking. In October 2024 it relaunched as Height 2.0 and made a much larger bet, describing itself as an autonomous project collaboration tool. The pitch was that project management is mostly repetitive judgement: triaging incoming bugs, spotting duplicates, chasing stale tasks, keeping specs aligned with what the code actually does. Height put a reasoning engine behind those jobs and gave every plan, including the free one, access to it. It was a genuinely interesting product and it did not find a market fast enough. That is worth saying plainly rather than glossing, because the failure carries information: teams liked the idea of an agent that tidies the backlog, and not enough of them would change tools for it. Tidying is not the expensive part of the work. ## How the two compared, for the record Linear figures read from linear.app/pricing on 21 August 2026. Height figures reflect its pricing at the time of shutdown and are historical. | What decided it | Linear | Height (until Sept 2025) | | --- | --- | --- | | Status | Active and growing | Discontinued 24 September 2025 | | Entry paid price | Basic $10 per user per month, billed yearly | Team plan around $8.50 per member per month | | Free tier | 250 issues, 2 teams, unlimited members | Unlimited tasks and members, with AI features included | | Core idea | Remove friction so humans track work faster | Have the tool do the project management itself | | Interface | Keyboard-first, list and board | Chat-first, with tasks emerging from conversation | | AI | Agent platform available on all tiers | Reasoning engine for triage, de-duplication and spec upkeep | | Company | Independent, well funded | Wound down after roughly $18 million raised | ## What to do if you are still on a Height export Teams that left Height in 2025 mostly landed in one of four places, for different reasons. - **Linear, for the ones who wanted the speed** — The closest replacement for Height's craft and pace. You lose the autonomous triage entirely and gain Triage as a human-run inbox. - **Shortcut, for the ones who wanted the structure** — Epics, milestones and up to five custom workflows on the Team plan at $8.50 per user per month, which is close to what Height charged. - **ClickUp, for the ones who wanted everything in one place** — Cheapest of the group at $7 per user per month billed yearly, and the most work to keep tidy. - **Notion, for the ones whose tasks were really documents** — A common landing spot for smaller Height teams, and a mistake for anyone who was relying on the tracker to drive execution. ## A third option worth knowing about Height's thesis was that a project tool should do project management. The part it chose to automate was the housekeeping: triage, duplicates, stale specs. That is real work, and it is not the work anybody is short of. Nobody has ever missed a deadline because the backlog had duplicates in it. Polaris makes the same general bet in a different place. Instead of automating the maintenance of the record, it puts AI workers on the roster as teammates and gives them the work itself. Humans and agents are rows in the same members table, so assignment is identical for both. Assign a task to an AI worker and a real cloud machine wakes for it, runs a live tool loop with web search, ticks the acceptance criteria on the task, and posts the finished output as a comment with any files it produced. The machine never closes the task; a person does, and rates it. Worth saying given the subject of this page: an autonomous project tool has already failed in this market once, and Polaris is in free public beta with no customers, no revenue and no case studies. What exists is a working product, recorded demos including an unstaged machine session, and pricing that charges nothing for the software and roughly two dollars per human-equivalent hour of work actually delivered, itemised on a work log you can challenge. Judge it on that rather than on the category. ## Questions people ask **When did Height shut down?** Height stopped operating on 24 September 2025. The closure was announced by founder Michaël Villar in March 2025, giving customers roughly six months to export their data and migrate. The company had raised around $18 million, including a Series A led by Redpoint Ventures. **What was Height 2.0?** Height 2.0 launched in October 2024 and repositioned the product as an autonomous project collaboration tool. An embedded reasoning engine handled triage of incoming bugs, backlog de-duplication, spec updates as work changed, and surfacing blockers. Those capabilities were available on every plan, including the free one. **What is the closest replacement for Height?** Linear for teams that valued the speed and design, Shortcut for teams that valued the structure and price. Neither reproduces Height's autonomous triage. Linear's agent platform is the nearest thing in a live product, and it works differently. **Can you still export data from Height?** No. The service went dark on 24 September 2025 and the export window closed with it. Teams that saved a CSV or JSON export during the notice period can still import that file into Linear, Shortcut or ClickUp, all of which accept CSV. **Does Linear have autonomous project management?** Not in Height's sense. Linear has an agent platform available across tiers and Triage Intelligence on the Business plan, which assists with routing incoming issues. The human remains the one grooming the backlog and deciding what matters, which is the opposite of the position Height took. ## Related - https://www.polarishq.co/alternatives/height - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/compare/linear-vs-shortcut - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/agentic-project-management --- --- title: "Trello vs monday.com: a board, or a board that computes" description: "Trello is half the price at every tier and free for ten collaborators. monday.com turns a board into a database with automations. Seat blocks checked Aug 2026." url: https://www.polarishq.co/compare/trello-vs-monday section: Comparisons updated: 2026-08-21 --- # Trello vs monday.com Both put cards in columns. Only one of them expects the columns to do arithmetic. ## The short answer Trello suits teams that want a shared board and nothing more, and it is roughly half the price of monday.com at every tier, with a free plan for up to ten collaborators. monday.com suits teams that need boards to behave like databases with typed columns, automations and dashboards, from $9 per seat per month with a three-seat minimum. Prices checked 21 August 2026. - **Trello Standard:** $5 / user / mo, annual - **monday.com Basic:** $9 / seat / mo, annual - **Free tiers:** Trello 10 collaborators · monday 2 seats - **Checked:** 21 August 2026 ## The verdict Trello wins for any team whose work fits on a board and whose main requirement is that everyone actually uses it. Adoption is effectively free, nothing needs configuring, and the free plan takes ten collaborators per workspace against monday.com's two seats. At $5 per user per month for Standard it is also the cheapest credible option in this category. monday.com wins the moment the board needs to compute something. Typed columns, formulas, status-driven automations, dashboards built from board widgets, and unlimited free viewers so stakeholders can watch without a seat. That is a different product doing a different job, and Trello cannot be extended into it with Power-Ups without producing something worse than either. The price gap is real and it widens because of how monday.com sells seats. Paid plans start at three seats and seats come in blocks of 3, 5, 10, 15 and 20, so a team of six buys ten. Six people on monday.com Standard is $120 a month. Six people on Trello Standard is $30. ## Head to head Read from trello.com/pricing and monday.com/pricing on 21 August 2026. | What decides it | Trello | monday.com | | --- | --- | --- | | Free tier | Up to 10 collaborators per workspace, 10 boards | Up to 2 seats, 3 boards, 3 docs | | Entry paid price | Standard $5 per user per month annually, $6 monthly | Basic $9 per seat per month annually | | Next tier up | Premium $10 annually, $12.50 monthly | Standard $12, then Pro $19 per seat per month | | Seat minimum | None | Three seats, sold in blocks of 3, 5, 10, 15, 20 | | Column model | Cards with labels, members, dates and custom fields | Typed columns including formula, mirror, dependency and status | | Automation | Butler on every tier, with rising limits | 250 actions per month on Standard, 25,000 on Pro | | Views beyond board | Calendar, timeline, table, dashboard and map on Premium | Table, kanban, calendar, timeline and Gantt by tier | | Viewers | Board members count toward the collaborator limit | Unlimited free viewers on paid plans | | Best fit | Small teams and simple pipelines | Operations teams that need reporting off the board | ## Six people, per month Annual list rates, August 2026. monday.com rounds six people up to a ten-seat block. - **$30** — Trello Standard. $5 per user, six users, no minimum. - **$120** — monday.com Standard. $12 per seat, ten seats billed for six people. - **$0** — Trello Free. Ten collaborators per workspace, ten boards. monday.com's free plan stops at two seats. ## What each is genuinely good at **Trello** - Adoption with no training, including from people who resist software - A free tier generous enough to run a real small team indefinitely - Butler automation available on every plan, including free - Power-Ups for the specific thing you need without paying for a platform - Half the price of monday.com at the equivalent tier **monday.com** - Boards that replace operational spreadsheets outright - Automations that a non-technical operations lead can build unaided - Dashboards that report across boards without exporting anything - Unlimited free viewers, so stakeholders never consume a seat - Adjacent CRM and dev products if the company later standardises on one vendor ## Who should pick which A seven-person studio tracking client work through five stages should use Trello, stay free or pay $35 a month, and put the difference somewhere useful. A thirty-person operations team tracking two hundred vendor onboardings with SLAs, owners and a weekly report should use monday.com. Trello will hold the cards and will not answer the question the report is asking. The pattern to avoid is buying monday.com for a Trello-shaped problem because someone liked the demo. The demo is genuinely excellent and it is showing you a database, which you may not need. ## A third option worth knowing about Both of these bill per person for a place to put cards, and monday.com bills for people who do not exist because seats come in blocks. If the reason you are comparing them is the invoice rather than the feature list, the third option is a different billing model rather than a cheaper board. Polaris is an all-in-one workspace where lanes are shared between list and board view, so the same work reads either way without maintaining two things. The software is free: unlimited humans, unlimited tasks, unlimited workstreams, unlimited documents, no seats, no tiers, no minimum and no block sizes. Documents and team chat are in the same product rather than in two more subscriptions. Billing starts only when an AI worker on your roster delivers work, at roughly two dollars per human-equivalent hour, estimated by an open formula from observable effort and logged job by job so any line can be challenged from the record. Polaris is in free public beta and has no customers yet. ## Questions people ask **Is Trello cheaper than monday.com?** Substantially. On 21 August 2026 Trello Standard listed at $5 per user per month annually against monday.com Basic at $9 per seat, and monday.com adds a three-seat minimum with block-based purchasing. For a six-person team the real gap was $30 against $120 a month at the comparable tiers. **Can you import Trello boards into monday.com?** Yes. monday.com publishes a Trello importer that maps boards to boards, lists to groups and cards to items. Attachments and comments transfer. Power-Up data and Butler automations do not, so any logic living in a Power-Up needs exporting separately before you move. **Does monday.com have a free plan?** Yes, capped at two seats with three boards and three docs. It is a trial in practice rather than a plan a team can live on. Trello's free plan allows up to ten collaborators per workspace, which is why small teams comparing free tiers land on Trello. **Can Trello do dashboards and reporting?** Trello Premium at $10 per user per month annually adds a dashboard view with basic charts across a board. Cross-board reporting is not native and needs a Power-Up. monday.com builds dashboards from widgets across multiple boards as a core feature, which is the main functional reason to pay the difference. **Which is better for a non-technical team?** Trello for adoption, monday.com for capability. Trello needs no explanation at all. monday.com needs an afternoon and someone willing to own the board structure, and in return it does things Trello will not do at any price. ## Related - https://www.polarishq.co/alternatives/trello - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/cost/trello-pricing - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/replace/monday-and-slack --- --- title: "Wrike vs Asana: proofing and approvals, or adoption" description: "Wrike caps its $10 Team plan at 15 users, then jumps to $25. Asana has no cliff but no proofing either. What that means for creative teams. Checked August 2026." url: https://www.polarishq.co/compare/wrike-vs-asana section: Comparisons updated: 2026-08-21 --- # Wrike vs Asana Two work platforms priced almost identically, aimed at two different departments. ## The short answer Wrike suits creative and professional services teams that need proofing, approvals, request forms and resource management in one tool, at $10 per user per month for Team and $25 for Business as checked on 21 August 2026. Asana suits cross-functional teams that value adoption and goal reporting, at $10.99 for Starter and $24.99 for Advanced billed annually. Wrike's Team plan caps at fifteen users. - **Wrike Team:** $10 / user / mo, 2 to 15 users - **Wrike Business:** $25 / user / mo, 5 to 200 users - **Asana Starter:** $10.99 / user / mo, annual - **Checked:** 21 August 2026 ## The verdict Choose Wrike if your work involves creative assets that need reviewing. Proofing with annotations directly on images, video and PDFs, approval workflows with named reviewers, and request forms with conditional logic that route intake to the right team are all native, and Asana has no real equivalent. For an in-house creative team or an agency running client review cycles, that is the entire decision. Choose Asana if the work is programmes rather than assets, and if adoption is a live risk. Asana is easier to learn by a clear margin, its goals and portfolio rollup give an executive view without configuration, and people who have never used a work platform can be productive in it within an hour. Wrike is more capable and more effortful, and the effort is where deployments fail. The headline prices are close enough to be a distraction. What is not close is Wrike's tier cap: the Team plan at $10 per user per month is limited to fifteen users, and the next plan up is $25. Growing from fifteen people to sixteen does not add one seat's cost, it multiplies the per-seat price by two and a half. Asana has no equivalent cliff. ## Head to head Read from wrike.com/price and asana.com/pricing on 21 August 2026. | What decides it | Wrike | Asana | | --- | --- | --- | | Entry paid price | Team $10 per user per month, 2 to 15 users | Starter $10.99 annually, $13.49 monthly | | Next tier up | Business $25 per user per month, 5 to 200 users | Advanced $24.99 annually, $30.49 monthly | | Tier caps | Team stops at 15 users, forcing the jump to Business | None | | How seats are sold | In groups: blocks of 5 up to 30 seats, 10 up to 100, 25 above that | Individually | | Free tier | Free plan with basic task management | Personal, up to 2 users for workspaces created after 12 Nov 2025 | | Proofing and approvals | Native annotation on images, video and PDFs, with approval routing | Not native; handled through comments or a separate tool | | Request intake | Custom request forms with conditional logic that create and route work | Forms that create tasks, without conditional routing | | Resource management | Native workload and capacity planning | Workload view on Advanced and above | | Learning curve | Days, and a deployment usually needs an owner | About an hour | > **The fifteen-user cliff** > > Fifteen people on Wrike Team is $150 a month. Sixteen people cannot stay on Team and land on Business at $25, which is $400 a month for one extra hire. Sixteen people on Asana Starter is $175.84. If you are between twelve and twenty people and growing, model the next twelve months before signing. ## The capabilities that decide it **Wrike** - Proofing with annotations on images, video and PDFs inside the task - Approval workflows with named reviewers and an audit of who approved what - Request forms with conditional logic that route work to the right team - Custom item types, so a campaign and a bug are different shapes of thing - Native resource and capacity planning rather than a view bolted on top **Asana** - Adoption without training, which is the most common cause of failed rollouts - Goals that roll into portfolios for a leadership view with no setup - One task visible in multiple projects without duplication - No tier cap forcing a price jump as the team grows - A much larger library of integrations and templates ## Three questions that settle it - **Does someone review visual work and mark it up?** — If yes, Wrike, and stop comparing. Proofing is the clearest functional gap between these two products and working around it means buying a third tool. - **How many people, and how many in a year?** — If you will cross fifteen within the contract term, Wrike Team is not the price you will pay. Model Business at $25 instead and compare that against Asana Advanced at $24.99. - **Has a previous rollout failed on adoption?** — If yes, Asana. Wrike's capability advantage is worth nothing in a workspace half the team avoids, and Wrike is the harder of the two to love. ## A third option worth knowing about Both of these platforms are priced on how many people might open them, and both have tier structures designed so that a growing team pays more per person over time. The work itself is unchanged by any of it. Polaris takes the opposite position. The software is free with no seats, no tiers and no caps: unlimited humans, unlimited tasks, unlimited workstreams and unlimited documents, with tasks, docs and team chat in one product. The bill is for work delivered by AI workers on your roster, at roughly two dollars per human-equivalent hour, estimated by an open formula from observable effort and logged job by job on the worker's work log, so any line on the invoice can be disputed from the record that produced it. For a services team the relevant part is what arrives. Assign a task to an AI worker and a real cloud machine wakes for it, works after you close your laptop, ticks the acceptance criteria on the task, and delivers as a comment with generated files attached. A person reviews it and closes it. Polaris is in free public beta with no customers yet, and it has no proofing or annotation feature, so a creative review workflow is not something it currently replaces. ## Questions people ask **Does Wrike's Team plan really stop at 15 users?** Yes. As listed on 21 August 2026, Wrike Team covers 2 to 15 users at $10 per user per month, and Business covers 5 to 200 users at $25. There is no intermediate tier, so a sixteenth hire moves the whole team to the higher rate. **Does Asana have proofing or annotation?** Not natively. Asana supports image and PDF attachments with comments, and some teams use pinned comments as a review record, but there is no annotation layer, no approval routing and no audit of who signed off. Creative teams that need this pair Asana with a dedicated proofing tool. **Can you migrate from Asana to Wrike?** Yes. Wrike provides an Asana importer covering projects, tasks, assignees, dates and attachments. Custom fields transfer with mapping. Dependencies, rules and multi-homed tasks need attention, because Wrike models folders and projects differently from Asana's project structure. **Which is better for an agency?** Wrike, in most cases, because agency work involves client review cycles, intake forms and billable capacity planning, and all three are native. The caveat is adoption: Wrike takes real setup, so it suits agencies with an operations lead and suits ones without a lot less. **How does Wrike sell seats?** In groups rather than individually. Wrike sells in blocks of five for accounts up to thirty seats, blocks of ten between thirty and a hundred, and blocks of twenty-five above that. A team of seventeen buys twenty seats, which is worth including in any comparison against Asana's per-user billing. ## Related - https://www.polarishq.co/alternatives/wrike - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/jira-vs-asana - https://www.polarishq.co/compare/basecamp-vs-asana - https://www.polarishq.co/cost/stack-cost-50-person-team - https://www.polarishq.co/glossary/per-seat-pricing --- --- title: "Todoist vs Asana: a personal list against a team system" description: "Todoist Business is $8 per user and enough for a team whose work is lists. Asana is $10.99 and models the plan. Where the line sits. Prices checked August 2026." url: https://www.polarishq.co/compare/todoist-vs-asana section: Comparisons updated: 2026-08-21 --- # Todoist vs Asana The question is not which is better. It is whether anyone besides the assignee needs to see the shape of the work. ## The short answer Todoist suits individuals and small teams whose work is a list of next actions, at $5 per month for Pro and $8 per user per month for Business, both billed annually. Asana suits teams where someone other than the assignee needs to see structure: dependencies, timelines, workload and goals, from $10.99 per user per month billed annually. Prices checked 21 August 2026. - **Todoist Pro:** $5 / mo, billed annually - **Todoist Business:** $8 / user / mo, annual - **Asana Starter:** $10.99 / user / mo, annual - **Checked:** 21 August 2026 ## The verdict Todoist is the better tool if the work belongs to individuals. It is the best pure task capture product in this comparison: natural-language input that parses dates as you type, keyboard and mobile ergonomics that make adding a task cost nothing, and a structure simple enough that people stay in it for years. Business at $8 per user per month is cheaper than Asana Starter at $10.99 and enough for a small team running shared project lists. Asana is the better tool the moment somebody who is not doing the work needs to understand it. A dependency chain, a timeline that shifts when a date moves, a workload view that shows who is over-committed, a goal that rolls up to a portfolio. Todoist has none of that and does not pretend to. Adding it would ruin the thing Todoist is good at. The practical line is a question about meetings. If your team currently holds a weekly meeting to work out what depends on what, Asana replaces part of that meeting. If your team's weekly meeting is people reading out what they did, Todoist is enough and Asana will be an expensive way to keep the same meeting. ## Head to head Read from todoist.com/pricing and asana.com/pricing on 21 August 2026. Todoist raised Pro pricing in December 2025. | What decides it | Todoist | Asana | | --- | --- | --- | | Entry paid price | Pro $5 per month billed annually, $7 monthly | Starter $10.99 per user per month annually, $13.49 monthly | | Team tier | Business $8 per user per month annually, $10 monthly | Advanced $24.99 annually, $30.49 monthly | | Free tier | 5 personal projects, 5 collaborators per project, 1 week of history | Personal, up to 2 users for workspaces created after 12 Nov 2025 | | Task capture | Natural language input that parses dates while typing | Standard form fields | | Dependencies | None | Native, with dates that shift downstream work | | Views | List, board and calendar layouts | List, board, timeline and calendar from Starter | | Reporting | Productivity trends and activity history | Goals, portfolios, workload, universal reporting on Advanced | | Team scale | Up to 1,000 members and 500 team projects on Business | No practical ceiling | | Best fit | Individuals and small teams running lists | Teams running plans that other people depend on | ## What each is genuinely best at - **Todoist: capture speed** — Typing a task with a due date, project and priority takes one line and no clicks. That sounds small and it is why people who have used Todoist for a decade are still using it. Tasks that cost nothing to file are tasks that get filed. - **Todoist: staying out of the way** — There is very little to configure, so there is very little to maintain. No workspace design, no field taxonomy, no rules to audit a year later. - **Asana: making a plan legible to people who are not in it** — The timeline, the dependency arrows and the workload view exist so somebody can look at a project and understand it without asking. That is the job Todoist does not do. - **Asana: accountability without nagging** — Every task has an owner and a date by default, and the product surfaces what is overdue to the people who need to know. Todoist tells you; Asana tells your team. ## Who should pick which A four-person consultancy where everyone knows what they are doing and needs a shared list per client should use Todoist Business at $8 a head, which is $32 a month against Asana Starter's $43.96. A four-person team delivering a fixed-date launch where the designer is blocked on the copywriter should use Asana. The dependency is the whole problem, and no amount of list discipline resolves it. A common and sensible arrangement is both, unofficially: Asana for the team's projects and Todoist for personal next actions, because people trust their own list more than a shared one. That is not tool sprawl, it is one paid seat and one $5 subscription, and it works. ## A third option worth knowing about The reason people keep a personal list alongside a team tool is that team tools are organised around projects and a person's day is not. What you need on Monday morning is not a project, it is the three things that cannot slip. Polaris is built with that as the home screen. Focus is a time-based lane of non-negotiables across three horizons, today, this week and the next thirty days, and it is pinned first in every view rather than being one tab among many. Workstreams hold the project-shaped work behind it, with lanes shared between list and board view so the same work reads either way. Alongside that, AI workers sit on the roster next to people. Assign one a task the same way you assign a colleague, and a cloud machine wakes for it and keeps working after your laptop closes, then delivers the result as a comment on the task with any files attached. You close it. The software is free with no seats or tiers; the bill is roughly two dollars per human-equivalent hour delivered, itemised on a work log you can challenge line by line. Polaris is in free public beta and has no customers yet. ## Questions people ask **How much does Todoist cost?** As checked on 21 August 2026: the Beginner tier is free with five personal projects and five collaborators per project, Pro is $5 per month billed annually or $7 monthly, and Business is $8 per user per month billed annually or $10 monthly. Todoist raised Pro pricing from $4 to $5 in December 2025. **Can Todoist replace Asana for a team?** For a small team whose work is genuinely a set of lists, yes, and at a lower price. It cannot replace Asana where work has dependencies, needs a timeline, or has to be legible to someone outside the team, because Todoist has no model for any of those things. **Can you import Todoist tasks into Asana?** Yes, via CSV. Todoist exports projects as CSV or as a template file, and Asana imports CSV with column mapping for assignee, due date and section. Sub-tasks, recurring rules and natural-language date patterns do not survive the trip and need setting up again in Asana. **Does Todoist have a free plan for teams?** The free Beginner tier allows five personal projects with up to five collaborators each, which supports very light shared work. Team features such as a shared workspace, team roles, shared templates and centralised billing require the Business plan at $8 per user per month billed annually. **Is Todoist good for project management?** It is good for task management, which is a different job. There is no dependency model, no timeline, no capacity planning and no portfolio view. Teams that describe themselves as needing project management usually need at least one of those, and that is the point at which Todoist stops being the answer. ## Related - https://www.polarishq.co/alternatives/todoist - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/compare/basecamp-vs-asana - https://www.polarishq.co/compare/trello-vs-monday - https://www.polarishq.co/cost/stack-cost-5-person-team - https://www.polarishq.co/glossary/focus-lane --- --- title: "Replace your stack: 18 tool-by-tool migration plans" description: "Eighteen migration plans for real tool combinations. What imports, what has to be rebuilt by hand, what you should keep and connect instead of moving." url: https://www.polarishq.co/replace section: Replace your stack updated: 2026-08-21 --- # Replacing the stack you already pay for Every plan here says which of your tools has an importer, which has a connector, and which has neither. ## The short answer Replacing a work stack means moving docs, tasks and team chat out of separate per-seat products into one workspace. Polaris holds all three for free and adds AI workers billed at roughly two dollars per human-hour delivered. Notion, Slack, Linear and GitHub can be connected instead of migrated. Jira, Confluence, Asana, Trello, ClickUp, Monday and Airtable have no connector, so those need an export and a rebuild. - **Combinations covered:** 18 - **Tools with an importer:** Notion only - **Tools you can connect instead:** Slack, Notion, Linear, GitHub - **Polaris software:** $0, no seats ## Why these pages exist Most migration content is written by the company that wants you to migrate, which is why it never mentions the parts that do not move. These pages do. Each one names the specific handoffs your combination is losing work in, then tells you which parts of the move are an export, which are a rebuild by hand, and which are not worth doing at all. The set is organised by what teams actually run rather than by tool. A team on Confluence and Jira has admins, space permissions and a compliance answer to give. A team on Trello and Slack has none of that and a much smaller bill. The same advice would be wrong for both. ## The two honest verdicts Not every combination should be replaced wholesale, and saying so is the point. **Connect, keep, retire the rest** - Your stack is mostly Notion, Slack, Linear or GitHub - Those four are in the Polaris connection catalog, authorised once for the whole org - AI workers read them through the connection with no export at all - You retire the per-seat products that are not in that list first **Export and rebuild** - Your tracker is Jira, Asana, Trello, ClickUp, Monday or Airtable - None has a Polaris importer and none is in the connection catalog - The realistic move is a CSV export plus a rebuild of the lanes you still use - Most teams find the rebuild smaller than expected because half the board was dead ## What every plan on this page covers - **The seam** — The specific handoff between your tools where work goes missing. Between a Notion spec and a Jira ticket it is the decision that changed after the spec was written. Between anything and Slack it is scrollback. - **The three-way sort** — What consolidates cleanly, what arrives with something missing, and what does not transfer at all. Relational data, automation rules and permission models are usually in the third category. - **The staged plan** — Stages you can stop after. Every plan puts new work in Polaris before it touches history, so an abandoned migration still leaves you somewhere sensible. - **The bill, before and after** — Per-seat subscriptions multiplied by headcount today, against free software plus metered delivery afterwards. The arithmetic lives on the cost pages, which quote real list prices with the date they were checked. > **What Polaris can actually import today** > > One thing. Notion pages, through an integration secret or a stored Notion connection, up to twenty-five selected pages per run and three levels of sub-pages, capped at sixty pages and fifteen hundred blocks. Every other tool in this cluster is a CSV export and a rebuild. Any page here that claimed otherwise would be found out in the first hour of a real migration. ## Every stack combination - [Replacing Notion and Jira](https://www.polarishq.co/replace/notion-and-jira) — The spec lives in one product, the work lives in another, and the two stop agreeing about four days in. - [Replacing Notion and Slack](https://www.polarishq.co/replace/notion-and-slack) — A stack with a writing tool and a talking tool, and no place where a piece of work has an owner and a date. - [Replacing Jira and Slack](https://www.polarishq.co/replace/jira-and-slack) — Engineering teams on Jira and Slack have a tracker, a chat tool, and specs living in ticket descriptions. - [Replacing Linear and Notion](https://www.polarishq.co/replace/linear-and-notion) — Two tools people choose deliberately, and a bill that is small enough that saving money is not the argument. - [Replacing Linear and Slack](https://www.polarishq.co/replace/linear-and-slack) — Issues in one place, everything anyone said about those issues in the other, and nothing written down. - [Replacing Trello and Slack](https://www.polarishq.co/replace/trello-and-slack) — A board that everyone can read and nobody maintains, plus a channel where the real assignments happen. - [Replacing Asana and Slack](https://www.polarishq.co/replace/asana-and-slack) — Rules that fire into Slack, replies that never come back, and a portfolio view built for people who do not do the work. - [Replacing Confluence and Jira](https://www.polarishq.co/replace/confluence-and-jira) — An Atlassian estate has administrators, permission schemes and an auditor, and none of those are content problems. - [Replacing Notion, Linear and Slack](https://www.polarishq.co/replace/notion-and-linear-and-slack) — The default startup stack, three subscriptions, and a decision that has to be written down three times to stay true. - [Replacing Notion, Jira and Slack](https://www.polarishq.co/replace/notion-and-jira-and-slack) — Two tools you can connect this afternoon and one that is a project with a start date and an owner. - [Replacing ClickUp and Slack](https://www.polarishq.co/replace/clickup-and-slack) — You already bought one all-in-one product. The Slack subscription next to it is the interesting fact. - [Replacing Monday.com and Slack](https://www.polarishq.co/replace/monday-and-slack) — A board configured by the person who reports on the work, used by the people who have to do it. - [Replacing Notion and Trello](https://www.polarishq.co/replace/notion-and-trello) — A page describing the project and a board describing the same project, kept in step by somebody's memory. - [Replacing Airtable and Slack](https://www.polarishq.co/replace/airtable-and-slack) — A relational base doing three jobs at once, and only one of those jobs should move. - [Replacing Confluence and Notion](https://www.polarishq.co/replace/confluence-and-notion) — Two subscriptions for writing things down, which almost always means one migration that never got finished. - [Replacing Jira, Confluence and Slack](https://www.polarishq.co/replace/jira-and-confluence-and-slack) — An Atlassian estate with a chat tool bolted to it, and exactly one piece of that you can move this quarter. - [Replacing Linear, Notion and GitHub](https://www.polarishq.co/replace/linear-and-notion-and-github) — Three tools a technical team chose on purpose, and none of them has to move anywhere. - [Replacing Asana and Notion](https://www.polarishq.co/replace/asana-and-notion) — A project brief in the tracker, a project page in the doc tool, and a client who has read one of them. ## The numbers behind the plans - [cost](https://www.polarishq.co/cost) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [alternatives](https://www.polarishq.co/alternatives) - [integrations](https://www.polarishq.co/integrations) ## Questions people ask **Do I have to move everything at once?** No, and no plan in this cluster recommends it. Every staged migration here starts by pointing new work at Polaris while the old tools stay open in read-only mode. Teams that stall halfway still end up with a working setup rather than a half-broken one. **Which tools can Polaris read without a migration?** Slack, Notion, Linear and GitHub are in the connection catalog, along with Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web search. Connections are authorised once for the whole organisation and stored server-side, so AI workers can use them while browsers cannot read the credentials back. **What happens to years of history in the old tracker?** It stays where it is. Every plan here recommends leaving the incumbent readable rather than trying to reproduce closed tickets from four years ago. Export the archive to CSV for your records, move the work that is still live, and stop paying for seats once nobody needs write access. **Is Polaris actually free, or does the price arrive later?** The software is free with no seat count and no tier. Unlimited humans, tasks, workstreams and docs are included, and the Chief of Staff is in every organisation from the first sign-in. Billing starts only when an AI worker delivers finished work, at roughly two dollars per human-equivalent hour, itemised on a work log you can challenge line by line. **How mature is Polaris?** It is in free public beta. There is a working product, a mobile app and four recorded demos including an unstaged machine session, but no customers to point at and no case studies. A migration plan from a beta product deserves more scepticism than one from an established vendor, which is why these pages state their limits in detail. ## Related - https://www.polarishq.co/alternatives - https://www.polarishq.co/cost - https://www.polarishq.co/integrations - https://www.polarishq.co/compare - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cloud-claude-code --- --- title: "Replace Notion and Jira: a real migration plan" description: "What actually moves when you consolidate Notion and Jira. Notion pages import with hard caps, Jira issues do not, and the spec-to-ticket seam closes." url: https://www.polarishq.co/replace/notion-and-jira section: Replace your stack updated: 2026-08-21 --- # Replacing Notion and Jira The spec lives in one product, the work lives in another, and the two stop agreeing about four days in. ## The short answer Consolidating Notion and Jira moves two things into one workspace: the written spec and the ticket that implements it. Polaris imports Notion pages directly, capped at twenty-five top-level pages and fifteen hundred blocks per run. Jira has no importer and no connector, so issues are exported to CSV and rebuilt as tasks, or the current sprint is run to zero before the switch. - **Notion import:** 25 pages, 1500 blocks per run - **Jira importer:** None - **Polaris software:** $0, unlimited people - **Billing:** ~$2 per human-hour delivered ## The seam nobody writes down Notion holds the spec. Jira holds the ticket. The ticket description is a paraphrase of the spec written by whoever groomed the backlog that week, and it stops being accurate the first time someone changes their mind in a meeting. Nothing updates the Notion page, because updating it is nobody's job. So the team ends up with two documents that disagree, and the tiebreaker is whoever remembers the conversation. Engineers build the ticket. Product reads the page. The gap between them is where rework comes from, and it is invisible on both dashboards because both tools think they are healthy. That gap is the actual reason to consolidate. Bill savings are real, but a team that closes the spec-to-ticket seam and keeps paying for two products has still won something. ## What moves, what arrives incomplete, what stays put Sorted by how much manual work each row costs you. | What you have now | Where it ends up | How clean | | --- | --- | --- | | Notion pages and sub-pages | Polaris Docs, as a nested tree | Imports directly, within the caps | | Notion databases used as trackers | Rebuilt as workstreams and lanes | No transfer. Properties do not come across | | Jira issues, live sprint | Tasks in a workstream, retyped or CSV-mapped | Manual. Usually two hours for a live sprint | | Jira workflows, schemes, screens | Nothing equivalent | Does not move. Most teams do not miss them | | Jira issue history and closed epics | Left in Jira, read-only | Do not attempt. Export for the archive instead | | Sprint reports and velocity charts | Not reproduced | Genuinely lost. Say so before you commit | ## What the Notion importer really does Taken from the shipped function, not from a roadmap. - **It lists what the integration can see** — You paste an integration secret or connect Notion once for the whole organisation. The picker shows every page shared with that integration and preselects the top-level ones. - **It carries text, structure and to-dos** — Paragraphs, three heading levels, bulleted and numbered lists, to-dos with their checked state, quotes, code, callouts, dividers and toggles all arrive. Sub-pages come along to three levels deep and become nested docs. - **It skips media and tables, and counts what it skipped** — Images, files, embeds, tables, databases, synced blocks and column layouts are dropped, and the import screen tells you how many blocks it dropped. Numbered lists arrive as bullets. Toggles arrive as bullets and stop folding. - **It has hard stops** — Twenty-five selected pages per run, sixty pages and fifteen hundred blocks total, four levels of indent, four thousand characters per block. A large workspace needs several runs, and page properties, comments and version history never come across. ## A four-stage migration Stage one takes an afternoon. Nothing after it is urgent. 1. **Import the docs that are still true** — Do not import the whole Notion workspace. Pick the pages your team opened in the last quarter, run the import, and read the skipped-block count. If it is high, those pages were mostly screenshots and tables, which tells you something about how much of your documentation was actually text. 2. **Rebuild the live sprint by hand** — Export the current sprint from Jira to CSV, then create the tasks in a Polaris workstream with owners and dates set as you go. Retyping thirty tickets takes about two hours and produces a backlog with no dead items in it, which a CSV import would not. 3. **Run one sprint in parallel** — Keep Jira open and read-only for one full cycle. Product writes in Polaris Docs, engineering works the Polaris tasks, and nobody grooms two backlogs. If the team drifts back to Jira, you have learned something cheap. 4. **Hire the worker that closes the seam** — Assign an AI worker the standing job of reading the doc and the task and flagging where they disagree. A cloud machine wakes per task, works the tool loop, and delivers the difference as a comment with a file. You close the task; the machine never does. 5. **Cancel the Jira seats, not the Jira account** — Drop write seats once nobody needs them and keep the archive readable if your retention policy asks for it. Cancelling the account is the step teams regret, and it saves almost nothing compared to dropping seats. > **When to stay on Jira** > > If your release process depends on Jira workflow transitions, if auditors read your issue history, or if a Jira automation gates a deploy, this migration is a project and not an afternoon. Polaris has no workflow engine and no equivalent of screens and schemes. Move the docs, connect what you can, and leave the tracker alone until that dependency is gone. ## Read next - [alternatives/jira](https://www.polarishq.co/alternatives/jira) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [replace/notion-and-jira-and-slack](https://www.polarishq.co/replace/notion-and-jira-and-slack) - [replace/confluence-and-jira](https://www.polarishq.co/replace/confluence-and-jira) ## Questions people ask **Can Polaris import Jira issues directly?** No. There is no Jira importer and Jira is not in the connection catalog. The realistic path is a CSV export from Jira for your archive, plus a manual rebuild of the sprint or backlog that is still live. Teams usually find the manual rebuild useful because it forces a decision on every stale ticket. **What happens to Notion databases when I import?** Database rows do not come across. The importer carries page content, headings, lists, to-dos, quotes, code, callouts and dividers, plus sub-pages three levels deep. A Notion database used as a project tracker has to be rebuilt as a Polaris workstream with lanes, which takes minutes but is not automatic. **How long does the whole move take for a fifteen-person product team?** Plan on one working day for the docs import and the sprint rebuild, then one full sprint in parallel before you cut over. The long pole is never the data. It is getting product managers to stop writing specs in the old place, which takes about two weeks of consistent nudging. **Do we lose our sprint velocity history?** Yes. Polaris does not reproduce Jira sprint reports, burndown charts or velocity metrics, and there is no import path for them. Export the charts you care about as PDFs before you drop seats. Most teams discover they were looking at velocity out of habit rather than to make decisions. **What does the bill look like afterwards?** Notion and Jira are both per-seat and both scale with headcount. Polaris software is free with no seat count, and the only charge is roughly two dollars per human-equivalent hour that an AI worker delivers, logged job by job. A team that assigns no agent work pays nothing at all. ## Related - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/replace/notion-and-jira-and-slack - https://www.polarishq.co/replace/linear-and-notion --- --- title: "Replace Notion and Slack: what to move, what to keep" description: "Notion and Slack are both in the Polaris connection catalog, so this migration is mostly not a migration. What to connect, what to import, what to add." url: https://www.polarishq.co/replace/notion-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Notion and Slack A stack with a writing tool and a talking tool, and no place where a piece of work has an owner and a date. ## The short answer Notion and Slack together give a team a place to write and a place to talk, but no place where work is owned and dated. Both are in the Polaris connection catalog, so neither has to be abandoned. The realistic move is to add the missing tracker, route Slack messages into an Inbox as prefilled task suggestions, and import the Notion pages that are still true. - **Tools needing export:** Neither - **What the stack is missing:** An owner and a date - **Polaris software:** $0, no seats ## This combination has a specific shape Teams land on Notion and Slack without a tracker for a good reason. For the first year, the work is obvious and the team is small enough that a Slack message counts as an assignment. Notion holds the plan, Slack holds the conversation, and everyone can see everything. It breaks in a recognisable way. Someone builds a Notion database with a Status column and calls it the tracker. It works for a month. Then the statuses go stale, because updating a database row is a separate act from doing the work, and the person who did the work has already moved on to the next thing. Meanwhile the decision that made half those rows obsolete is in a Slack thread from a Tuesday. ## The two seams in this stack - **The Notion database that is pretending to be a tracker** — Notion databases are excellent at holding structured information and poor at demanding an owner and a date. Rows accumulate without either. Nobody notices, because a stale row looks exactly like a live one. - **The decision that stayed in scrollback** — The message that changed the plan sat in a channel, got seven reactions and scrolled away. The Notion page still says the old thing. Anyone who was not online that afternoon is now working from a document that is quietly wrong. ## The numbers that decide this one - **14** — connections in the catalog. Slack and Notion are two of them - **~60s** — to hire an AI worker. an interview in chat, then a SKILL.md you can edit - **$0** — for the software. unlimited people, tasks, workstreams and docs - **~$2** — per human-hour delivered. estimated by an open formula, logged per job ## The plan, in three stages There is no export in any of them. 1. **Connect Slack and watch the Inbox fill** — Authorise Slack once for the organisation. Messages that look like work arrive in the Polaris Inbox as prefilled task suggestions with a bucket, a lane, labels and an owner already chosen. Nothing becomes a task without a click, which is the difference between a helpful capture tool and a system that quietly invents work. 2. **Import the Notion pages you actually open** — Run the importer against the pages your team touched this quarter rather than the whole workspace. Sub-pages come along three levels deep. Databases do not, so rebuild the one you were using as a tracker as a workstream with lanes, which is a fifteen-minute job and the point of the exercise. 3. **Put the non-negotiables in the Focus lane** — Focus is a time-based lane across three horizons: today, this week, the next thirty days. Everything the team must not drop goes there first. It is pinned at the top of every view, which is the structural answer to a stack where nothing had a date. > **The honest recommendation** > > You are not replacing Notion and Slack so much as adding the layer they never had. Connect both, keep writing in whichever place your team already trusts, and judge Polaris on one thing only: whether an AI worker you hired in a minute hands back work you would have paid a person to do. If it does not, you have lost nothing, because the software was free. ## Read next - [integrations/slack](https://www.polarishq.co/integrations/slack) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [cost/slack-pricing](https://www.polarishq.co/cost/slack-pricing) - [replace/notion-and-linear-and-slack](https://www.polarishq.co/replace/notion-and-linear-and-slack) - [replace/notion-and-trello](https://www.polarishq.co/replace/notion-and-trello) ## Questions people ask **Do I have to leave Slack to use Polaris?** No. Slack is in the connection catalog, authorised once for the whole organisation and stored server-side. Connecting it means Slack messages arrive in the Polaris Inbox as task suggestions and AI workers can read the channels they are given access to. Plenty of teams keep Slack for conversation and use Polaris for everything that needs an owner. **How much of my Notion workspace can I import at once?** Twenty-five selected pages per run, with sub-pages three levels deep, stopping at sixty pages or fifteen hundred blocks. A workspace bigger than that needs several runs. Text, headings, lists, to-dos, quotes, code, callouts and dividers arrive. Images, embeds, tables and databases are skipped, and the importer reports how many blocks it dropped. **Does a Slack message become a task automatically?** Never without a click. The Inbox shows a suggestion with the bucket, lane, labels and owner already filled in, and you approve or discard it. Signals become suggestions rather than silent tasks, because a capture system that creates work on its own gets muted within a week. **We have no tracker at all. Is that actually a problem?** It is a problem at the point where somebody has to ask in Slack what the status of something is. If nobody asks that question yet, you probably do not need this page. If it is asked twice a week, the Notion database with the stale Status column is already the workaround, and it is costing more attention than a real tracker would. ## Related - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/replace/notion-and-linear-and-slack - https://www.polarishq.co/replace/notion-and-trello --- --- title: "Replace Jira and Slack: an engineering migration plan" description: "A stack with a tracker and a chat tool and no docs layer. What a Jira export gives you, why Slack does not need replacing, and where the specs should live." url: https://www.polarishq.co/replace/jira-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Jira and Slack Engineering teams on Jira and Slack have a tracker, a chat tool, and specs living in ticket descriptions. ## The short answer Jira and Slack cover tracking and conversation but leave no home for written specification, so requirements end up in ticket descriptions and channel threads. Polaris adds a docs tree alongside tasks and chat. Slack is in the connection catalog and can stay. Jira has no importer, so live issues are rebuilt by hand and the closed history stays in Jira, read-only. - **Slack:** Connects, no export - **Jira:** No importer, manual rebuild - **What you gain:** A docs layer this stack never had - **Billing after:** ~$2 per human-hour delivered ## The ticket description is doing a job it is bad at In a Jira and Slack shop, the specification is whatever is in the ticket description on the day someone picks it up. That field was designed to describe one unit of work, and it is being asked to hold context, rationale, edge cases and the reason a previous approach was abandoned. So the rationale goes to Slack instead, where it is searchable in theory and gone in practice. Six months later somebody asks why the retry logic works the way it does, and the honest answer is that three people knew and two of them have left. The gap is not tracking. Jira tracks well. The gap is that this stack has nowhere durable to write, and no amount of ticket hygiene fixes that. ## Where each piece lands **Stays roughly as it is** - Slack, connected once for the organisation and left alone - GitHub, also in the connection catalog, so pull requests stay where they are - Closed Jira issues, readable in Jira with write seats dropped - Whatever CI and deploy tooling reads the repo rather than the tracker **Changes shape** - Live issues become tasks in a workstream, retyped rather than imported - Ticket descriptions get split: the work becomes a task, the reasoning becomes a doc - Jira automations that post to Slack are replaced by Inbox suggestions - Anything that depended on workflow transitions needs a different answer ## Migration in four stages 1. **Write down the three things everyone asks about** — Before moving any ticket, create three docs for the questions your team answers repeatedly in Slack. Deploy process, on-call expectations, and whichever part of the system nobody understands. This is the layer the stack was missing, and building it first proves the point faster than any tracker migration. 2. **Connect Slack and GitHub** — Both are in the connection catalog and authorised once, org-wide, with credentials stored server-side. Slack messages start arriving as prefilled task suggestions in the Inbox. AI workers you hire later can read both without you granting anything again. 3. **Move the current sprint, not the backlog** — Export the sprint from Jira to CSV for reference and create the tasks in Polaris by hand. Leave the backlog behind on purpose. If an item matters, someone will raise it again within a fortnight, and if nobody does, it was never going to be built. 4. **Give a worker the boring half of the sprint** — Hire an AI worker with GitHub, Slack and web search, and assign it release notes, dependency triage, or the bug reproduction nobody wants. A machine wakes per task, ticks its own acceptance criteria, and posts the work as a comment with files attached. An engineer reviews and closes it. ## The Jira features with no equivalent Check this list before you commit to anything. | Jira feature | Polaris equivalent | What to do | | --- | --- | --- | | Workflow schemes and transitions | None | Keep Jira for anything a deploy gate reads | | Custom fields and screens | Labels and lanes | Rebuild only the two or three fields you filter on | | JQL saved filters | None | Rebuild as lanes, which is coarser and usually enough | | Sprint and velocity reports | None | Export as PDF before dropping seats | | Issue linking and epics | Workstreams and lanes | Flatter model. Deep epic trees do not survive it | | Automation rules | AI workers assigned standing tasks | Different mechanism, so plan a rewrite rather than a port | > **The dependency to check first** > > If anything outside Jira reads Jira, a migration is not a content move. Deploy pipelines that transition issues, compliance reports built on issue history, and integrations that create tickets from alerts all have to be rewritten or left in place. Polaris has no workflow engine to receive them. ## Read next - [alternatives/jira](https://www.polarishq.co/alternatives/jira) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [integrations/github](https://www.polarishq.co/integrations/github) - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [replace/jira-and-confluence-and-slack](https://www.polarishq.co/replace/jira-and-confluence-and-slack) - [replace/notion-and-jira-and-slack](https://www.polarishq.co/replace/notion-and-jira-and-slack) ## Questions people ask **Can we keep Slack and still use Polaris chat?** Yes, and most teams should for the first month. Slack connects to Polaris rather than being replaced by it, so messages arrive in the Inbox as task suggestions while the channels keep working. Teams that try to move conversation on day one usually end up with two half-used chat tools. **Is there any way to import Jira issues in bulk?** Not into Polaris. Jira has no importer and is not in the connection catalog. Export to CSV for your archive, then create the live items by hand. A fifteen-person team's active sprint is typically two hours of typing, and the resulting board has no dead tickets in it. **What replaces our Jira automation rules?** AI workers assigned standing tasks, which is a different mechanism rather than a port. A rule that moved an issue on a status change has no equivalent. A rule that generated a weekly summary does, because a worker can be given that job with real tool access and it delivers the summary as a comment with a file. **Where do specs live in Polaris?** In Docs, a nested tree with a block editor, markdown shortcuts, to-dos and sub-pages. Files are versioned and support review and comments. It is the layer a Jira and Slack stack does not have, and for most engineering teams it is the part of this migration that changes daily life the most. **How does the pricing compare to Jira plus Slack?** Jira and Slack are both per-seat and both grow with headcount, including for people who barely open them. Polaris software is free with no seat count, and billing happens only when an AI worker delivers work, at roughly two dollars per human-equivalent hour. The cost pages have the current list prices with the date they were checked. ## Related - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/github - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/replace/notion-and-jira-and-slack - https://www.polarishq.co/replace/linear-and-slack --- --- title: "Replace Linear and Notion? Read this before you do" description: "Both Linear and Notion connect to Polaris rather than needing migration. The case for keeping them, and the narrow case for consolidating anyway." url: https://www.polarishq.co/replace/linear-and-notion section: Replace your stack updated: 2026-08-21 --- # Replacing Linear and Notion Two tools people choose deliberately, and a bill that is small enough that saving money is not the argument. ## The short answer Linear and Notion are both in the Polaris connection catalog, which makes wholesale replacement the least compelling option for this stack. Connecting both lets AI workers read issues and pages without an export. Teams that consolidate anyway usually do it to get delivery rather than tracking, since a Polaris task assigned to a worker wakes a cloud machine that returns finished work as a comment. - **Linear:** Connects, no export - **Notion:** Connects and imports - **Recommended first move:** Connect, do not migrate ## Nobody arrives at this stack by accident Teams on Linear and Notion picked both on purpose. Linear because cycles impose a rhythm that Jira never did, and because the keyboard-first interaction model means people actually update their issues. Notion because writing in it is pleasant, and pleasant writing tools get used. A page whose job is to argue you were wrong about that would be a bad page. Two subscriptions is not tool sprawl, the bill is small at most team sizes, and the seam between an issue and a document is narrower here than in almost any other combination on this site. So the question is not whether to leave. It is whether the two tools plus a pile of individual AI subscriptions is the shape you want, and whether the work your team keeps deferring is work a machine could have finished overnight. ## The seam that does exist Small, but it is the same one every time. - **The doc ages and the issue does not know** — Linear issues get closed. The Notion page describing the approach stays at the version it was written at. There is no mechanism connecting the two, and the mismatch surfaces months later when someone new reads the doc and builds the wrong thing. - **AI usage is stuck on individual laptops** — Both tools have AI features, and separately your engineers are running coding agents locally. None of that is visible to the team, none of it is assignable, and all of it stops when a laptop closes. That is the actual gap in this stack, and it is not a tracking problem. ## What to do instead of migrating Three stages, none of which is a cutover. 1. **Connect Linear and Notion, change nothing else** — Both are authorised once for the organisation and stored server-side. Your team keeps working in Linear and Notion exactly as before. Polaris becomes readable context rather than another place to check. 2. **Hire one worker and give it real work** — Say what keeps slipping. The Chief of Staff runs a short interview, every answer a click, and the capabilities become a SKILL.md file you can open and edit. Give it Linear, Notion and web search. The whole thing takes about a minute. 3. **Judge it on one delivery** — Assign a task with acceptance criteria you would give a contractor. A cloud machine picks it up, runs a live tool loop, ticks the criteria as it goes and posts the result as a comment with files. You close the task and rate it. If that delivery is not worth roughly two dollars an hour, stop there and you have paid nothing. ## When consolidating is the right call **Keep Linear and Notion** - Cycles and triage are working and the team updates issues without nagging - Engineering is the only group that needs a tracker - Your Notion workspace is heavy on databases, tables and embedded media - Two subscriptions at your headcount is not a number anyone has complained about **Move to one workspace** - Non-engineering teams have been locked out of the tracker and are working in spreadsheets - Per-seat cost has started scaling with hires who barely open either tool - You are also paying for several AI subscriptions with nothing shared between them - You want assignment to work the same way for a person and for an agent > **The part a migration page usually hides** > > Polaris is in free public beta with no customers to point at. Linear and Notion are mature products with years of polish. On tracking and writing alone, this is not a fair fight, and pretending otherwise would waste your afternoon. The thing Polaris has that they do not is a roster where humans and AI workers sit in the same table and get assigned work the same way. ## Read next - [integrations/linear](https://www.polarishq.co/integrations/linear) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [alternatives/linear](https://www.polarishq.co/alternatives/linear) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [cost/linear-pricing](https://www.polarishq.co/cost/linear-pricing) - [replace/linear-and-notion-and-github](https://www.polarishq.co/replace/linear-and-notion-and-github) - [replace/notion-and-linear-and-slack](https://www.polarishq.co/replace/notion-and-linear-and-slack) ## Questions people ask **If Linear connects, why would I move off it at all?** For most teams on this stack, you would not, at least not first. The connection lets AI workers read and act on Linear issues while your team keeps using Linear. Consolidation makes sense later, when non-engineering groups need the same workspace or when per-seat cost stops matching the value. **Will my Notion databases survive an import?** No. The importer carries page content: paragraphs, headings, lists, to-dos with their checked state, quotes, code, callouts, dividers and toggles, plus sub-pages three levels deep. Databases, tables, embeds and images are skipped and counted. A Notion-heavy team should expect to rebuild any database it relies on. **Can an AI worker create Linear issues for me?** Linear is in the connection catalog and connections are authorised once, org-wide, with credentials stored server-side so workers can use them and browsers cannot read them back. What any individual worker does with a connection depends on the SKILL.md you write for it, which is a file you can read and edit rather than a hidden prompt. **What does this cost compared to what I pay now?** Linear and Notion are both per-seat. Polaris software is free with no seats and no tiers, and charges roughly two dollars per human-equivalent hour that an AI worker delivers. Running Polaris alongside both tools costs nothing until you assign agent work, which is why connecting first is a genuinely free experiment. **Does Polaris have anything like Linear cycles?** Not in that form. Polaris uses a Focus lane across three horizons, today, this week and the next thirty days, pinned first in every view, plus workstreams with shared lanes across list and board views. It is a different discipline from a two-week cycle with a committed scope, and a team that loves cycles should count that as a real loss. ## Related - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/replace/notion-and-linear-and-slack --- --- title: "Replace Linear and Slack: the small-eng-team plan" description: "Linear and Slack both connect to Polaris. The real gap in this stack is written context, and the real question is whether agents should be teammates." url: https://www.polarishq.co/replace/linear-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Linear and Slack Issues in one place, everything anyone said about those issues in the other, and nothing written down. ## The short answer Linear and Slack cover issue tracking and conversation with no documentation layer between them, so context survives only as long as somebody remembers the thread. Both connect to Polaris without an export. The consolidation adds versioned docs and a shared roster where AI workers are assigned tasks the same way people are, and a cloud machine runs each task to delivery. - **Export required:** None - **Missing layer:** Written context - **Hire a worker:** ~60 seconds, in chat - **Software cost:** $0 ## The thread is the documentation, which is the problem A small engineering team on Linear and Slack has genuinely good hygiene by most standards. Issues get updated, cycles complete, and the channel is fast. What it does not have is a place where a decision goes to survive. So the reasoning lives in threads. Why the schema looks like that, why the third-party library was dropped, what the constraint was that made the obvious approach impossible. All of it is in Slack, which means all of it is findable only by someone who already knows what to search for. The failure is delayed and expensive. It shows up when a new engineer joins, when a decision gets relitigated for the second time, or when the person who held the context leaves. ## Where each piece of this stack ends up | Today | After | Effort | | --- | --- | --- | | Linear issues | Connected, or rebuilt as workstream tasks | None if connected | | Slack channels | Connected, feeding the Inbox as suggestions | One authorisation, org-wide | | Decisions in threads | Docs pages with versions and comments | Manual, and worth doing by hand | | Linear cycles | Focus lane across three horizons | Different model. Expect an adjustment | | Slack huddles and DMs | Not replaced | Keep Slack for these | | Local coding agents | Workers on the roster with editable SKILL.md files | New capability rather than a migration | ## Three stages, smallest first 1. **Connect both and leave them running** — Slack and Linear are authorised once for the whole organisation. Credentials are verified live and stored server-side. Nothing about how your team works changes at this stage, which is the point. 2. **Write the five decisions down** — Pick the five questions your team has answered twice in Slack and turn each into a doc. Versioned, commentable, in a nested tree. This is the layer the stack was missing and it takes an afternoon, not a migration. 3. **Move the roster, not the issues** — Hire two AI workers, give them Slack, Linear, GitHub and web search, and assign them the work your cycle keeps pushing. Humans and agents are rows in the same members table, so assignment is identical. The machine delivers as a comment and ticks its acceptance criteria; a person closes the task. > **What this stack is actually short of** > > Not tracking. Linear is better at issues than Polaris is. What Linear and Slack cannot do is keep working after everyone closes their laptops. If your team already runs coding agents locally and wishes teammates could see them, assign to them, and read what they produced, that is the gap this consolidation fills. ## Read next - [integrations/linear](https://www.polarishq.co/integrations/linear) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [alternatives/linear](https://www.polarishq.co/alternatives/linear) - [cost/linear-pricing](https://www.polarishq.co/cost/linear-pricing) - [cost/slack-pricing](https://www.polarishq.co/cost/slack-pricing) - [cloud-claude-code/claude-code-for-teams](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) - [replace/linear-and-notion-and-github](https://www.polarishq.co/replace/linear-and-notion-and-github) ## Questions people ask **Do Linear issues import into Polaris?** There is no bulk importer for Linear issues. Linear is in the connection catalog instead, so Polaris and AI workers can read it through an org-wide authorisation with no export at all. Teams that want issues to live in Polaris rebuild the active cycle by hand, which is usually under an hour for a small team. **What replaces Linear cycles?** The Focus lane, which is time-based across today, this week and the next thirty days, and is pinned first in every view. It is a looser discipline than a committed two-week cycle. If your team's rhythm depends on cycle commitment and velocity, that is a real loss and worth weighing before you move issues. **Can an AI worker respond to something in a Slack channel?** Slack messages arrive in the Polaris Inbox as prefilled task suggestions with bucket, lane, labels and owner already set, and a human clicks to turn one into a task. Once it is a task you can assign it to an AI worker exactly as you would to a person. Nothing becomes a task, or gets acted on, without that click. **Is this cheaper than Linear plus Slack?** The software is free with no seats, so the subscription line goes to zero. What replaces it is usage: roughly two dollars per human-equivalent hour delivered by an AI worker, estimated by an open formula and logged job by job on the work log. A team that assigns no agent work pays nothing. ## Related - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/replace/linear-and-notion - https://www.polarishq.co/replace/jira-and-slack --- --- title: "Replace Trello and Slack: a plan for small teams" description: "Trello boards export as JSON but do not import into Polaris. What a small team should actually rebuild, what to abandon, and why cost is not the argument here." url: https://www.polarishq.co/replace/trello-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Trello and Slack A board that everyone can read and nobody maintains, plus a channel where the real assignments happen. ## The short answer Trello and Slack suit small teams because both are easy to start and cheap to run, and both stop working at the point where a card needs an owner and a deadline. Trello boards export as JSON but have no Polaris importer, so lists are rebuilt as lanes by hand. Slack connects instead, feeding an Inbox where messages become prefilled task suggestions. - **Trello import:** None. Rebuild lists as lanes - **Slack:** Connects, no export - **Typical rebuild:** One board, under an hour ## Cost is not why you would leave Trello Trello is one of the cheapest ways to give a small team a shared view of work, and any page that opens with the bill is arguing in bad faith. A team of six is not being bled dry by a Trello subscription. Check the pricing page if you want the number, but it is not the reason anyone searches for this. The reason is that the board stops being true. Cards accumulate in a list called In Progress that nobody has looked at since spring. Half of them have no member assigned and no due date, because Trello lets a card exist perfectly happily without either. Meanwhile the actual assignment happened in Slack, where somebody said they would handle it, and the card was never updated because updating the card is a separate chore from doing the work. What a small team needs is not a bigger board. It is fewer places where a piece of work can hide. ## Board archaeology, the honest inventory Before rebuilding anything, sort your cards into these four piles. - **Live work with a named person** — Rebuild these first as tasks with an owner and a date. Usually a surprisingly small number, often under twenty for a team of six. - **Live work with nobody attached** — These are the ones the board was hiding. Each needs a decision now: give it an owner and a horizon, or delete it. Do not carry an unowned card across a migration. - **Reference cards used as documents** — Onboarding checklists, client notes, process descriptions living in card descriptions and attachments. These belong in Docs, not on a board, and rewriting them is the moment the team notices how much of the board was never task-shaped. - **Archaeology** — Anything untouched for a quarter. Export the board to JSON, keep the file, and leave it. If something in there mattered, it will come back on its own. ## A weekend-sized migration This is one of the smaller moves in this cluster, and it should be treated that way. 1. **Export the board and stop looking at it** — Trello exports a board as JSON from the board menu. Save the file somewhere durable. It is your archive and it is not going to be imported anywhere, so treat the export as an insurance policy rather than a migration step. 2. **Rebuild the lists as lanes, live work only** — Create one workstream and recreate your lists as lanes. Lanes are shared between list and board view, so you organise once and look at it either way. Carry across only the cards from the first two piles above, and give every one an owner and a date as you type it. 3. **Connect Slack so assignments stop escaping** — One authorisation for the whole team. When somebody volunteers for something in a channel, the message arrives in the Inbox as a suggestion with the bucket, lane, labels and owner already filled in, and one click turns it into a real task. This closes the exact seam that made the board go stale. 4. **Try one AI worker on the chore nobody wants** — The weekly client update, the competitor check, the invoice chase. Hiring takes about a minute in chat and produces a SKILL.md you can edit. The worker delivers as a comment with a file attached and never marks its own task done. > **When to stay on Trello** > > If your board is genuinely simple, everyone updates it, and the only complaint is the monthly charge, stay. The saving is small and the disruption is real. This move is worth making when work is being lost between the board and the channel, or when you want a teammate who will actually finish the weekly report. ## Read next - [alternatives/trello](https://www.polarishq.co/alternatives/trello) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [cost/trello-pricing](https://www.polarishq.co/cost/trello-pricing) - [cost/stack-cost-5-person-team](https://www.polarishq.co/cost/stack-cost-5-person-team) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [replace/notion-and-trello](https://www.polarishq.co/replace/notion-and-trello) - [replace/monday-and-slack](https://www.polarishq.co/replace/monday-and-slack) ## Questions people ask **Can I import a Trello board into Polaris?** No. Trello has no Polaris importer and is not in the connection catalog. Export the board to JSON from Trello's board menu for your records, then rebuild the lists as lanes and the live cards as tasks. For a small team this is under an hour and removes the dead cards in the process. **What about Trello Power-Ups and butler automations?** Neither transfers. Polaris has no Power-Up equivalent and no rule builder. Recurring work that a Butler rule created is handled differently, by assigning an AI worker a standing task with acceptance criteria, which produces a delivered result rather than a newly created empty card. **Do card attachments come across?** Not automatically. Attachments on Trello cards have to be downloaded and re-uploaded, and there is no bulk path. Most small teams find only a handful of cards have attachments worth keeping, and the rest are screenshots that nobody has opened since the day they were posted. **Is Polaris really free for a six-person team?** The software is free with no seat count and no tier, so six people costs the same as sixty: nothing. Charges start only when an AI worker delivers work, at roughly two dollars per human-equivalent hour, itemised on a work log you can challenge line by line. A team that never assigns agent work never gets a bill. ## Related - https://www.polarishq.co/alternatives/trello - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/trello-pricing - https://www.polarishq.co/cost/stack-cost-5-person-team - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/replace/notion-and-trello - https://www.polarishq.co/replace/asana-and-slack - https://www.polarishq.co/for/small-business --- --- title: "Replace Asana and Slack: a cross-team migration plan" description: "Asana rules, portfolios and custom fields have no Polaris equivalent. What a cross-functional team rebuilds, and why the automation rewrite is the real work." url: https://www.polarishq.co/replace/asana-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Asana and Slack Rules that fire into Slack, replies that never come back, and a portfolio view built for people who do not do the work. ## The short answer Asana and Slack pair a structured tracker with an unstructured conversation layer, and the handoff between them runs one way. Asana rules post into Slack; the discussion that follows never returns to the task. Asana has no Polaris importer or connector, so projects are exported to CSV and rebuilt as workstreams. Slack connects, turning messages into prefilled task suggestions. - **Asana import:** None. CSV export, manual rebuild - **Asana rules:** No equivalent. Plan a rewrite - **Slack:** Connects, org-wide, once - **Software cost after:** $0 ## The one-way notification Asana is good at structure. Custom fields, sections, dependencies, portfolios rolling up across projects, and rules that fire when something changes. Most of that structure was built by an operations person who wanted a cross-functional team to be legible to whoever asks about status. The rule fires and posts to Slack. Someone replies in the thread with the thing that actually matters, which is that the dependency changed or the client moved the date. That reply does not go back to Asana. The task keeps its old due date and its green custom field, and the portfolio rolls up a number that is wrong. This is a different failure from a stale Trello board. Nothing here looks neglected. The dashboard is full, the rules are running, and the reporting is confidently incorrect. ## The three-way sort Asana carries more structure than most trackers, which means more of it has nowhere to go. | Asana feature | Consolidates | Notes | | --- | --- | --- | | Tasks, subtasks, assignees, due dates | Cleanly | Rebuilt as tasks in a workstream with owners and dates | | Sections and project structure | Cleanly | Sections become lanes, shared across list and board views | | Custom fields | Partially | Labels cover categorical fields. Numeric and formula fields do not move | | Portfolios and workload | Not at all | No roll-up reporting layer. This is a real loss for programme managers | | Rules and automation | Not at all | Rewritten as standing tasks assigned to AI workers, which is a different mechanism | | Forms and intake | Not at all | Intake moves to the Inbox, fed by Slack rather than by a public form | | Comments and attachment history | Not at all | Export the CSV, keep Asana readable, do not attempt to reproduce | ## Migrating a cross-functional team Do this one team at a time. A whole-company cutover across marketing, ops and design in one week does not survive contact with the first stakeholder. 1. **Inventory the rules before anything else** — List every automation rule that is actually firing. For each one, write what outcome it produces rather than what it does. Rules that move a task on a status change have no equivalent. Rules that produce a recurring artefact, a report or a summary, can become a standing task for an AI worker instead. 2. **Export projects to CSV for the record** — Asana exports a project to CSV and JSON. This is your archive. It does not import into Polaris, and trying to reproduce comment threads is not worth anyone's week. 3. **Rebuild one team's live projects** — Pick the team with the fewest external dependencies. Recreate their projects as workstreams and their sections as lanes. Set an owner and a date on every task as you go, which is the moment you discover how many Asana tasks had a due date but no real owner. 4. **Connect Slack and move intake** — Requests that arrived through Asana forms now arrive in channels, land in the Inbox as prefilled suggestions, and become tasks with one click. This is a downgrade in structure and an upgrade in the number of requests that get captured at all. 5. **Replace the reporting rule with a worker** — The weekly status roll-up your portfolio was producing becomes a task assigned to an AI worker with the connections it needs. A cloud machine wakes, does the work, ticks its own acceptance criteria and delivers a file as a comment. Somebody reads it and closes the task. > **Do not move if portfolios are load-bearing** > > If an executive reads an Asana portfolio to make resourcing decisions, Polaris has nothing to replace it with. There is no roll-up reporting across workstreams and no workload view. Consolidating anyway means somebody rebuilds that report by hand or hands it to an AI worker as a recurring job, and you should decide which before you start, not after. ## Read next - [alternatives/asana](https://www.polarishq.co/alternatives/asana) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [cost/asana-pricing](https://www.polarishq.co/cost/asana-pricing) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [replace/asana-and-notion](https://www.polarishq.co/replace/asana-and-notion) - [replace/monday-and-slack](https://www.polarishq.co/replace/monday-and-slack) ## Questions people ask **Is there an Asana importer?** No. Asana is not in the Polaris connection catalog and there is no bulk import path. Asana exports projects to CSV and JSON, which serves as your archive, and live projects are rebuilt as workstreams with lanes. For a twenty-person cross-functional team, expect a day per team rather than an afternoon. **What happens to our Asana rules?** They do not port. Polaris has no rule builder. Rules whose purpose was to move a task through states simply disappear, and most teams find that acceptable. Rules that produced something, a report or a digest, are rebuilt as standing tasks assigned to an AI worker, which delivers the artefact rather than triggering a status change. **Do custom fields survive?** Partly. Categorical fields become labels, which covers the common case of a status, a priority or a team. Numeric fields, formula fields and anything a portfolio was summing do not move. Audit which fields you actually filter or report on before you start, because most projects have several nobody has read in a year. **Can we keep Slack?** Yes, and for a cross-functional team you probably should for the first two months. Slack is in the connection catalog, authorised once for the organisation with credentials stored server-side. Connecting it feeds the Polaris Inbox with task suggestions while every existing channel keeps working exactly as it does now. **What does the bill look like after?** Asana and Slack are both per-seat and both scale with headcount, including for stakeholders who only read. Polaris software is free with unlimited people, and charges only for delivered agent work at roughly two dollars per human-equivalent hour. The saving is largest exactly where per-seat pricing hurts most, which is read-only stakeholders. ## Related - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/replace/asana-and-notion - https://www.polarishq.co/replace/clickup-and-slack --- --- title: "Replace Confluence and Jira: the honest difficulty" description: "The hardest migration in this set. Space permissions, Jira macros inside pages, admin and audit requirements, and why a one-team pilot beats a cutover." url: https://www.polarishq.co/replace/confluence-and-jira section: Replace your stack updated: 2026-08-21 --- # Replacing Confluence and Jira An Atlassian estate has administrators, permission schemes and an auditor, and none of those are content problems. ## The short answer Confluence and Jira form a governed estate rather than two products, with shared identity, space and project permissions, admin controls and retention obligations. Neither has a Polaris importer or connector. Confluence exports spaces as XML, HTML or PDF and Jira exports issues as CSV, but page macros that render live Jira data export as dead snapshots. A single-team pilot is the only realistic starting point. - **Importers available:** Neither product - **Biggest hidden cost:** Permissions and macros - **Recommended scope:** One team, not the estate - **Polaris auth:** Passwordless email code only ## This is an estate, not a stack Every other page in this cluster describes tools a team chose. An Atlassian estate is usually something a company decided, with an administrator who owns it, a permission scheme somebody documented, and at least one process that exists because an auditor asked for it. That changes what a migration means. The content is the easy part. The hard parts are that spaces have restricted pages whose restrictions are not visible in an export, that identity flows through Atlassian Access rather than through each product, and that somebody has to answer for where the data lives afterwards. The specific trap is the Jira macro. A Confluence page that renders a live issue table looks like documentation and is actually a query. Export the space and those macros come out as dead snapshots or as nothing at all. Teams discover this after the export, which is the wrong time. ## What an Atlassian shop has that this cluster's other readers do not **Constraints to check before starting** - Space and page-level restrictions that do not survive an export - Identity through Atlassian Access, with SSO and directory sync - Retention and residency commitments made to customers or regulators - Deploy or release gates that read Jira issue transitions - An administrator whose job includes saying no to this **What Polaris offers against those** - Passwordless email-code sign-in, and no password path at all - Row-level security in Postgres, with agents and humans in the same members table - Connections authorised once, org-wide, stored server-side and unreadable by browsers - A per-task work log that itemises what an AI worker did and how long it was billed for - No permission-scheme equivalent, which has to be said plainly ## A pilot, not a programme Five stages that stop being reversible only at the last one. 1. **Pick a team with no compliance surface** — Internal tooling, growth, or a team whose work never appears in an audit. The pilot is about whether people prefer working this way, and running it inside a regulated workflow answers a different question badly. 2. **Inventory the macros in that team's space** — Search their Confluence space for Jira issue macros, page includes and any blueprint-driven template. Each is a page that will not export as what it looks like. Decide for each whether it becomes a written document, a Polaris task list, or something you keep in Confluence. 3. **Rewrite rather than export** — For a single team's living documentation, rewriting is usually faster than converting an XML space export, and it removes a decade of pages nobody opens. Export the full space to PDF for the archive and keep Confluence readable. 4. **Rebuild the current board by hand** — Export the team's live issues from Jira to CSV, then create them as Polaris tasks with owners and dates. Leave every closed issue behind. Anything a release gate transitions stays in Jira until that gate is rewritten. 5. **Give the pilot team an AI worker and measure one thing** — Whether work comes back finished. Hiring takes about a minute and produces a SKILL.md file the team can read and edit. The machine delivers as a comment with files, ticks its acceptance criteria, and never closes its own task. If that changes the team's output, you have a case to take to the administrator. If it does not, you cancelled nothing. > **Say this to your administrator before you say anything else** > > Polaris is in free public beta. There are no customers to reference, no compliance certifications to cite and no enterprise agreement. For an Atlassian estate that is a serious objection, not a formality. The defensible version of this project is a pilot on non-regulated work, run in parallel, with Confluence and Jira untouched. ## Read next - [alternatives/confluence](https://www.polarishq.co/alternatives/confluence) - [alternatives/jira](https://www.polarishq.co/alternatives/jira) - [cost/confluence-pricing](https://www.polarishq.co/cost/confluence-pricing) - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [cost/stack-cost-50-person-team](https://www.polarishq.co/cost/stack-cost-50-person-team) - [replace/jira-and-confluence-and-slack](https://www.polarishq.co/replace/jira-and-confluence-and-slack) - [replace/confluence-and-notion](https://www.polarishq.co/replace/confluence-and-notion) ## Questions people ask **Can we import a Confluence space into Polaris?** No. Confluence has no Polaris importer and is not in the connection catalog. Confluence can export a space as XML, HTML or PDF, which is a fine archive and not an import path. For a single team's living pages, rewriting into Polaris Docs is usually faster than converting an export and produces documentation somebody has actually read this year. **What happens to Jira macros embedded in Confluence pages?** They are queries rendered as content, so they do not survive an export as living data. A page showing an open-issues table becomes a snapshot at best. Identify these before you export, and decide per page whether the content becomes a document, a Polaris lane, or stays in Confluence. **How do page restrictions and space permissions map across?** They do not map. Polaris has organisations with members and row-level security in Postgres, not Confluence's space and page restriction model. A team relying on restricted pages for confidentiality needs to check that Polaris's model is sufficient before moving anything, and for some content the honest answer will be that it stays in Confluence. **Does Polaris support SSO or SCIM?** Authentication is passwordless email codes, with no password path in the product. There is no SAML SSO or directory sync. For an organisation whose identity requirements run through Atlassian Access or an equivalent, that is a blocking difference and it should be raised at the start of the conversation rather than the end. **Is the cost saving worth this much work?** For a large Atlassian estate the licence saving is real and the migration effort is larger than any figure on the cost pages accounts for. The stronger argument is delivery rather than price: free software plus roughly two dollars per human-equivalent hour of AI-delivered work, itemised and challengeable. If that is not compelling on its own, the licence maths probably will not carry the project. ## Related - https://www.polarishq.co/alternatives/confluence - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/stack-cost-50-person-team - https://www.polarishq.co/replace/jira-and-confluence-and-slack - https://www.polarishq.co/replace/confluence-and-notion - https://www.polarishq.co/replace/notion-and-jira --- --- title: "Replace Notion, Linear and Slack without migrating" description: "All three of these tools are in the Polaris connection catalog. The plan is to connect them, close the three-way seam, and retire subscriptions later." url: https://www.polarishq.co/replace/notion-and-linear-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Notion, Linear and Slack The default startup stack, three subscriptions, and a decision that has to be written down three times to stay true. ## The short answer Notion, Linear and Slack are all in the Polaris connection catalog, so this is the one combination in this cluster that needs no export at all. Connect all three, authorised once for the organisation, and AI workers read them directly. The seam this stack loses work in is the round trip: a decision made in chat has to be written into a page and reflected in an issue, by hand, every time. - **Tools needing export:** Zero - **Subscriptions today:** Three, all per-seat - **Polaris software:** $0, unlimited people - **Hire a worker:** ~60 seconds, in chat ## The three-way round trip This is the stack that gets recommended in every thread about what a modern software team should run, and the recommendation is not wrong. Each of the three is very good at its job. The cost is not any one of them. It is the round trip between them. A decision gets made in a Slack thread. To survive, it has to be written into the Notion page that describes the approach, and reflected in the Linear issue that implements it. That is two acts of manual transcription, performed by a person who has already got what they needed from the conversation and has no incentive to do either. So the transcription happens sometimes. The page is right about half the things and the issue is right about a different half, and the only complete version was in a thread that scrolled past three weeks ago. ## What connecting all three actually gives you Each connection is authorised once for the whole organisation, verified live, and stored server-side so workers can use the credentials and browsers cannot read them back. - **Slack stops being a leak** — Messages that look like work arrive in the Polaris Inbox as prefilled task suggestions with a bucket, lane, labels and owner already chosen. One click makes a task. No click means nothing happens, which is why the Inbox stays usable. - **Notion becomes readable to your workers** — An AI worker given the Notion connection reads the pages you point it at. Separately, the importer can copy pages into Polaris Docs if you want them there, up to twenty-five per run with sub-pages three levels deep. - **Linear stays exactly where it is** — No export, no rebuild, no arguing with the team about cycles. The connection is there so workers can see the issues, and your engineers carry on in the tool they like. - **GitHub is in the catalog too** — If your team is on GitHub, that is a fourth connection and it comes from the same list. A worker with Linear, GitHub, Slack and web search has most of what a junior engineer would need on their first week. ## Three stages, no cutover 1. **Connect the three, tell nobody** — Authorise Slack, Notion and Linear. Nothing about anyone's day changes. This stage exists so the next one is possible, and it takes minutes. 2. **Hire the Chief of Staff's first recruit** — The Chief of Staff is on your roster from the first sign-in. Tell it what keeps falling through, answer a short interview where every question is a click, and the capabilities become a SKILL.md file you can open and edit. Give the new worker the three connections plus web search. 3. **Assign the transcription problem itself** — The standing job of reading a channel, a page and an issue and reporting where they disagree is exactly the work nobody volunteers for. A cloud machine wakes per task, runs a live tool loop, ticks its own acceptance criteria and delivers as a comment with files. A person closes the task. That loop is the whole product. ## Retire now, or retire later You do not have to decide this on day one, and you should not. **Reasons to keep all three running** - Your engineers update Linear issues without being asked - Notion holds databases, tables and media that the importer would skip - Slack carries huddles, DMs and every external shared channel - Three subscriptions at your headcount is not a number anyone has complained about **Reasons to consolidate for real** - Per-seat cost is scaling with hires who open one of the three - Non-engineering teams cannot get into the tracker and are working in spreadsheets - You are also paying for individual AI subscriptions with nothing shared between them - You want assignment to work identically for a person and for an agent ## Read next - [integrations/slack](https://www.polarishq.co/integrations/slack) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [integrations/linear](https://www.polarishq.co/integrations/linear) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) - [replace/linear-and-notion-and-github](https://www.polarishq.co/replace/linear-and-notion-and-github) - [replace/notion-and-jira-and-slack](https://www.polarishq.co/replace/notion-and-jira-and-slack) ## Questions people ask **Do I have to cancel anything to start?** Nothing. All three tools are in the connection catalog, so the starting position is Polaris running alongside your existing stack with no export and no cutover. The software is free with no seat count, so running both costs nothing until you assign work to an AI worker. **What is the difference between connecting Notion and importing from Notion?** Connecting authorises the whole organisation once so AI workers can read Notion through stored credentials. Importing copies pages into Polaris Docs, up to twenty-five selected pages per run with sub-pages three levels deep, capped at sixty pages and fifteen hundred blocks. Most teams connect first and import only the pages they want to edit in Polaris. **Which of the three should we drop first?** Usually none of them, for at least a month. When teams do drop one, it tends to be whichever has the most read-only seats attached, because that is where per-seat pricing costs the most for the least use. The tracker is normally the last to go, because tracker habits are the hardest to move. **How does this compare to just adding an AI subscription to each tool?** Per-tool AI features work inside their own product and stop at its boundary, which leaves the round trip between the three exactly where it was. A Polaris worker holds connections to all three at once and delivers a finished artefact as a comment with files. It is also billed by delivered work rather than per seat per month. **Is Polaris ready for a team that already likes its tools?** It is in free public beta with a working product, a mobile app and recorded demos, and no customers to point at. A team that is happy with Notion, Linear and Slack should treat this as an addition rather than a replacement, connect all three, and judge it on whether a worker delivers something worth paying for. ## Related - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/replace/linear-and-notion - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/replace/notion-and-jira-and-slack --- --- title: "Replace Notion, Jira and Slack: a staged plan" description: "Two of these three connect to Polaris and one does not. Do the free half first, then decide about Jira as a separate project with its own timeline." url: https://www.polarishq.co/replace/notion-and-jira-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Notion, Jira and Slack Two tools you can connect this afternoon and one that is a project with a start date and an owner. ## The short answer Notion, Jira and Slack split cleanly into two halves. Notion and Slack are in the Polaris connection catalog and need no export, so that half is done in an afternoon. Jira has no importer and no connector, and moving off it means a CSV export plus a manual rebuild of live issues. Treating those halves as one project is why these migrations stall. - **Connect today:** Notion and Slack - **Separate project:** Jira - **Jira import path:** CSV export, manual rebuild - **Software cost after:** $0, unlimited people ## Two of these are not like the third This is the most common stack in companies between twenty-five and a hundred people, and it usually arrived in layers. Slack came first because everyone needed to talk. Notion came second because someone wanted the handbook somewhere better than a shared drive. Jira came third, when engineering got big enough that a board in Notion stopped being credible. Because it arrived in layers, it can leave in layers too, and that is the useful insight for this combination. Notion and Slack are both in the connection catalog, which means the first half of this migration involves no export, no cutover and no argument. Jira is a different kind of task, with a real cost and a real risk, and it deserves its own timeline. Most failed consolidations here fail because someone scoped all three as one initiative, ran into the Jira workflow scheme in week two, and lost the momentum that the easy half would have given them. ## The three seams, and which half each belongs to | Seam | Where work is lost | Fixed by | | --- | --- | --- | | Slack to anywhere | Decisions scroll past and never reach a task | The free half. Connect Slack, approve Inbox suggestions | | Notion to Jira | The spec and the ticket stop agreeing after grooming | The free half, partly. A worker can flag the differences | | Jira to Notion | Closed issues never update the page that described them | The Jira half. Only closing when tracking moves | | Jira to Slack | Bot notifications that nobody reads and cannot be replied to | The free half. Inbox suggestions replace the firehose | ## Half one: this afternoon No export, no risk, and it costs nothing because the software is free. 1. **Connect Slack and Notion** — Two authorisations, org-wide, credentials verified live and stored server-side. Slack starts feeding the Inbox with prefilled task suggestions. Notion becomes readable to any AI worker you give the connection to. 2. **Import the handbook, not the workspace** — Run the Notion importer over the pages people actually open. Twenty-five per run, sub-pages three levels deep, capped at sixty pages and fifteen hundred blocks. Read the skipped count afterwards, because it tells you how much of your documentation was tables and screenshots rather than words. 3. **Put the company's non-negotiables in Focus** — Focus is the home screen, a time-based lane across today, this week and the next thirty days, pinned first in every view. It is the one thing this stack has no equivalent of, because a Jira board answers what a team is doing and never answers what must not slip. ## Half two: the Jira decision A separate project with its own owner. Run it only after the first half has been live for a month. 1. **Find what outside Jira reads Jira** — Deploy gates, alerting integrations that open tickets, compliance reports built on issue history. Every one of these has to be rewritten or left in place. Polaris has no workflow engine and no issue transitions for them to hook into. 2. **Rebuild one team's live board** — CSV export for the archive, manual creation for the live work, an owner and a date on every task as you type it. Leave the backlog behind deliberately. Items that matter get raised again within a fortnight. 3. **Drop write seats, keep the archive** — Cancel seats rather than the account once nobody needs write access. Keeping the history readable costs a fraction of what full seats cost and removes the only genuinely irreversible step from this plan. > **The reason to do half one even if half two never happens** > > A team with Slack and Notion connected can hire an AI worker in about a minute and give it both, plus web search. The worker gets a SKILL.md file you can read and edit, runs on a cloud machine that wakes per task, and delivers finished work as a comment with files attached. None of that required touching Jira, and it is the part that changes what your team can get done. ## Read next - [replace/notion-and-jira](https://www.polarishq.co/replace/notion-and-jira) - [replace/jira-and-slack](https://www.polarishq.co/replace/jira-and-slack) - [replace/jira-and-confluence-and-slack](https://www.polarishq.co/replace/jira-and-confluence-and-slack) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [cost/stack-cost-50-person-team](https://www.polarishq.co/cost/stack-cost-50-person-team) ## Questions people ask **Why not move all three at once?** Because two of them require no work and one of them is a project. Bundling them means the easy half waits on the hard half, and the hard half is where a Jira workflow scheme or a deploy gate stops everything. Connect Slack and Notion now, and put a date on the Jira conversation separately. **Can we run Polaris and Jira side by side indefinitely?** Yes. Polaris software is free with no seat count, so running it alongside an existing stack has no licence cost. Plenty of teams end up with docs and cross-functional work in Polaris and engineering tickets still in Jira, and that is a legitimate steady state rather than a failed migration. **What does the Notion importer skip?** Images, files, embeds, tables, databases, synced blocks and column layouts, and it reports how many blocks it dropped. Numbered lists arrive as bullets and toggles arrive as bullets that no longer fold. Page properties, comments and version history never come across. Text, headings, lists, to-dos, quotes, code, callouts and dividers all do. **Does connecting Slack mean Polaris posts into our channels?** Connecting brings Slack signals into the Polaris Inbox as suggestions that a human approves. Signals become suggestions rather than silent tasks. What any individual AI worker is allowed to do with a connection is set by the SKILL.md you write for it, which is a file you can read and edit rather than a hidden prompt. **How much does this save a fifty-person company?** Three per-seat subscriptions multiplied by fifty people is the figure to beat, and the cost pages carry the current list prices with the date they were checked. After the move the software line is zero and the variable line is roughly two dollars per human-equivalent hour of delivered agent work, logged job by job and challengeable from the work log. ## Related - https://www.polarishq.co/replace/notion-and-jira - https://www.polarishq.co/replace/jira-and-slack - https://www.polarishq.co/replace/notion-and-linear-and-slack - https://www.polarishq.co/replace/jira-and-confluence-and-slack - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/stack-cost-50-person-team --- --- title: "Replace ClickUp and Slack: consolidating twice" description: "ClickUp already promised all-in-one and you still pay for Slack. What that says about the stack, what exports, and what a second consolidation should aim at." url: https://www.polarishq.co/replace/clickup-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing ClickUp and Slack You already bought one all-in-one product. The Slack subscription next to it is the interesting fact. ## The short answer ClickUp already includes docs, chat, whiteboards and dashboards, so a team running ClickUp alongside Slack has consolidated once and kept paying for a second chat tool anyway. ClickUp exports tasks to CSV and has no Polaris importer or connector; its Space, Folder and List hierarchy is rebuilt as workstreams and lanes. Slack is in the connection catalog and stays. - **ClickUp import:** None. CSV export, manual rebuild - **Slack:** Connects, no export - **Hierarchy change:** Space, Folder, List becomes lanes ## The Slack subscription is the evidence ClickUp ships chat. It ships docs, whiteboards, dashboards, goals and forms. A team that bought ClickUp bought all of that, and the fact that the same team is still paying for Slack says something specific: the surface area was added, and the habit did not move. That pattern repeats inside the product. Most ClickUp accounts have custom statuses somebody configured with real care, views that were built once and never opened again, and automations that fire into a space nobody watches. The tool is not the problem. Having more places to configure than work to do is. So a second consolidation should be judged on a different question than the first one was. Not how many features come in the box, but how much work comes back finished. ## What to check in your own account before deciding This audit takes twenty minutes and settles the argument either way. - **Count the views nobody opened this month** — List, board, calendar, gantt, timeline, workload, per space. If most of them have not been opened, you are paying for optionality rather than for work. - **Count the custom statuses per list** — Custom statuses are ClickUp's best feature and its most common source of drift, because every list ends up with a slightly different set and nothing rolls up cleanly across them. - **Count how many decisions happened in ClickUp Chat** — If the answer is near zero, the chat feature is dead weight on your bill and Slack is your real chat tool. That is fine, and it is the argument for connecting Slack rather than trying to move conversation again. - **Count the automations that produce something** — Most automations change a status. A few produce an artefact. Only the second kind has a Polaris equivalent, in the form of a standing task assigned to an AI worker that delivers a file. ## Where ClickUp's pieces land | ClickUp | Polaris | How clean | | --- | --- | --- | | Spaces, Folders, Lists | Workstreams with lanes | Flatter. Three levels collapse to two | | Tasks and subtasks | Tasks in lanes | Rebuilt from CSV export by hand | | Custom statuses | Lanes and labels | Partially. Per-list status sets do not survive | | ClickUp Docs | Polaris Docs | Rewritten. No importer, and export fidelity is poor | | Dashboards and goals | Nothing equivalent | Does not move. Hand the reporting to a worker instead | | Whiteboards | Nothing equivalent | Does not move at all | | ClickUp Chat | Polaris chat, or the Slack connection | Depends which one your team actually uses | ## The second consolidation, in three moves 1. **Decide which chat tool is real, and drop the other** — One of ClickUp Chat and Slack is carrying your conversations and the other is a line on an invoice. If it is Slack, connect it to Polaris and stop paying for chat twice. This decision alone often justifies the audit. 2. **Export tasks and rebuild only the live lists** — ClickUp exports tasks to CSV. Rebuild the lists that had activity this month as lanes in a workstream, with an owner and a date on every task. The lists that had no activity are the answer to whether the hierarchy was helping. 3. **Replace the dashboard with a delivered report** — Instead of rebuilding a dashboard nobody reads, assign an AI worker the standing job of producing the summary the dashboard was for. It runs on a cloud machine, ticks its own acceptance criteria, and delivers a file as a comment. Somebody reads it, which is more than the dashboard was getting. > **What you genuinely lose** > > Whiteboards, dashboards, goals, gantt and workload views, forms, and the per-list custom status model. Polaris has none of these and none are on this page's roadmap to claim. If your team lives in a timeline view or runs intake through ClickUp forms, this migration removes something you use daily, and no amount of AI delivery compensates for that on its own. ## Read next - [alternatives/clickup](https://www.polarishq.co/alternatives/clickup) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [cost/clickup-pricing](https://www.polarishq.co/cost/clickup-pricing) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [replace/monday-and-slack](https://www.polarishq.co/replace/monday-and-slack) - [replace/asana-and-slack](https://www.polarishq.co/replace/asana-and-slack) ## Questions people ask **Can ClickUp tasks be imported into Polaris?** No. ClickUp has no Polaris importer and is not in the connection catalog. ClickUp exports tasks to CSV, which is your archive, and the live lists are rebuilt as lanes by hand. Teams usually rebuild far fewer lists than they expected, because the audit reveals how many were dormant. **What happens to ClickUp Docs?** They have to be rewritten. There is no importer for ClickUp Docs and no connector, and export fidelity is poor enough that copying and pasting the pages people actually open is usually the faster route. The only document importer Polaris ships is for Notion. **We use ClickUp Chat, not Slack. Does that change anything?** It simplifies the move, because Polaris includes team chat and you have no second chat subscription to reconcile. It also means the Slack connection is irrelevant to you, so the whole migration is a single-product rebuild rather than a connect-plus-rebuild. **Is there any equivalent to ClickUp automations?** Not as rules. Automations that change a status have no equivalent and simply go away. Automations that generate something, a report or a recurring checklist, become standing tasks assigned to AI workers, which deliver a finished artefact rather than triggering a state change. That is a rewrite, not a port. **How does the cost compare?** ClickUp and Slack are both per-seat. Polaris software is free with unlimited people and no tiers, and bills roughly two dollars per human-equivalent hour of AI-delivered work, logged job by job. For a team that consolidated to ClickUp specifically to reduce subscriptions, the second consolidation removes the subscription line entirely. ## Related - https://www.polarishq.co/alternatives/clickup - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/clickup-pricing - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/replace/monday-and-slack - https://www.polarishq.co/replace/asana-and-slack --- --- title: "Replace Monday.com and Slack: an ops team plan" description: "Monday sells seats in blocks and meters automation runs. What that costs an ops team, what exports, and why the board is usually built for the wrong person." url: https://www.polarishq.co/replace/monday-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Monday.com and Slack A board configured by the person who reports on the work, used by the people who have to do it. ## The short answer Monday.com is configured centrally and read by managers, which is why its columns are usually more detailed than the people doing the work will maintain. Boards export to Excel or CSV and there is no Polaris importer or connector, so items are rebuilt as tasks in lanes. Slack is in the connection catalog, so messages become prefilled task suggestions in an Inbox rather than another notification. - **Monday import:** None. Excel or CSV export - **Slack:** Connects, org-wide, once - **Polaris seats:** Unlimited, $0 - **Polaris billing:** ~$2 per human-hour delivered ## Who the board was built for Monday boards are usually built by an operations manager, an agency producer or a founder who needed visibility across several streams at once. They are configured carefully, with status columns, people columns, timeline columns and a dashboard on top that answers the question a stakeholder keeps asking. The people doing the work experience the same board differently. To them it is a set of fields to update after the fact, on a screen they open because they were asked to, describing work they have already finished. So the update happens on Thursday afternoon for the whole week, in one pass, from memory. That is why the seam in this stack runs through Slack. The real status was given in a channel on Tuesday, in a sentence, to the person who asked. The board learned about it two days later in a less accurate form. ## Two things about Monday's pricing that change the maths Both are structural rather than a number, and both are worth checking on the current pricing page before you commit either way. - **Seats are sold in blocks** — You do not always add one seat for one person. Crossing a block boundary can cost more than the person's actual usage justifies, which is why teams end up sharing a login for a stakeholder who only reads. Polaris has no seat count at all, so read-only stakeholders cost nothing. - **Automation and integration runs are metered per month** — Most plans include an allowance of automation actions, and heavy boards can exhaust it, which is a cost that scales with how much you automate rather than how many people you employ. Polaris has no rule engine and therefore no allowance, and recurring work is handled by assigning a standing task to an AI worker instead. ## Rebuilding an ops board Four stages. The first one is the one people skip and then regret. 1. **Ask who reads each column** — Go column by column across your main board and name the person who reads it. Columns with no reader get deleted rather than migrated. In most ops boards this removes between a third and half of the configuration, and it is the single largest saving in this whole plan. 2. **Export to Excel and rebuild the surviving structure** — Monday exports a board to Excel or CSV. Groups become lanes in a Polaris workstream, items become tasks with an owner and a date, and the surviving columns become labels. Subitems flatten, which is a real change and usually an improvement for a board that had grown three levels deep. 3. **Connect Slack so status stops arriving twice** — One authorisation for the organisation. When somebody reports progress in a channel, it lands in the Inbox as a suggestion with bucket, lane, labels and owner already filled in. One click and the task is updated in the same place the work lives, which removes the Thursday afternoon catch-up entirely. 4. **Hand the dashboard's job to a worker** — The stakeholder wanted a report, not a dashboard. Hire an AI worker in about a minute, give it the connections it needs, and assign it the recurring summary. It runs on a cloud machine, ticks its acceptance criteria, and delivers the report as a comment with a file. You close the task after reading it. ## Before and after, for a twenty-person operations team **Now** - Two per-seat subscriptions, one of them sold in seat blocks - A monthly allowance of automation runs that heavy boards can exhaust - Stakeholders who only read still occupy paid seats - Status collected twice, once in chat and once in a column **After** - No software cost, no seat count, unlimited stakeholders - No rule engine and no automation allowance, because recurring work is assigned rather than triggered - Roughly two dollars per human-equivalent hour that a worker delivers - Every billed hour itemised on a work log you can challenge line by line ## Read next - [alternatives/monday](https://www.polarishq.co/alternatives/monday) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [cost/monday-pricing](https://www.polarishq.co/cost/monday-pricing) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [replace/clickup-and-slack](https://www.polarishq.co/replace/clickup-and-slack) - [replace/asana-and-slack](https://www.polarishq.co/replace/asana-and-slack) ## Questions people ask **Can Monday boards be imported into Polaris?** No. Monday is not in the Polaris connection catalog and there is no importer. Monday exports a board to Excel or CSV for your archive, and the live groups and items are rebuilt as lanes and tasks by hand. The column audit that comes with the rebuild usually removes more configuration than it recreates. **What replaces Monday dashboards?** Nothing, structurally. Polaris has no roll-up dashboard or widget layer. The practical replacement is to assign the report the dashboard was producing to an AI worker as a recurring task, which delivers a written summary as a comment with a file. That is a different artefact from a live dashboard and it suits stakeholders who wanted an answer rather than a screen. **Do subitems survive the move?** They flatten. Polaris uses workstreams with lanes and tasks rather than nested items, so a board with items and subitems becomes a flatter structure. Teams with deep subitem trees should test one board before committing, because for some workflows that flattening loses genuine information. **We have stakeholders who only look at the board. What do they cost?** Nothing. Polaris software is free with no seat count and no tiers, so read-only stakeholders, clients and executives cost the same as everyone else. On a per-seat product sold in seat blocks, those same people are often the reason a team crossed into a higher tier. **Is Polaris established enough for an operations team to depend on?** It is in free public beta with a working product, a mobile app and recorded demos, and no customers to cite. An operations team whose board is the company's source of truth should run one workstream in parallel for a month before moving anything that a client or an executive depends on. ## Related - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/replace/clickup-and-slack - https://www.polarishq.co/replace/asana-and-slack --- --- title: "Replace Notion and Trello: the same work, written twice" description: "A stack with a doc tool, a board and no chat. What imports from Notion, what a Trello rebuild involves, and the team chat you gain that you never had." url: https://www.polarishq.co/replace/notion-and-trello section: Replace your stack updated: 2026-08-21 --- # Replacing Notion and Trello A page describing the project and a board describing the same project, kept in step by somebody's memory. ## The short answer Notion and Trello give a small team a place to write and a place to track, with no chat tool between them, so coordination happens in email, direct messages or in person. Notion pages import into Polaris Docs within fixed caps. Trello has no importer, so lists are rebuilt as lanes. Consolidating this stack adds team chat that neither tool provided. - **Notion:** Imports, 25 pages per run - **Trello:** No importer. Rebuild lists as lanes - **New capability:** Team chat this stack never had ## The duplication is the whole problem This stack has an unusual shape. Notion holds a project page with the plan, the context and a list of what needs doing. Trello holds a board with cards for what needs doing. Those two lists are the same list, written twice, in two products, by the same person, on two different days. Keeping them in step is nobody's job and there is no mechanism for it, so they drift. The page says one thing, the board says another, and the team resolves it by asking whoever wrote them. On a team of five that works, which is exactly why the duplication survives for years. The other unusual thing is what is missing. There is no chat tool here. Coordination happens in email, in direct messages on whatever people already use, or across a desk. That means the coordination is invisible and unsearchable, and it means consolidating this stack adds something rather than only removing something. ## The move, item by item | What you have | Where it goes | Effort | | --- | --- | --- | | Notion project pages | Docs, as a nested tree | Imports. A small workspace often fits in one run | | Notion databases | Workstreams with lanes | Rebuilt. Properties do not transfer | | Trello lists | Lanes, shared across list and board views | Manual, and usually quick | | Trello cards with a member and a date | Tasks with an owner and a horizon | Manual. Type them as you decide they still matter | | Trello cards with neither | A decision, then a task or the bin | This is where the value is | | Card attachments | Re-uploaded by hand | No bulk path. Most are not worth moving | | Nothing | Team chat | New. Nothing to migrate because nothing existed | ## A migration a five-person team can do in a morning 1. **Import Notion first, because it is the fastest win** — Small workspaces often clear the whole import in a single run. Twenty-five pages per run, sub-pages three levels deep, sixty pages and fifteen hundred blocks in total. Text, headings, lists, to-dos with their checked state, quotes, code, callouts and dividers all arrive. Images, tables, databases and embeds are skipped and counted. 2. **Rebuild the board from the page, not from the board** — This is the trick that pays for the whole exercise. Open the Notion project page, not the Trello board, and create tasks from what the page says is actually needed. Then check the board for anything real that the page missed. You will find that one of the two lists was mostly obsolete, and the duplication ends there. 3. **Start using chat for the coordination that was invisible** — The decisions that were happening by email or across a desk now happen where the work is. This is the part that feels different a week in, because for the first time the reasoning sits next to the task rather than in somebody's sent folder. > **The five-person case for doing this at all** > > Two subscriptions at this size is not an expensive bill and it is not the reason to move. The reason is that a team this small cannot afford to lose a week of work to two lists disagreeing, and cannot afford to hire the person who would do the weekly admin. The software is free, so the experiment costs an hour of your morning and nothing else. ## Read next - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [alternatives/trello](https://www.polarishq.co/alternatives/trello) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [cost/trello-pricing](https://www.polarishq.co/cost/trello-pricing) - [cost/stack-cost-5-person-team](https://www.polarishq.co/cost/stack-cost-5-person-team) - [replace/trello-and-slack](https://www.polarishq.co/replace/trello-and-slack) - [replace/notion-and-slack](https://www.polarishq.co/replace/notion-and-slack) ## Questions people ask **Which should I move first, Notion or Trello?** Notion, because it is the only one with an importer and it takes minutes. Once the pages are in Polaris Docs, rebuild the board from what the pages say rather than from the cards. That order surfaces the duplication instead of copying it into a new tool. **Will my whole Notion workspace fit in one import?** For a small team, usually. The limits are twenty-five selected pages per run, sub-pages to three levels, and a hard stop at sixty pages or fifteen hundred blocks. A workspace heavy on images and tables will report a high skipped-block count, because those are not carried across. **We do not use Slack. Do we need to add it?** No. Polaris includes team chat, so this combination gains a chat layer without adding a subscription. Slack is in the connection catalog if you ever adopt it, but a team coordinating by email and in person has nothing to connect and nothing to migrate. **What does it cost for a team of five?** The software is free with no seat count, so five people cost nothing. Charges start only when an AI worker delivers work, at roughly two dollars per human-equivalent hour, estimated by an open formula and logged job by job. Nothing delivered means nothing billed. ## Related - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/alternatives/trello - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/cost/trello-pricing - https://www.polarishq.co/cost/stack-cost-5-person-team - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/replace/trello-and-slack - https://www.polarishq.co/replace/notion-and-slack --- --- title: "Replace Airtable and Slack? Keep the database" description: "Airtable is a relational database, not a tracker, and none of it moves to Polaris. The honest plan is a split: keep the base, move the human work." url: https://www.polarishq.co/replace/airtable-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Airtable and Slack A relational base doing three jobs at once, and only one of those jobs should move. ## The short answer Airtable is a relational database with views on top, so linked records, rollups, formulas, interfaces and automations have no equivalent in Polaris and no migration path. Slack is in the connection catalog and needs no export. The workable plan for this combination is a split: keep Airtable as the database, move the human work that was living in its views, and connect Slack. - **Airtable data:** Does not move. Keep the base - **Slack:** Connects, no export - **What moves:** The task views only - **Polaris software:** $0, no seats ## Airtable is doing three jobs and you only want to move one In almost every account, an Airtable base has grown into three things at once. It is a database, holding records with real relationships between them. It is a set of views, one of which somebody uses as a task list. And it is an application, with interfaces built for people who never open the underlying tables. Only the middle one belongs in a work tool. The database is a database, and Polaris has nothing that resembles linked records, rollup fields, lookups or formulas. The interfaces are an application layer, and Polaris has no equivalent of those either. Any page that told you to replace Airtable with a task tracker would be telling you to throw away a data model in exchange for a to-do list. Do not do that. The parts of this stack that are worth consolidating are the human coordination and the Slack scrollback, not the base. ## The split, drawn explicitly This is the only page in this cluster where the recommendation is to keep the tracker-shaped tool. **Stays in Airtable** - Tables, records and every link between them - Rollups, lookups, formula fields and computed values - Interfaces built for people who never touch the grid - Automations, scripts and anything that writes back into the base - Whatever an external form or a customer-facing process feeds **Moves into Polaris** - The view somebody was using as a task list, rebuilt as a workstream - The recurring work that a person does after reading a record - Decisions currently living in Slack threads about what a record means - Written process documentation that was stuck in a long text field - The reports somebody assembles by hand from several views ## The split migration Three stages, and the base is never touched. 1. **Identify the one view that is really a to-do list** — There is usually exactly one: a grid or kanban filtered to open items with an assignee field. That view is the only part of Airtable a work tool should take over. Everything else in the base is either data or an application. 2. **Rebuild that view as a workstream, and stop syncing it** — Create the lanes, create the live tasks with owners and dates, and then decide deliberately that the Airtable view no longer needs to reflect them. Two systems trying to hold the same task state without a sync is worse than either one alone, and there is no sync between Airtable and Polaris to save you. 3. **Connect Slack and give a worker the reporting job** — Slack connects once for the whole organisation and feeds the Inbox with prefilled task suggestions. The manual report that somebody assembles from several Airtable views each week becomes a standing task for an AI worker, delivered as a comment with a file, closed by a person. > **The failure mode to avoid** > > Teams that try to move a relational base into a task tool end up with tasks that have lost their relationships and a spreadsheet on the side that quietly becomes the real system. If your Airtable base has more than a handful of linked tables, treat this migration as out of scope and connect Slack instead. Airtable is not in the Polaris connection catalog, so the base stays a separate system either way. ## Read next - [alternatives/airtable](https://www.polarishq.co/alternatives/airtable) - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [cost/slack-pricing](https://www.polarishq.co/cost/slack-pricing) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [replace/notion-and-slack](https://www.polarishq.co/replace/notion-and-slack) - [replace/clickup-and-slack](https://www.polarishq.co/replace/clickup-and-slack) ## Questions people ask **Can Polaris import an Airtable base?** No. There is no Airtable importer and Airtable is not in the connection catalog, so neither records nor structure can be brought across. Airtable exports tables to CSV, which loses every link between them. The recommendation on this page is to keep the base rather than to attempt a move. **What happens to linked records and rollups?** They have no equivalent. Polaris organises work as workstreams with lanes and tasks, and has no relational model, no lookups and no formula fields. A base whose value comes from relationships between tables should stay in Airtable, and any plan that says otherwise is asking you to lose data structure for no gain. **Can an AI worker read our Airtable base?** Not directly. The connection catalog is fixed and Airtable is not on it. The catalog covers Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web search. Work that depends on base contents means somebody pastes the relevant records into the task. **So what does this stack actually gain?** A place where human work has an owner and a date, chat that sits next to that work, and AI workers that deliver finished output as comments with files. The database stays exactly as it is. Whether that is worth adopting a second tool depends on how much of your team's week is coordination rather than data entry. **Does keeping Airtable mean I still pay for it?** Yes, and this page is not pretending otherwise. Polaris software is free with no seat count, so the addition costs nothing, and billing only starts at roughly two dollars per human-equivalent hour when an AI worker delivers work. The Airtable line on your invoice stays, because the base is worth what it costs. ## Related - https://www.polarishq.co/alternatives/airtable - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/replace/clickup-and-slack - https://www.polarishq.co/compare/notion-vs-airtable --- --- title: "Replace Confluence and Notion: paying twice to write" description: "Two documentation tools usually means a migration that stalled. What imports from Notion, what a Confluence export really gives you, and how to finish it." url: https://www.polarishq.co/replace/confluence-and-notion section: Replace your stack updated: 2026-08-21 --- # Replacing Confluence and Notion Two subscriptions for writing things down, which almost always means one migration that never got finished. ## The short answer A team paying for both Confluence and Notion is usually mid-migration, with old documentation in one and new documentation in the other and no rule about which is authoritative. Notion pages import into Polaris Docs within fixed caps. Confluence has no importer and no connector, so its spaces export as XML, HTML or PDF and living pages get rewritten rather than converted. - **Notion:** Imports, within caps - **Confluence:** No importer. Export or rewrite - **Real gap in this stack:** No tracker at all ## Nobody chooses two documentation tools This combination is almost never a decision. It is a state. Either engineering was on Confluence and a newer team started in Notion, or a migration to Notion was announced, got two thirds done, and then the person driving it changed jobs. The result is a documentation estate with no authoritative half. New people are told the handbook is in Notion and the technical documentation is in Confluence, which is true until it is not. Search has to be run twice. Links point across a boundary that neither product knows exists. The important observation about this stack, though, is what is not in it. There is no tracker. Two writing tools and nothing that holds a task with an owner and a date, which means the work items are scattered through page bodies as checkboxes nobody reviews. ## Finishing the migration you already started Three decisions, in this order. - **Decide which half is dead** — Sort both estates by last edit date. Pages untouched for a year are archive, not documentation, and archive does not need to be migrated anywhere. In most stalled migrations this is well over half of the total and finishing suddenly looks achievable. - **Import the Notion half and rewrite the Confluence half** — Notion has an importer and Confluence does not, so the effort is asymmetric. Notion pages come across at twenty-five per run with sub-pages three levels deep, capped at sixty pages and fifteen hundred blocks. Confluence spaces export to XML, HTML or PDF for the archive, and the pages people still open get rewritten by hand. - **Add the thing neither tool was** — The checkboxes buried in page bodies become tasks with an owner and a date, in a workstream, with the Focus lane holding whatever must not slip. This is the part that changes how the team works, and it is available only because the consolidation went to a product that holds both docs and tasks. ## What each source gives you The asymmetry between these two is the planning constraint. | Source | Path into Polaris | What is lost | | --- | --- | --- | | Notion pages | Direct import, 25 per run | Images, tables, databases, embeds, properties, comments | | Notion sub-page trees | Import to three levels deep | Anything nested deeper than three levels | | Confluence pages | Manual rewrite | Everything not retyped, so choose carefully | | Confluence space exports | XML, HTML or PDF archive | Not an import path. Storage only | | Confluence page restrictions | No equivalent | The whole restriction model | | Confluence templates and blueprints | No equivalent | Rebuild as a doc you copy | > **The rule that ends the ambiguity** > > Pick one date and declare that from that day, no new page is created in either old tool. Do not try to finish moving history first. A documentation estate becomes authoritative the moment new writing has one destination, and it stays split forever if you sequence it the other way round. ## Read next - [alternatives/confluence](https://www.polarishq.co/alternatives/confluence) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [compare/notion-vs-confluence](https://www.polarishq.co/compare/notion-vs-confluence) - [cost/confluence-pricing](https://www.polarishq.co/cost/confluence-pricing) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [replace/confluence-and-jira](https://www.polarishq.co/replace/confluence-and-jira) ## Questions people ask **Can Confluence pages be imported into Polaris?** No. Confluence has no Polaris importer and is not in the connection catalog. Confluence exports a space as XML, HTML or PDF, which is an archive rather than an import path. Pages people still read get rewritten by hand, which for a stalled migration is usually a smaller set than anyone expects. **How much of Notion comes across?** Paragraphs, three heading levels, bulleted and numbered lists, to-dos with their checked state, quotes, code, callouts, dividers and toggles, plus sub-pages three levels deep. Images, files, embeds, tables, databases, synced blocks and columns are skipped and counted. Page properties, comments and version history do not come across at all. **Which tool should we stop writing in first?** Both, on the same day. The failure mode in this stack is a phased approach where new writing continues in one old tool while history moves out of the other, which produces three destinations instead of two. Freeze new page creation in both, then move history at whatever pace suits you. **We use Jira with Confluence. Does that change the plan?** Substantially, because a Confluence page that renders a live Jira issue table is a query rather than a document and will not survive an export as living data. If Jira is in your stack, read the Confluence and Jira plan first, since the macro inventory has to happen before anything else. **What does one workspace cost compared to two documentation subscriptions?** Confluence and Notion are both per-seat and you are currently paying both for the same job. Polaris software is free with no seat count, and bills roughly two dollars per human-equivalent hour when an AI worker delivers work. For a team paying twice to write things down, the subscription line goes to zero. ## Related - https://www.polarishq.co/alternatives/confluence - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/replace/confluence-and-jira - https://www.polarishq.co/replace/notion-and-jira --- --- title: "Replace Jira, Confluence and Slack: the enterprise path" description: "Only one of these three connects to Polaris. Why Slack is the beachhead, how to stage a move by team rather than by tool, and what governance blocks it." url: https://www.polarishq.co/replace/jira-and-confluence-and-slack section: Replace your stack updated: 2026-08-21 --- # Replacing Jira, Confluence and Slack An Atlassian estate with a chat tool bolted to it, and exactly one piece of that you can move this quarter. ## The short answer Jira, Confluence and Slack form the standard enterprise stack, and only Slack is in the Polaris connection catalog. Neither Atlassian product has an importer, so both stay in place while a single team runs work in parallel. Staging by team rather than by tool is what makes this survivable, because a tool-by-tool cutover hits permission schemes, deploy gates and retention policy in the first fortnight. - **Connects today:** Slack only - **Atlassian importers:** None for either product - **Staging unit:** One team, not one tool - **Polaris status:** Free public beta ## Why tool-by-tool fails here and team-by-team does not The instinct with three tools is to move the easiest one first, and in a smaller company that instinct is right. In an enterprise estate it is wrong, because each of these three products is wired into things outside itself. Jira transitions gate deploys. Confluence spaces carry restrictions somebody committed to in an audit. Slack has retention policy, compliance export and shared channels with customers. Move one tool across the whole company and you inherit all of that on day one. Move one team across all three and you inherit only that team's share, which is usually close to none if you choose the team carefully. So the unit of migration here is a team with no compliance surface, running everything in parallel while the estate carries on untouched. That is a slower story than most migration pages tell, and it is the one that survives contact with an administrator. ## The governance questions to answer before you start Every one of these has a real answer, and some of the answers are blocking. - **How do people sign in** — Polaris uses passwordless email codes and has no password path at all. There is no SAML SSO and no directory sync. In an organisation whose identity runs through a central provider, this is the question that decides whether a pilot is even permitted. - **Where does the data sit and who can read it** — Polaris runs on Supabase Postgres with row-level security, and connection credentials are stored server-side so agents can use them and browsers cannot read them back. There is no equivalent of Confluence space restrictions or Jira permission schemes. - **What does an AI worker do with a connection** — A worker's capabilities are a SKILL.md file you can open, read and edit, not a hidden prompt. Its tool access comes from the org's connection catalog, authorised once. Its work is logged job by job, and the bill is itemised on that log and challengeable from it. - **What is the vendor risk** — Polaris is in free public beta. No customers to reference, no certifications to cite, no enterprise agreement. Bring this up first rather than letting a security reviewer find it, because a pilot proposed honestly gets a different answer from one that gets discovered. ## Staging by team Five stages. The estate is untouched until the fourth. 1. **Connect Slack for the pilot team's channels only** — Slack is the one free move in this stack. Messages arrive in the Polaris Inbox as prefilled task suggestions and a person clicks to make each one a task. Nothing about the Slack workspace itself changes, and nothing is exported. 2. **Run the pilot team's new work in Polaris, old work in Jira** — No migration, no CSV. Work started after the pilot begins is created in Polaris. Everything already in flight finishes in Jira. Within a sprint the split is obvious and within two the old tool is mostly quiet. 3. **Write the team's living documentation fresh** — Do not convert a Confluence space export. Write the ten pages this team actually reads, in Polaris Docs, and inventory which of their existing pages are Jira macros rather than documents, because those never export as what they look like. 4. **Hire workers and measure delivered hours** — This is the only stage that produces evidence an executive cares about. Give workers the connections they need, assign real work, and read the work log. Each entry carries the estimated human-equivalent hours from an open formula, so the pilot produces a defensible number rather than a sentiment. 5. **Decide about seats, one team at a time** — Only after a team has run in parallel for a quarter do you drop its Jira and Confluence write seats. The estate stays readable. Nothing in this plan requires cancelling an Atlassian contract, and any plan that did would not get approved. > **The three blockers that end this project early** > > A deploy pipeline that transitions Jira issues, a compliance report built on Confluence page history, and an identity requirement for SAML SSO. Polaris has no answer to any of the three. Find out which of them apply to your pilot team in week one, not week six. ## Read next - [replace/confluence-and-jira](https://www.polarishq.co/replace/confluence-and-jira) - [replace/jira-and-slack](https://www.polarishq.co/replace/jira-and-slack) - [replace/notion-and-jira-and-slack](https://www.polarishq.co/replace/notion-and-jira-and-slack) - [cost/stack-cost-50-person-team](https://www.polarishq.co/cost/stack-cost-50-person-team) - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [cost/confluence-pricing](https://www.polarishq.co/cost/confluence-pricing) - [integrations/slack](https://www.polarishq.co/integrations/slack) ## Questions people ask **Which of these three can Polaris actually connect to?** Slack, and only Slack. Jira and Confluence are not in the connection catalog and neither has an importer. The catalog is fixed: Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web search. **Does Polaris support SSO for an enterprise directory?** No. Sign-in is passwordless email codes and there is no password path or SAML support in the product. For organisations that mandate SSO on every application, this is a blocking difference and it should be raised before a pilot is scoped rather than after. **Can we keep Jira for engineering and use Polaris elsewhere?** Yes, and for a large estate that is often the sensible steady state. Polaris software is free with no seat count, so running it alongside Atlassian costs nothing on the licence line. Engineering keeps its workflow schemes and deploy gates while other teams get docs, tasks and AI workers in one place. **How do we justify this to a finance team?** With the work log rather than the licence saving. Each AI worker job records estimated human-equivalent hours from an open formula, billed at roughly two dollars an hour, and any line can be challenged from the log itself. A quarter of pilot data gives finance a delivered-work number, which is a stronger case than a subscription comparison during a period when you are paying for both. **What happens to Confluence pages that embed live Jira data?** They do not survive an export as living content. A page rendering an open-issues table is a query, and exported it becomes a snapshot or nothing. Inventory these first, and decide per page whether it becomes written documentation, a Polaris lane, or stays in Confluence permanently. ## Related - https://www.polarishq.co/replace/confluence-and-jira - https://www.polarishq.co/replace/jira-and-slack - https://www.polarishq.co/replace/notion-and-jira-and-slack - https://www.polarishq.co/cost/stack-cost-50-person-team - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail --- --- title: "Replace Linear, Notion and GitHub: connect all three" description: "All three are in the Polaris connection catalog, so nothing here needs migrating. What connecting them gives an AI worker, and the seam it closes." url: https://www.polarishq.co/replace/linear-and-notion-and-github section: Replace your stack updated: 2026-08-21 --- # Replacing Linear, Notion and GitHub Three tools a technical team chose on purpose, and none of them has to move anywhere. ## The short answer Linear, Notion and GitHub are all in the Polaris connection catalog, so this combination requires no export, no rebuild and no cutover. Each is authorised once for the organisation with credentials stored server-side. An AI worker given all three plus web search can read the issue, the specification and the repository, then deliver finished work as a comment with files attached. - **Migration required:** None - **Connections used:** Linear, Notion, GitHub, web search - **Hire a worker:** ~60 seconds, in chat - **Billing:** ~$2 per human-hour delivered ## The stack is fine. The agent situation is not. A team on Linear, Notion and GitHub has made good choices and does not need a project management lecture. Issues get updated, the repository is the truth about what shipped, and the documentation is written in a tool people enjoy using. What that team usually also has is several engineers running coding agents on their own laptops. Those agents are fast and genuinely useful and completely invisible. A teammate cannot see what one is doing, cannot assign it anything, and cannot read what it produced. When the laptop closes, the work stops. That is the gap this consolidation addresses, and it is not a tracking gap. It is that the most productive tool on the team is single-player and nobody else can reach it. ## The seam between three tools that are each individually correct - **The merged pull request and the stale design doc** — GitHub knows the implementation changed. Linear knows the issue closed. The Notion page describing the design still says what was planned in March, and nothing on either side has any reason to tell it otherwise. - **Context assembled by hand, every time** — Anyone picking up unfamiliar work opens three tabs and reconstructs the story: what the issue asks, what the doc intended, what the code actually does. That reconstruction happens fresh for every person and is thrown away afterwards. ## There is no migration, so this is the whole plan 1. **Authorise the three connections** — Linear, Notion and GitHub, each connected once for the whole organisation. Credentials are verified live and stored server-side, so workers can use them and browsers cannot read them back. Nobody's daily tools change. 2. **Hire a worker and read its SKILL.md** — Describe what keeps slipping. The Chief of Staff runs a short interview where every answer is a click, then the capabilities become a real SKILL.md file. Open it. Edit it. This is the part that convinces technical teams, because the worker's behaviour is a file under your control rather than an opaque prompt. 3. **Assign work that spans all three tools** — Reconciling a doc against a merged change, drafting release notes from a milestone, triaging incoming issues against existing ones. A cloud machine wakes for the task, runs a live tool loop with real web search, ticks its own acceptance criteria and posts the result as a comment with files. An engineer reviews and closes it. ## Local agent against team worker The comparison this stack's readers actually care about. **An agent on a laptop** - One person can see it and nobody can assign to it - Stops when the machine sleeps or the terminal closes - Produces output in a directory somebody has to share manually - No shared record of what was tried, what worked, or what it cost **A worker on the roster** - Sits in the same members table as your people, with the same assignment flow - A cloud machine wakes per task and keeps going after everyone logs off - Delivers as a comment on the task, with generated files attached - Ticks acceptance criteria as it goes, and never closes its own task ## Read next - [integrations/github](https://www.polarishq.co/integrations/github) - [integrations/linear](https://www.polarishq.co/integrations/linear) - [integrations/notion](https://www.polarishq.co/integrations/notion) - [cloud-claude-code/claude-code-for-teams](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) - [cloud-claude-code/cloud-agents-vs-local-agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) - [replace/linear-and-notion](https://www.polarishq.co/replace/linear-and-notion) - [replace/notion-and-linear-and-slack](https://www.polarishq.co/replace/notion-and-linear-and-slack) ## Questions people ask **Do I have to move issues out of Linear or code out of GitHub?** Neither. Both are in the connection catalog and stay exactly where they are. This is the one combination in this cluster with no export, no rebuild and no cutover, which also means there is very little to undo if you decide against it. **What can an AI worker actually do with the GitHub connection?** Its tool access comes from the org's connection catalog, and what it does with that access is defined in its SKILL.md, a file you can read and edit. Connections are authorised once, verified live, and stored server-side. Every job the worker runs is recorded on a work log with the estimated hours that were billed. **How is this different from running Claude Code locally?** A local agent is one person's tool: invisible to teammates, unassignable, and stopped when the laptop closes. A Polaris worker is a row in the same members table as your people, gets assigned tasks the same way, runs on a cloud machine that wakes per task, and delivers its output as a comment with files that anyone on the team can read. **Can I import my Notion docs as well as connecting them?** Yes, and the two are separate. Connecting lets workers read Notion through stored credentials. Importing copies pages into Polaris Docs, twenty-five per run with sub-pages three levels deep, capped at sixty pages and fifteen hundred blocks, skipping images, tables, databases and embeds. **What does it cost to try this?** Nothing until a worker delivers something. The software is free with unlimited people, tasks, workstreams and docs, and billing is roughly two dollars per human-equivalent hour of delivered work, estimated by an open formula and logged job by job. Connecting three tools and hiring a worker costs zero until you assign it a task. ## Related - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/replace/linear-and-notion - https://www.polarishq.co/replace/notion-and-linear-and-slack - https://www.polarishq.co/for/technical-founders --- --- title: "Replace Asana and Notion: the brief written twice" description: "Asana project briefs and Notion pages describe the same project in two products. What exports, what does not, and the chat layer this stack never had." url: https://www.polarishq.co/replace/asana-and-notion section: Replace your stack updated: 2026-08-21 --- # Replacing Asana and Notion A project brief in the tracker, a project page in the doc tool, and a client who has read one of them. ## The short answer Asana and Notion overlap on purpose: Asana projects carry briefs and descriptions, Notion pages carry the same context in more detail, and neither is authoritative. Asana has no Polaris importer or connector, so projects export to CSV and are rebuilt as workstreams. Notion pages import within fixed caps. Neither tool provides team chat, which consolidation adds. - **Asana:** CSV export, manual rebuild - **Notion:** Imports, 25 pages per run - **Gained:** Team chat, in the same place as the work ## Two tools that both think they hold the brief Asana projects have an overview with a brief, a description field, and often a set of tasks whose descriptions carry more context than anyone intended them to. Notion has a page for the same project, with the research, the decisions and the version the client saw. Neither is wrong and neither is complete. The account manager works from the Notion page. The person doing the delivery works from the Asana task. When the scope changes, one of the two gets updated, and which one depends entirely on where the conversation happened. The other thing worth naming is the absence. There is no chat tool in this stack. Coordination runs through email, through Asana task comments, and through direct messages on whatever people already have open, which means half the decisions about a project are in an inbox that nobody else can search. ## Where the pieces end up Notion is the half with an import path. Asana is the half with the work. | Source | Destination | How it gets there | | --- | --- | --- | | Notion project pages | Docs, nested | Imports. Images, tables and databases are skipped | | Asana projects and sections | Workstreams with lanes | Rebuilt by hand from a CSV export | | Asana tasks with owner and date | Tasks in lanes | Rebuilt. Fast, because the data is already clean | | Asana custom fields | Labels | Categorical fields only. Numeric and formula fields do not move | | Asana portfolios and workload | Nothing equivalent | Does not move. Assign the reporting to a worker instead | | Asana project briefs | Merged into the Docs page | Manual, and this merge is the point of the migration | | Email and direct-message coordination | Team chat | New capability. Nothing to migrate | ## A four-stage move for a client-services team Sequenced so that no live client project is ever mid-migration. 1. **Import Notion first and read the skipped count** — Twenty-five pages per run, sub-pages three levels deep, sixty pages and fifteen hundred blocks in total. If the skipped-block count comes back high, your project pages were mostly screenshots and embedded tables, which tells you how much of the brief was never text in the first place. 2. **Merge each Asana brief into its Notion page** — Do this before touching tasks. One page per project, holding everything, with one date at the top saying when it was last true. This single act removes the ambiguity that made the stack confusing, and it is worth doing even if you stop the migration here. 3. **Rebuild only projects that started this quarter** — Export everything from Asana to CSV for the archive. Recreate as workstreams only the projects with live delivery, and let finished client work stay in Asana read-only. Client projects that ended do not need to exist twice. 4. **Move coordination into chat and hire for the reporting** — The status emails become messages next to the work. The weekly client update that somebody assembles by hand becomes a standing task for an AI worker with the connections it needs, delivered as a comment with a file, closed by the account manager after reading it. > **The test for whether this move is worth it** > > Ask one delivery person and one account manager where the current scope of a live project is written. If they name different places, this migration will pay for itself in avoided rework. If they name the same place, your team has already solved the duplication by convention and the only remaining argument is the bill and the AI workers. ## Read next - [alternatives/asana](https://www.polarishq.co/alternatives/asana) - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [cost/asana-pricing](https://www.polarishq.co/cost/asana-pricing) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [replace/asana-and-slack](https://www.polarishq.co/replace/asana-and-slack) - [replace/notion-and-trello](https://www.polarishq.co/replace/notion-and-trello) - [for/agencies](https://www.polarishq.co/for/agencies) ## Questions people ask **Which tool do we move first?** Notion, because it is the only one of the two with an importer and it takes minutes rather than days. Once the pages are in Polaris Docs, merge each Asana project brief into the matching page. Only then rebuild tasks, so the tasks get created from a single agreed description rather than two competing ones. **Does anything import from Asana?** No. Asana is not in the Polaris connection catalog and there is no importer. Asana exports projects to CSV and JSON, which serves as an archive. Live projects are rebuilt as workstreams with lanes, and for a client-services team that usually means only the current quarter's work. **We coordinate by email today. Is adding chat actually an improvement?** It is when the decision needs to sit next to the work rather than in one person's inbox. This stack loses scope changes to email threads that only two people can see. Whether that is worth changing how your team communicates depends on how often somebody has had to forward a thread to explain what was agreed. **What about client access?** Polaris software is free with no seat count, so adding people costs nothing regardless of how much they read. On a per-seat tracker, client-facing collaborators are one of the more expensive line items, which is why agencies often keep clients out of the tool entirely and end up back in email. **Can an AI worker write our weekly client update?** It can be assigned that as a standing task with acceptance criteria, given the connections it needs from the catalog. A cloud machine wakes for the task, works the tool loop, ticks the criteria as it goes and delivers the update as a comment with a file. A person reads it and closes the task, because the machine never marks its own work done. ## Related - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/replace/asana-and-slack - https://www.polarishq.co/replace/notion-and-trello - https://www.polarishq.co/for/agencies --- --- title: "What your work stack costs — 16 pricing pages, Aug 2026" description: "Real list prices for Notion, Jira, Linear, Slack, Asana, ClickUp, monday, Confluence and Trello, checked August 2026, with the stack arithmetic line by line." url: https://www.polarishq.co/cost section: Cost updated: 2026-08-21 --- # What your work stack actually costs Sixteen pages of pricing arithmetic you can check yourself, with the billing basis named on every row. ## The short answer Polaris costs nothing to run and bills roughly two dollars per human-equivalent hour an AI worker delivers. A five-person stack of Notion, Slack and Trello runs about $1,875 a year at the list prices checked in August 2026, and a fifty-person stack with AI subscriptions passes $37,000. These pages show the arithmetic, name the billing basis, and label every estimate. - **Pricing pages:** 16 - **Polaris software:** $0, no seats - **Polaris billing:** ~$2 per human-hour delivered - **Prices checked:** 21 August 2026 ## Why most pricing comparisons are wrong Vendors quote the annual-billing rate in large type and the monthly rate behind a toggle. The gap is real money: Slack Pro is $7.25 per active user per month on annual billing and $8.75 on monthly, and Asana Starter is $10.99 annual against $13.49 monthly. A comparison page that mixes the two produces a number that matches nobody's invoice. So every row in this cluster carries its billing basis. Where a vendor shows only the annual rate, the page says only the annual rate. Where a vendor publishes no list price at all, the page says that too, and quotes nothing. ## Four stacks, four totals List prices checked 21 August 2026, arithmetic shown in full on each page. - **$1,875** — 5 people / year. Notion Plus, Slack Pro, Trello Premium, monthly billing - **$4,650** — 10 people / year. Notion Business, Slack Pro, Linear Basic - **$15,447** — 25 people / year. Notion Business, Slack Business+, Asana Starter, monthly billing - **$37,200** — 50 people / year. Same three plus Claude Team and Copilot seats ## How these pages are built The same discipline on all sixteen. - **The vendor's own page, or nothing** — Every figure came from the vendor's pricing page on 21 August 2026. Atlassian publishes no readable per-user list price for Jira or Confluence, so those two pages quote none and explain the calculator instead. - **The multiplication is on the page** — Rate, billing basis, seat count, monthly subtotal, annual subtotal. You can check a stack-cost page line by line without leaving it. - **Polaris usage numbers are labelled illustrative** — Polaris is in free public beta with no customers and no usage data. Where a page models a monthly bill it states the assumption out loud, in the sentence before the number. - **The hours formula is printed, not summarised** — Base pickup time, searches, finished prose, checklist items, comments and files, with the clamps. It is on the page so you can argue with it. ## Two things a bill can charge for **A place to put work** - Priced per seat, so the bill tracks headcount - A second tool doubles the seat count, a third triples it - AI arrives as another per-seat line on top of the base plan - The invoice never tells you what got done **Work that came back finished** - Software at zero, with no seat count and no tier - Billing starts when an AI worker delivers something - Hours estimated by a published formula, logged job by job - Nothing delivered means nothing billed > **Where Polaris actually is** > > Free public beta. No customers, no revenue, no case studies. The product works and the demos are real recordings, one of them an unedited machine session sped up. Any monthly figure attached to Polaris on these pages is arithmetic on a stated assumption, not a measurement. ## Every cost page - [Notion pricing, checked August 2026](https://www.polarishq.co/cost/notion-pricing) — Two seat prices and a second meter for AI, which is the part that surprises people on renewal. - [Jira pricing, checked August 2026](https://www.polarishq.co/cost/jira-pricing) — The only vendor in this cluster whose per-user price you cannot read off a page. - [Linear pricing, checked August 2026](https://www.polarishq.co/cost/linear-pricing) — A generous free tier with two hard caps, and paid rates published on yearly billing only. - [Slack pricing, checked August 2026](https://www.polarishq.co/cost/slack-pricing) — The one vendor here that bills per active user, which helps more than most people realise. - [Asana pricing, checked August 2026](https://www.polarishq.co/cost/asana-pricing) — One of the few vendors that prints both billing rates, and the gap is twenty-three percent. - [ClickUp pricing, checked August 2026](https://www.polarishq.co/cost/clickup-pricing) — Cheap seats with an AI add-on that can cost more than the seat it sits on. - [monday.com pricing, checked August 2026](https://www.polarishq.co/cost/monday-pricing) — Seats sold in blocks, AI sold in credits, and four separate products under one brand. - [Confluence pricing, checked August 2026](https://www.polarishq.co/cost/confluence-pricing) — Priced by a function rather than a number, and bundled with the rest of Atlassian by design. - [Trello pricing, checked August 2026](https://www.polarishq.co/cost/trello-pricing) — The cheapest tracker on this list, and the one teams outgrow fastest. - [Tool stack cost for a 5-person team](https://www.polarishq.co/cost/stack-cost-5-person-team) — Three subscriptions, five people, and the arithmetic laid out so you can check it against your own invoices. - [Tool stack cost for a 10-person team](https://www.polarishq.co/cost/stack-cost-10-person-team) — The size where the free tiers run out on all three products at roughly the same time. - [Tool stack cost for a 25-person team](https://www.polarishq.co/cost/stack-cost-25-person-team) — The size where paying monthly instead of annually costs $1,650 a year on its own. - [Tool stack cost for a 50-person team](https://www.polarishq.co/cost/stack-cost-50-person-team) — Three work tools and two AI subscriptions, which is where most companies of this size actually are. - [Per-seat pricing and usage pricing charge for different things](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) — One model prices access. The other prices output. Almost every argument about software cost is really about which of those you are buying. - [The cost of AI subscriptions for a team](https://www.polarishq.co/cost/cost-of-ai-subscriptions) — Five figures a year at fifty people, and no shared record of what any of it produced. - [What an AI worker costs, and how the hours are counted](https://www.polarishq.co/cost/what-an-ai-worker-costs) — The whole formula is on this page, including the parts that make it an estimate rather than a measurement. ## Related reading - [alternatives](https://www.polarishq.co/alternatives) - [replace](https://www.polarishq.co/replace) - [glossary/per-seat-pricing](https://www.polarishq.co/glossary/per-seat-pricing) - [glossary/usage-based-pricing](https://www.polarishq.co/glossary/usage-based-pricing) - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [glossary/tool-sprawl](https://www.polarishq.co/glossary/tool-sprawl) ## Questions people ask **Is the Polaris software really free, or is there a seat charge hiding somewhere?** There is no seat charge. Unlimited humans, tasks, workstreams, docs, connections and the Chief of Staff are included in every organisation at no cost. The only revenue line is roughly two dollars per human-equivalent hour that an AI worker delivers, and nothing delivered means nothing billed. **How do I know the hours are not inflated?** The formula is published and every job writes its own inputs to the work log: how many searches ran, how many characters of finished prose came back, how many checklist items were ticked, how many comments and files were produced. You can recompute any line yourself and challenge it. That is the whole reason the formula is simple rather than clever. **What happens if an AI worker does bad work?** You do not close the task. Agents deliver and humans close, always, in both directions of the product: the machine posts its work as a comment and ticks acceptance criteria, and a person decides whether that counts. You close the task and rate it, or you send it back. **Why do your competitor prices differ from other comparison sites?** Most comparison pages quote the annual-billing rate as if it were the monthly price, which understates a monthly-billed invoice by twenty to twenty-five percent on several of these vendors. Every row here names its billing basis, and the check date is 21 August 2026. **Can I keep the tools I already pay for?** Yes, and for most teams that is the sensible first move. Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp and Instagram are all in the Polaris connection catalog, so an AI worker can use them without anyone migrating anything. The saving only appears when you stop paying for a seat somewhere. ## Related - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/alternatives - https://www.polarishq.co/replace - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/glossary/human-equivalent-hours --- --- title: "Notion pricing 2026: $10 Plus, $20 Business, plus AI credits" description: "Notion's Free, Plus, Business and Enterprise plans with the credit-based AI pricing on top, the ten-person arithmetic, and what the same work costs on Polaris." url: https://www.polarishq.co/cost/notion-pricing section: Cost updated: 2026-08-21 --- # Notion pricing, checked August 2026 Two seat prices and a second meter for AI, which is the part that surprises people on renewal. ## The short answer Notion lists Free at zero, Plus at $10 per member per month and Business at $20 per member per month on notion.com/pricing, checked 21 August 2026, with up to twenty percent off for yearly billing. AI is metered separately: Custom Agents are free to try, then $10 per 1,000 monthly Notion credits. Polaris charges nothing for software. - **Plus:** $10 / member / month - **Business:** $20 / member / month - **AI credits:** $10 per 1,000 monthly - **Checked:** 21 August 2026 ## Notion plans as listed Read from notion.com/pricing on 21 August 2026. The page shows these per-member rates with a yearly toggle marked "Save up to 20% with yearly" and does not print a separate yearly figure. | Plan | Per member / month | What it adds | | --- | --- | --- | | Free | $0 | Trial of Notion AI, up to 10 guests, 5MB file uploads, 7-day page history, block limit once a workspace has two or more members | | Plus | $10 | Unlimited file uploads, unlimited blocks, custom forms and sites, unlimited charts, basic connections | | Business | $20 | Notion Agent, AI Meeting Notes, Enterprise Search in beta, SAML SSO, private teamspaces, premium connections | | Enterprise | Custom | Zero data retention with LLM providers, audit logs, DLP and SIEM connections, customer success manager | ## The credit meter is the part to model Notion credits are an add-on for Business and Enterprise that pay for Custom Agents, Workers and AI usage beyond the plan allowance. Monthly credits cost $10 per 1,000 and reset each cycle with no carryover. Annual credits cost $13 per 1,000 and stay available until renewal. Notion's own help page states both figures. Workers, Notion's agent product, is listed as free to try and begins consuming credits on 15 October. So a Business workspace has one predictable line and one variable line, and the variable line is the one nobody forecasts correctly in the first quarter. ## Ten people on Notion Business, with credits Arithmetic at the listed rates. The credit row is an assumption, marked as such, not a Notion quote. | Line | Rate | Quantity | Per month | Per year | | --- | --- | --- | --- | --- | | Notion Business seats | $20 / member / month | 10 members | $200.00 | $2,400.00 | | AI credits (assumed 3,000 / month) | $10 per 1,000 monthly credits | 3,000 credits | $30.00 | $360.00 | | Total | | | $230.00 | $2,760.00 | | Polaris software | $0 | unlimited members | $0.00 | $0.00 | ## Four things that move a Notion bill - **Members and guests are billed differently** — Notion charges by seats assigned in the workspace. Guests get page-level access and are limited by a plan guest cap rather than billed as members, which is why guest-heavy workspaces look cheap until someone is promoted. - **The Free plan gets tight at two members** — Notion's free tier applies a block limit to workspaces with two or more members, which is the point most small teams discover they are on a paid product. - **Monthly credits expire, annual credits are dearer** — Monthly credits cost less per thousand but reset with no carryover. Annual credits cost $13 per 1,000 and last until renewal. Choosing wrongly costs about thirty percent. - **Notion is a place to write, not a thing that writes** — Agents and credits move that line, and they are worth trying. The base product's job is still to hold what people already produced. ## Notion Business against Polaris **Notion Business** - $20 per member per month, checked 21 August 2026 - AI on a second meter at $10 per 1,000 monthly credits - Docs, databases and Notion Agent in one workspace - Tasks and team chat still live in other subscriptions for most teams **Polaris** - $0 for the software, no seats, no tiers - Docs, tasks and team chat in the same product - AI workers you hire in about sixty seconds, with an editable SKILL.md - Roughly $2 per human-equivalent hour an AI worker delivers, itemised > **Connecting beats migrating** > > Notion is in the Polaris connection catalog. You can authorise it once for the whole organisation and let AI workers read and write in the Notion you already have, without moving a page. Your Notion bill does not change until you decide to stop paying for seats. ## Where to go next - [alternatives/notion](https://www.polarishq.co/alternatives/notion) - [replace/notion-and-jira](https://www.polarishq.co/replace/notion-and-jira) - [replace/notion-and-slack](https://www.polarishq.co/replace/notion-and-slack) - [cost/confluence-pricing](https://www.polarishq.co/cost/confluence-pricing) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [integrations/notion](https://www.polarishq.co/integrations/notion) ## Questions people ask **Is Notion's $10 the monthly or the annual rate?** Notion's pricing page displays $10 for Plus and $20 for Business per member per month and carries a toggle marked "Save up to 20% with yearly" without printing a separate yearly figure. Treat $10 and $20 as the rates shown on 21 August 2026 and check the toggle for your own billing choice before you budget. **Do I need Notion AI credits to use AI in Notion?** Custom Agents are free to try and then consume credits at $10 per 1,000 monthly credits, and Workers is free to try now with credit consumption starting on 15 October. Credits are an add-on for Business and Enterprise workspaces. **Is Polaris really free, or does the cost show up somewhere else?** The software is free with no seat count. The cost shows up only when an AI worker delivers work, at roughly two dollars per human-equivalent hour, and every hour is itemised on that worker's log. If no worker delivers anything in a month, the bill for that month is zero. **How do I know Polaris is not inflating the hours?** The estimate comes from a published formula applied to what the job actually did: base pickup time, web searches at about twelve minutes each, finished prose at about ninety characters a minute, and fixed overheads for checklist items, comments and files. The inputs are on the work log, so you can recompute a line and dispute it. **Can Polaris and Notion run side by side?** Yes. Notion is one of the connections an AI worker can be given, authorised once and stored server-side, so a worker can pull context out of Notion and write results back. Teams usually run both for a while and only cut seats once the habit has moved. ## Related - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/replace/notion-and-jira - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/integrations/notion --- --- title: "Jira pricing 2026: why Atlassian publishes no list price" description: "Jira Cloud's Free plan caps at 10 users, and Atlassian publishes no readable per-user price above it. What the calculator does, and how to budget around it." url: https://www.polarishq.co/cost/jira-pricing section: Cost updated: 2026-08-21 --- # Jira pricing, checked August 2026 The only vendor in this cluster whose per-user price you cannot read off a page. ## The short answer Jira Cloud's Free plan covers up to 10 users. Atlassian publishes no readable per-user list price for Standard or Premium: on 21 August 2026 the Jira pricing page, the Standard page, the Premium page and the licensing page all routed to an interactive calculator instead of a figure. Per-user rates step down as seat count rises, and annual billing discounts vary by seat count. - **Free plan:** Up to 10 users - **Standard / Premium:** Calculator only - **Billing basis:** Highest seat count in the cycle - **Checked:** 21 August 2026 ## What we checked, and what we found On 21 August 2026 we read atlassian.com/software/jira/pricing, atlassian.com/software/jira/standard, atlassian.com/software/jira/premium and atlassian.com/licensing/jira. None of them prints a per-user price. The Standard page says annual subscriptions "may offer a discount depending on the number of users purchased" and sends you to the calculator; the others link to the same calculator. That is not an oversight. Atlassian prices on continuous user tiers, so there is no single number to print. Every other vendor in this cluster publishes a rate. Atlassian publishes a function. We are not going to invent a figure to fill the gap, and you should be sceptical of comparison pages that do. ## What Atlassian does document Facts taken from Atlassian's own support and licensing pages on 21 August 2026. | Fact | Source page | What it means for a budget | | --- | --- | --- | | Free plan supports up to 10 users | Jira Cloud plans, Atlassian Support | The eleventh hire is a pricing event, not a seat | | Standard and Premium support up to 100,000 users | Jira Cloud plans, Atlassian Support | Tier choice is about features and support, not capacity | | Each user's unit price depends on the tier that user falls into | Atlassian licensing | The marginal seat is cheaper than the average seat, so per-user quotes from other teams do not transfer | | The bill uses the highest seat count at any point in the cycle | Atlassian licensing | Removing a user mid-cycle does not reduce that cycle's invoice | | Annual discounts depend on the number of users purchased | Jira Standard plan page | There is no flat annual percentage to plan against | | Eligible community and academic customers get 75% off cloud | Atlassian licensing | Worth checking before any migration maths | ## Three Jira cost traps that are not on the price page - **The user tier, not the user** — Because the rate steps down at thresholds, adding people can lower your per-user cost and still raise your bill. Finance teams that budget by multiplying last quarter's per-user rate by next quarter's headcount get it wrong in both directions. - **Maximum quantity billing** — Your invoice is set by the highest number of seats assigned during the cycle. A contractor added for two weeks is billed as a full seat for the cycle, and deactivating them does not claw it back. - **Jira rarely arrives alone** — Most Jira teams also run Confluence and Slack. Three subscriptions across the same headcount is the normal shape, and it is the shape the stack-cost pages in this cluster model. ## Jira against Polaris **Jira Cloud** - Free to 10 users, then priced by a calculator - A workflow engine most trackers cannot match - Per-user rate falls with scale, total bill rises with headcount - Tracks work precisely and performs none of it **Polaris** - $0 for the software at any headcount - Tasks, docs and team chat in one product - Humans and AI workers are the same kind of teammate, assigned identically - Roughly $2 per human-equivalent hour an AI worker delivers > **When Jira is still the right answer** > > If your organisation runs regulated workflows with conditional transitions, mandatory fields and audited approval gates, Jira's workflow engine is doing real work that a simpler tracker cannot replicate. Price is a bad reason to leave that behind. Sprawl and idle seats are better ones. ## Where to go next - [alternatives/jira](https://www.polarishq.co/alternatives/jira) - [cost/confluence-pricing](https://www.polarishq.co/cost/confluence-pricing) - [replace/jira-and-slack](https://www.polarishq.co/replace/jira-and-slack) - [replace/confluence-and-jira](https://www.polarishq.co/replace/confluence-and-jira) - [compare/jira-vs-linear](https://www.polarishq.co/compare/jira-vs-linear) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) ## Questions people ask **Why does this page not give a Jira price per user?** Because Atlassian does not publish one. On 21 August 2026 every Atlassian page we checked routed to an interactive calculator rather than printing a rate, and the rate itself changes with your seat count. Quoting a single number would be a guess dressed as a fact. **Is Jira free for small teams?** Jira Cloud's Free plan supports up to 10 users, per Atlassian's own plan documentation. Above that you move to Standard or Premium and the calculator decides your rate based on how many seats you buy. **Does removing a Jira user lower this month's bill?** No. Atlassian bills on the maximum number of seats assigned to a product at any point during the billing cycle, so a seat that existed for one day counts for the whole cycle. Mid-cycle additions are prorated, removals are not refunded. **What would the same work cost on Polaris?** The software costs nothing regardless of headcount. Work that an AI worker delivers is metered at roughly two dollars per human-equivalent hour, estimated by a published formula and logged line by line. A team that never assigns anything to an AI worker pays nothing at all. **Can Polaris work alongside Jira rather than replacing it?** Not directly today. Jira is not in the current Polaris connections catalog, which covers Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web research. Teams on Jira usually start Polaris on a separate stream of work rather than mirroring the tracker. ## Related - https://www.polarishq.co/alternatives/jira - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/replace/jira-and-slack - https://www.polarishq.co/replace/confluence-and-jira - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team --- --- title: "Linear pricing 2026: $10 Basic, $16 Business, billed yearly" description: "Linear's Free, Basic, Business and Enterprise plans with the exact yearly-billing rates, the Free plan's issue and team caps, and the ten-person arithmetic." url: https://www.polarishq.co/cost/linear-pricing section: Cost updated: 2026-08-21 --- # Linear pricing, checked August 2026 A generous free tier with two hard caps, and paid rates published on yearly billing only. ## The short answer Linear lists Free at zero with unlimited members but only 2 teams and 250 issues, Basic at $10 per user per month and Business at $16 per user per month, both marked billed yearly on linear.app/pricing, checked 21 August 2026. Enterprise is custom and annual only. Polaris charges nothing for software and meters delivered work instead. - **Basic:** $10 / user / month, yearly - **Business:** $16 / user / month, yearly - **Free plan cap:** 2 teams, 250 issues - **Checked:** 21 August 2026 ## Linear plans as listed Read from linear.app/pricing on 21 August 2026. Both paid rates are shown with the note "Billed yearly"; the page does not print a separate monthly-billing rate, so this page does not quote one. | Plan | Rate | Billing basis | Caps and additions | | --- | --- | --- | --- | | Free | $0 | n/a | Unlimited members, 2 teams, 250 issues, agent platform and Linear Agent included | | Basic | $10 / user / month | Billed yearly | 5 teams, unlimited issues, unlimited file uploads, admin roles | | Business | $16 / user / month | Billed yearly | Unlimited teams, private teams and guests, Triage Intelligence, Loops, Code Intelligence, Insights, Asks, Zendesk and Intercom | | Enterprise | Custom | Annual only | SAML and SCIM, invoice or PO billing, advanced org modeling, migration support | ## The 250-issue ceiling is the real trigger Linear's free tier gives unlimited members, which reads as generous, and then caps you at 2 teams and 250 issues. A working engineering team crosses 250 issues in a couple of months. That is the moment the decision arrives, and it arrives on issue count rather than headcount, which is unusual and worth planning for. Linear also includes its agent platform and Linear Agent on the free plan, with coding sessions and Loops drawing on AI credits. The credit rates were not printed on the pricing page when we checked, so this page does not state them. ## Ten people on Linear, and what the rest of the stack adds Linear rates as listed on yearly billing. Slack is shown at both of its published rates because Slack prints both. | Line | Rate | Basis | Seats | Per month | Per year | | --- | --- | --- | --- | --- | --- | | Linear Basic | $10 / user / month | Yearly | 10 | $100.00 | $1,200.00 | | Linear Business | $16 / user / month | Yearly | 10 | $160.00 | $1,920.00 | | Slack Pro (annual) | $7.25 / active user / month | Annual | 10 | $72.50 | $870.00 | | Slack Pro (monthly) | $8.75 / active user / month | Monthly | 10 | $87.50 | $1,050.00 | | Linear Basic + Slack Pro, annual basis | | | 10 | $172.50 | $2,070.00 | | Polaris software | $0 | n/a | unlimited | $0.00 | $0.00 | ## Where Linear earns its money, honestly - **Cycle discipline** — Linear's opinionated cycles, triage and keyboard-first speed are the reason engineers defend it in tool reviews. That is a real product advantage, not a marketing line, and it is the main reason not to leave. - **The team cap, not the seat price** — Basic allows 5 teams. Companies that model teams as squads hit that ceiling long before they mind the $10, which pushes them to Business at $16 and a sixty percent jump in the seat line. - **It still only tracks** — Linear is a superb record of what should happen. The work itself is done by the people whose names are on the issues, and every one of those people is a seat. ## Linear Business against Polaris **Linear Business** - $16 per user per month, billed yearly, checked 21 August 2026 - Unlimited teams, private teams and guests - Issue tracking that engineers actually enjoy - Docs and team chat live in other subscriptions **Polaris** - $0 for the software, unlimited humans and workstreams - Tasks, docs and team chat in one place - Assign a task to an AI worker and a cloud machine wakes for it - Roughly $2 per human-equivalent hour delivered, challengeable line by line > **Linear is a connection, not a casualty** > > Linear is in the Polaris connections catalog. Authorise it once for the organisation and an AI worker can read and write Linear issues while your engineers keep the tool they like. Nothing about your Linear bill changes until you choose to change it. ## Where to go next - [alternatives/linear](https://www.polarishq.co/alternatives/linear) - [compare/jira-vs-linear](https://www.polarishq.co/compare/jira-vs-linear) - [replace/linear-and-slack](https://www.polarishq.co/replace/linear-and-slack) - [replace/linear-and-notion](https://www.polarishq.co/replace/linear-and-notion) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [integrations/linear](https://www.polarishq.co/integrations/linear) ## Questions people ask **What does Linear cost per month if I do not want annual billing?** Linear's pricing page showed only the yearly-billing rates when checked on 21 August 2026, so this page does not quote a monthly figure. Assume the monthly rate is higher, as it is with every other vendor in this cluster, and check the page's toggle before budgeting. **How long does Linear's free plan last for a real team?** Until you cross 2 teams or 250 issues, whichever comes first. Issue count is usually the binding constraint, and an active engineering team reaches 250 issues within a few months of normal work. **Does Polaris replace Linear?** It can, and for some teams it should, but connecting is the lower-risk first step. Linear is in the Polaris connections catalog, so AI workers can operate on your existing issues while people keep using Linear. The bill only changes when you stop paying for seats. **What is the catch on two dollars per human-hour?** There is no catch, but there is a limit worth knowing: the hour figure is an estimate from observable effort, not a stopwatch. It comes from a published formula, it is logged per job with its inputs, and you can challenge any line from the work log. That auditability is the point. **Who marks work as done in Polaris?** A person does. An AI worker ticks its acceptance-criteria checklist and posts its deliverable as a comment with any files it produced, but it never closes the task. You close it and rate it, which is also the moment the work counts. ## Related - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/compare/jira-vs-linear - https://www.polarishq.co/compare/linear-vs-shortcut - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/integrations/linear --- --- title: "Slack pricing 2026: $8.75 Pro monthly, $7.25 annual" description: "Slack's Free, Pro, Business+ and Enterprise+ rates on both billing bases, how per-active-user Fair Billing actually works, and the 90-day free history limit." url: https://www.polarishq.co/cost/slack-pricing section: Cost updated: 2026-08-21 --- # Slack pricing, checked August 2026 The one vendor here that bills per active user, which helps more than most people realise. ## The short answer Slack lists Pro at $8.75 per active user per month on monthly billing and $7.25 on annual, and Business+ at $18 monthly and $15 annual, checked on slack.com/pricing on 21 August 2026. Free keeps 90 days of message history. Slack bills only members who took an action within a 28-day period. Polaris includes team chat at no cost. - **Pro:** $8.75 monthly / $7.25 annual - **Business+:** $18 monthly / $15 annual - **Free history:** 90 days - **Checked:** 21 August 2026 ## Slack plans on both billing bases Read from slack.com/pricing on 21 August 2026. Slack prints both rates, which is more than most vendors in this cluster do. | Plan | Monthly billing | Annual billing | Gap | | --- | --- | --- | --- | | Free | $0 | $0 | 90 days of message history, searchable | | Pro | $8.75 / active user / month | $7.25 / active user / month | $1.50 per user per month, 21% more on monthly | | Business+ | $18 / active user / month | $15 / active user / month | $3.00 per user per month, 20% more on monthly | | Enterprise+ | Contact sales | Contact sales | Not published | ## Fair Billing is genuinely fair, and worth using Slack's policy states that members who take an action in Slack at any time within a 28-day period are considered active for billing purposes. If someone you have already paid for goes inactive, Slack adds a prorated credit for the unused time. New members added mid-cycle are charged prorated for the remainder. That makes Slack the cheapest tool in a typical stack to over-provision and the most expensive to under-audit, because dormant accounts cost nothing but half-active ones cost full price. Compare that with Atlassian, which bills on the highest seat count assigned at any point in the cycle whether or not anyone logged in. ## Slack across four team sizes Straight multiplication at the listed rates. Actual invoices will be lower wherever people go inactive for a full 28-day window. | Team size | Pro, monthly basis | Pro, annual basis | Business+, annual basis | Business+ per year | | --- | --- | --- | --- | --- | | 5 active users | $43.75 / month | $36.25 / month | $75.00 / month | $900.00 | | 10 active users | $87.50 / month | $72.50 / month | $150.00 / month | $1,800.00 | | 25 active users | $218.75 / month | $181.25 / month | $375.00 / month | $4,500.00 | | 50 active users | $437.50 / month | $362.50 / month | $750.00 / month | $9,000.00 | ## The three costs a Slack invoice does not show - **The 90-day wall on Free** — Free keeps 90 days of history. Most teams do not upgrade for features; they upgrade the first time a decision from four months ago cannot be found. - **Signal that never becomes work** — A request in a thread is not a task. It is a task-shaped message that decays. Polaris turns a Slack message into a prefilled task suggestion with bucket, lane, labels and owner already set, waiting for one click. Signals become suggestions, never silent tasks. - **Slack sits on top of the other two bills** — Nobody buys Slack instead of a tracker and a docs tool. It is the third subscription across the same headcount, which is why it appears in every stack-cost page in this cluster. ## Slack against Polaris **Slack Pro** - $8.75 per active user per month, monthly billing - Billing by activity, with prorated credits for the inactive - Where the signal lives, and where it dies - Team chat only; tasks and docs are separate subscriptions **Polaris** - Team chat included at $0 alongside tasks and docs - Inbox turns tool signals into prefilled task suggestions - AI workers you can talk to in the same place as your team - Roughly $2 per human-equivalent hour delivered, and nothing else > **Slack is a connection too** > > Slack is in the Polaris connections catalog, authorised once for the whole organisation with credentials stored server-side. That is how the Inbox works: a Slack message arrives as a suggested task with its fields already filled, and a person approves it. ## Where to go next - [alternatives/slack](https://www.polarishq.co/alternatives/slack) - [compare/slack-vs-microsoft-teams](https://www.polarishq.co/compare/slack-vs-microsoft-teams) - [replace/notion-and-slack](https://www.polarishq.co/replace/notion-and-slack) - [replace/jira-and-slack](https://www.polarishq.co/replace/jira-and-slack) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) ## Questions people ask **Does Slack charge for every member of the workspace?** No. Slack's Fair Billing policy counts a member as active if they took any action within a 28-day period, and adds a prorated credit when a paid member goes inactive. That makes a dormant account free and a barely-used account full price. **How much more does Slack cost on monthly billing?** About twenty-one percent more on Pro, at $8.75 against $7.25 per active user per month, and exactly twenty percent more on Business+, at $18 against $15. Both figures were read from slack.com/pricing on 21 August 2026. **Does Polaris replace Slack?** Polaris includes team chat in the free workspace, so it can. Most teams start by connecting Slack instead, which lets the Inbox turn messages into prefilled task suggestions while everyone keeps using Slack. The saving arrives only when seats are actually cancelled. **What does the Polaris Inbox do with a Slack message?** It proposes a task with bucket, lane, labels and owner already filled in, and waits. Nothing is created until a person clicks. Signals become suggestions, not silent tasks, which is the difference between a helpful Inbox and an unusable one. **If Polaris chat is free, where does the money come from?** From delivered work only, at roughly two dollars per human-equivalent hour, estimated by a published formula and logged job by job on the worker's work log. Chat, tasks, docs, connections and unlimited human seats cost nothing. ## Related - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/compare/slack-vs-microsoft-teams - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/replace/linear-and-slack - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/glossary/per-seat-pricing --- --- title: "Asana pricing 2026: $10.99 Starter annual, $13.49 monthly" description: "Asana's Personal, Starter, Advanced and Enterprise rates on both billing bases, the AI Studio credit allowances, and the twenty-five-person arithmetic in full." url: https://www.polarishq.co/cost/asana-pricing section: Cost updated: 2026-08-21 --- # Asana pricing, checked August 2026 One of the few vendors that prints both billing rates, and the gap is twenty-three percent. ## The short answer Asana lists Starter at $10.99 per user per month on annual billing and $13.49 on monthly, and Advanced at $24.99 annual and $30.49 monthly, checked on asana.com/pricing on 21 August 2026. Personal is free with a hard teammate cap. AI Studio credits are included per billing account by tier. Polaris charges nothing for software. - **Starter:** $10.99 annual / $13.49 monthly - **Advanced:** $24.99 annual / $30.49 monthly - **Monthly premium:** About 23% - **Checked:** 21 August 2026 ## Asana plans on both billing bases Read from asana.com/pricing on 21 August 2026. | Plan | Annual billing | Monthly billing | AI Studio credits | Seats | | --- | --- | --- | --- | --- | | Personal | $0 | $0 | n/a | Capped at a small team; the page showed 2 users when checked | | Starter | $10.99 / user / month | $13.49 / user / month | 50K per billing account per month | Unlimited | | Advanced | $24.99 / user / month | $30.49 / user / month | 75K per billing account per month | Unlimited | | Enterprise | Contact sales | Contact sales | 200K per billing account per month | Unlimited | ## The Starter to Advanced jump is the expensive decision Starter to Advanced is $10.99 to $24.99 per user per month on annual billing, a 127 percent increase, and the features that force it are ordinary: portfolios, goals, workload management, and approvals with proofing. Teams do not upgrade to Advanced because they outgrew task management. They upgrade because a director wanted a portfolio view. AI Studio credits are allocated per billing account rather than per user, which means the allowance does not grow with your headcount even though the seat bill does. Asana also sells AI Teammates as an add-on with pricing through sales, so it is not a figure this page can quote. ## Twenty-five people on Asana, both bases Straight multiplication at the listed rates, with the annual saving shown. | Line | Rate | Seats | Per month | Per year | | --- | --- | --- | --- | --- | | Starter, annual billing | $10.99 | 25 | $274.75 | $3,297.00 | | Starter, monthly billing | $13.49 | 25 | $337.25 | $4,047.00 | | Advanced, annual billing | $24.99 | 25 | $624.75 | $7,497.00 | | Advanced, monthly billing | $30.49 | 25 | $762.25 | $9,147.00 | | Cost of paying monthly on Starter | | 25 | $62.50 | $750.00 | | Polaris software | $0 | unlimited | $0.00 | $0.00 | ## Three things worth knowing before renewal - **The free tier is now very small** — Asana's Personal plan showed a cap of 2 users when we checked on 21 August 2026. If you remember Asana Free as a ten-person plan, re-check before you plan around it. - **Credits are pooled, seats are not** — AI Studio credits are stated per billing account per month, so a fifty-person Starter workspace and a five-person Starter workspace get the same 50K allowance while paying ten times the seat bill. - **Paying monthly costs a fifth more** — Twenty-three percent on Starter, twenty-two percent on Advanced. On a twenty-five person Starter workspace that is $750 a year for the privilege of not committing. ## Asana Starter against Polaris **Asana Starter** - $10.99 per user per month annual, $13.49 monthly - Timeline, dashboards, unlimited automations, forms - AI credits pooled per billing account, not per person - Docs and team chat live in other subscriptions **Polaris** - $0 software, unlimited humans, no tier to outgrow - Tasks, docs and team chat in one product - AI workers assigned exactly like people, from the same member list - Roughly $2 per human-equivalent hour delivered, itemised per job > **Where Asana still wins** > > Portfolios and workload across dozens of concurrent projects is genuine programme management, and Asana Advanced does it well. If that is the job, this is not the page that talks you out of it. If the job is a team of fifteen tracking work they already understand, the Advanced line is expensive furniture. ## Where to go next - [alternatives/asana](https://www.polarishq.co/alternatives/asana) - [compare/asana-vs-monday](https://www.polarishq.co/compare/asana-vs-monday) - [compare/jira-vs-asana](https://www.polarishq.co/compare/jira-vs-asana) - [replace/asana-and-slack](https://www.polarishq.co/replace/asana-and-slack) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) ## Questions people ask **How much is Asana per user per month?** Starter is $10.99 per user per month on annual billing and $13.49 on monthly. Advanced is $24.99 annual and $30.49 monthly. Both were read from asana.com/pricing on 21 August 2026, and Enterprise pricing is not published. **Is Asana still free for small teams?** There is a free Personal plan, but its teammate cap showed as 2 users when we checked on 21 August 2026. That is much tighter than the free tier many teams remember, so verify it against the current page before planning a rollout on it. **What does Asana's AI cost?** Each paid tier includes an AI Studio credit allowance stated per billing account per month: 50K on Starter, 75K on Advanced and 200K on Enterprise. AI Teammates is a separate add-on priced through sales, so there is no list figure to quote. **What would the same work cost on Polaris?** The software is free at any headcount. Work delivered by an AI worker is metered at roughly two dollars per human-equivalent hour. As an illustration only, a twenty-five person team whose workers delivered 80 human-hours in a month would see about $160 for that month, against $274.75 for Asana Starter seats alone. **How is a Polaris hour calculated?** From observable effort: a base pickup time of fifteen minutes, twelve minutes per web search, finished prose at about ninety characters a minute, eight minutes per acceptance-criteria item ticked, ten minutes per comment addressed, ten minutes per file produced and five per PDF, clamped between five minutes and eight hours per session. Every input is on the work log. ## Related - https://www.polarishq.co/alternatives/asana - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/replace/asana-and-notion - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "ClickUp pricing 2026: $7 Unlimited, $12 Business, plus AI" description: "ClickUp's Unlimited and Business rates on both billing bases, the separate ClickUp Brain and Everything AI add-ons, and what a stacked AI bill comes to." url: https://www.polarishq.co/cost/clickup-pricing section: Cost updated: 2026-08-21 --- # ClickUp pricing, checked August 2026 Cheap seats with an AI add-on that can cost more than the seat it sits on. ## The short answer ClickUp lists Unlimited at $7 per user per month on annual billing and $10 on monthly, and Business at $12 annual and $19 monthly, checked on clickup.com/pricing on 21 August 2026. AI is a separate add-on: Brain AI at $7.20 annual or $9 monthly, Everything AI at $22.40 annual or $28 monthly. Polaris charges nothing for software. - **Unlimited:** $7 annual / $10 monthly - **Business:** $12 annual / $19 monthly - **Brain AI add-on:** $7.20 annual / $9 monthly - **Checked:** 21 August 2026 ## ClickUp plans and AI add-ons Read from clickup.com/pricing on 21 August 2026. The AI tiers are priced per user per month on top of a base plan, not instead of one. | Line | Annual billing | Monthly billing | Monthly premium | | --- | --- | --- | --- | | Free Forever | $0 | $0 | n/a | | Unlimited | $7 / user / month | $10 / user / month | 43% more | | Business | $12 / user / month | $19 / user / month | 58% more | | Enterprise | Custom | Custom | Not published | | Brain AI add-on | $7.20 / user / month | $9 / user / month | 25% more | | Everything AI add-on | $22.40 / user / month | $28 / user / month | 25% more | ## The add-on can exceed the product ClickUp Unlimited is $7 per user per month on annual billing. Everything AI is $22.40 per user per month on the same basis. A team that wants the full AI layer pays roughly three times more for the assistant than for the workspace it assists in, and both lines scale with headcount whether or not a given person uses the AI. That is the clearest example in this cluster of what per-seat AI pricing does. You are billed for access, so the finance question becomes who is allowed to have it rather than what any of it produced. ## Twenty people on ClickUp Business with Brain AI Annual-billing rates, straight multiplication. Both lines are per user per month. | Line | Rate | Seats | Per month | Per year | | --- | --- | --- | --- | --- | | ClickUp Business | $12.00 | 20 | $240.00 | $2,880.00 | | Brain AI add-on | $7.20 | 20 | $144.00 | $1,728.00 | | Total, annual billing | | 20 | $384.00 | $4,608.00 | | Same stack on monthly billing | $19 + $9 | 20 | $560.00 | $6,720.00 | | Polaris software | $0 | unlimited | $0.00 | $0.00 | ## What the ClickUp price list tells you about the model - **The monthly penalty is the largest here** — Business is 58 percent more on monthly billing, $19 against $12. No other vendor in this cluster charges that much for flexibility, and it is a deliberate push toward annual commitment. - **AI billed per seat is AI billed for nothing in particular** — A per-user AI charge is the same whether that user ran two prompts or two hundred. Access has a price; output does not. - **Credits underneath the add-on** — ClickUp also sells AI Super Credits alongside the per-seat AI tiers, so a full AI configuration can carry a fixed line and a variable one at the same time. ## Two ways to pay for AI at work **Per seat, per month** - $7.20 or $22.40 per user per month on annual billing - Charged for everyone with access, used or not - Cost rises with headcount, not with work - Nothing on the invoice names an output **Per hour delivered** - $0 until an AI worker delivers something - Roughly $2 per human-equivalent hour of delivered work - Every hour tied to a named job with its inputs logged - A quiet month costs nothing at all > **The honest limit on our side** > > Two dollars an hour sounds cheap and can add up if you assign a great deal of work. The difference is that the bill moves with output rather than headcount, and every line names the job that produced it. Polaris is in free public beta and has no usage data to publish, so any monthly figure here is arithmetic on a stated assumption. ## Where to go next - [alternatives/clickup](https://www.polarishq.co/alternatives/clickup) - [compare/clickup-vs-monday](https://www.polarishq.co/compare/clickup-vs-monday) - [compare/clickup-vs-notion](https://www.polarishq.co/compare/clickup-vs-notion) - [replace/clickup-and-slack](https://www.polarishq.co/replace/clickup-and-slack) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) ## Questions people ask **How much is ClickUp per user per month?** Unlimited is $7 per user per month on annual billing and $10 on monthly. Business is $12 annual and $19 monthly. Both were read from clickup.com/pricing on 21 August 2026, and Enterprise is custom. **Is ClickUp Brain included in the paid plans?** No. Brain AI is a separate per-user add-on at $7.20 per user per month on annual billing or $9 on monthly, and Everything AI is $22.40 annual or $28 monthly. The Free plan includes trial access to advanced AI features only. **Why does ClickUp charge so much more for monthly billing?** ClickUp's page shows up to thirty percent off standard plans for yearly billing, which works out to 43 percent more on Unlimited and 58 percent more on Business when paying monthly. It is the steepest monthly premium among the vendors checked in this cluster. **What does Polaris charge for its AI?** Nothing for access and roughly two dollars per human-equivalent hour of work an AI worker actually delivers. There is no per-seat AI charge, no credit pack and no tier. A person who never assigns anything to a worker generates no bill. **What if the delivered work is not good enough?** You do not close the task. An AI worker posts its result as a comment and ticks its acceptance criteria, but only a person closes and rates a task in Polaris. Work you send back is work you can argue about, and the work log shows exactly what the estimate was built from. ## Related - https://www.polarishq.co/alternatives/clickup - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/compare/clickup-vs-notion - https://www.polarishq.co/cost/monday-pricing - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/replace/clickup-and-slack --- --- title: "monday.com pricing 2026: $9 Basic, $12 Standard, $19 Pro" description: "monday work management rates per seat, the AI credit allowances in each tier, the seat-selector pricing model, and the arithmetic for a twenty-person team." url: https://www.polarishq.co/cost/monday-pricing section: Cost updated: 2026-08-21 --- # monday.com pricing, checked August 2026 Seats sold in blocks, AI sold in credits, and four separate products under one brand. ## The short answer monday work management lists Basic at $9, Standard at $12 and Pro at $19 per seat per month, with a yearly toggle marked "Yearly SAVE 18%", checked on monday.com/pricing on 21 August 2026. Free covers up to 2 seats. Each paid tier bundles AI credits: 1,000, 2,000 and 3,000 a month. Polaris charges nothing for software. - **Basic:** $9 / seat / month - **Standard:** $12 / seat / month - **Pro:** $19 / seat / month - **Checked:** 21 August 2026 ## monday work management, as listed Read from monday.com/pricing on 21 August 2026 with the billing toggle showing "Yearly SAVE 18%" against "Monthly". The rates below are the ones the page displayed; confirm your own toggle before budgeting. | Plan | Per seat / month | AI credits per month | Notes | | --- | --- | --- | --- | | Free | $0 | Basic allowance included | Up to 2 seats | | Basic | $9 | 1,000 | Unlimited items, 5GB storage | | Standard | $12 | 2,000 | Timeline, calendar and automation limits raised | | Pro | $19 | 3,000 | Private boards, time tracking, higher automation limits | | Enterprise | Custom | Custom | Not published | ## You are not buying one product monday sells work management, CRM, Service and Dev as separate priced products. CRM Basic listed at $12 per seat on annual billing and $18 on monthly, and Service Standard listed at $31 per seat on annual. A company that adopts monday for projects and then adds CRM is starting a second subscription with its own seat count, not extending the first. Seats are also bought through a selector rather than typed in freely, and the page defaults to 10 seats. Plan on buying to the next block rather than to your exact headcount. ## Twenty people on monday work management Straight multiplication at the listed per-seat rates. | Plan | Rate | Seats | Per month | Per year | | --- | --- | --- | --- | --- | | Basic | $9 | 20 | $180.00 | $2,160.00 | | Standard | $12 | 20 | $240.00 | $2,880.00 | | Pro | $19 | 20 | $380.00 | $4,560.00 | | Pro plus monday CRM Basic for 5 sales seats | $19 and $12 | 20 and 5 | $440.00 | $5,280.00 | | Polaris software | $0 | unlimited | $0.00 | $0.00 | ## Three cost mechanics specific to monday - **Credits are per account, seats are per person** — A Pro workspace gets 3,000 AI credits a month whether it has 10 seats or 200. The AI allowance does not scale with what you pay, which makes it a fixed budget that a growing team burns through faster each quarter. - **The seat selector rounds you up** — Seats are chosen from preset blocks. A team of 12 buys the block above 10, and the unused seats are billed the same as the used ones. - **Product sprawl inside one vendor** — Work management, CRM, Service and Dev each carry their own per-seat price. Consolidating onto monday can mean signing three subscriptions with the same logo on them. ## monday Pro against Polaris **monday Pro** - $19 per seat per month, checked 21 August 2026 - 3,000 AI credits per account per month, regardless of size - Separate subscriptions for CRM, Service and Dev - Seats bought in blocks, used or not **Polaris** - $0 software, unlimited humans, no blocks and no tiers - Tasks, docs and team chat in the same product - AI workers hired in about sixty seconds with an editable SKILL.md - Roughly $2 per human-equivalent hour delivered, logged per job > **One number worth writing down** > > A twenty-person monday Pro workspace is $4,560 a year before anyone opens the CRM. That is the budget an AI worker programme has to beat, and it is the comparison most teams never actually make because the two lines sit in different parts of the finance sheet. ## Where to go next - [alternatives/monday](https://www.polarishq.co/alternatives/monday) - [compare/asana-vs-monday](https://www.polarishq.co/compare/asana-vs-monday) - [compare/clickup-vs-monday](https://www.polarishq.co/compare/clickup-vs-monday) - [compare/trello-vs-monday](https://www.polarishq.co/compare/trello-vs-monday) - [replace/monday-and-slack](https://www.polarishq.co/replace/monday-and-slack) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) ## Questions people ask **How much is monday.com per seat?** Work management listed Basic at $9, Standard at $12 and Pro at $19 per seat per month on monday.com/pricing on 21 August 2026, with a billing toggle marked "Yearly SAVE 18%". CRM, Service and Dev are separately priced products with their own seat rates. **Is monday.com free for a small team?** The Free plan covers up to 2 seats, which is a trial rather than a working plan for a team. Beyond that you are on Basic at $9 per seat per month or higher, bought through a seat selector that works in blocks. **Are AI credits included with monday plans?** Yes, and the allowance is per account rather than per seat: 1,000 credits a month on Basic, 2,000 on Standard and 3,000 on Pro. A larger team pays more in seats and gets the same credit pool. **How would Polaris bill the same team?** Nothing for the software, at any seat count. Work delivered by AI workers is metered at roughly two dollars per human-equivalent hour. As an illustration only, if that twenty-person team's workers delivered 60 human-hours in a month, the bill for that month would be about $120 against $380 for monday Pro seats. **Who decides an AI worker's task is finished?** A person. The worker ticks its acceptance-criteria checklist as it goes and delivers the result as a comment with any files it produced, then stops. Closing and rating the task is a human action in Polaris, always. ## Related - https://www.polarishq.co/alternatives/monday - https://www.polarishq.co/compare/asana-vs-monday - https://www.polarishq.co/compare/clickup-vs-monday - https://www.polarishq.co/cost/clickup-pricing - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/glossary/per-seat-pricing --- --- title: "Confluence pricing 2026: free to 10 users, calculator above" description: "Confluence Cloud is free to 10 users and Atlassian publishes no readable per-user price above that. How the tiered billing works and what it does to a budget." url: https://www.polarishq.co/cost/confluence-pricing section: Cost updated: 2026-08-21 --- # Confluence pricing, checked August 2026 Priced by a function rather than a number, and bundled with the rest of Atlassian by design. ## The short answer Confluence Cloud's Free plan covers up to 10 users with 2GB of storage. Atlassian publishes no readable per-user list price for Standard or Premium: on 21 August 2026 the Confluence pricing page, the Standard page and the licensing page all routed to a calculator. Each user's rate depends on the tier they fall into, and the bill uses the highest seat count in the cycle. - **Free plan:** Up to 10 users, 2GB - **Standard / Premium:** Calculator only - **Community / academic:** 75% off cloud - **Checked:** 21 August 2026 ## What we could and could not read On 21 August 2026 we read atlassian.com/software/confluence/pricing, atlassian.com/software/confluence/standard and atlassian.com/licensing/confluence. None of them prints a per-user figure. The licensing page states that pricing is progressive with volume discounts as seat counts rise, that "each user's unit price is based on the pricing rate in which the user falls", and directs you to the calculator. This page therefore quotes no per-user price for Confluence. Comparison pages that do are either quoting one seat band as if it were the list rate, or repeating a third-party figure. Neither will match your invoice. ## What Atlassian does publish about Confluence billing All rows taken from Atlassian's own licensing and product pages on 21 August 2026. | Rule | What Atlassian says | Budget consequence | | --- | --- | --- | | Free tier | Up to 10 users, 2GB storage, community support | The eleventh person triggers a purchase decision | | Progressive pricing | Each user's unit price is based on the rate band that user falls into | Your average per-user cost is not anybody else's | | Maximum quantity billing | The bill uses the highest number of seats assigned at any point in the cycle | A seat that existed for a day is billed for the cycle | | Prorating | Mid-cycle additions are prorated; removals do not reduce the current period | Seat cleanups only pay off next cycle | | Annual discount | Available, and varies with the number of users purchased | No flat percentage to plan against | | Community and academic | 75% off cloud pricing for eligible customers | Check eligibility before any migration maths | ## The Confluence cost nobody puts in the business case - **It arrives with Jira** — Confluence is rarely bought alone. The usual shape is Jira plus Confluence plus Slack across the same headcount, three seat counts tracking one payroll. - **Rovo credits sit under the AI features** — Atlassian's paid tiers bundle Rovo AI, and that included AI is metered by a monthly credit allowance rather than being unlimited. Treat it as a second meter, not a feature. - **Space permissions are a real reason to stay** — Confluence's space-level permission model is more granular than most document tools, and for regulated documentation that granularity is the product. Price alone is a poor argument against it. ## Confluence against Polaris Docs **Confluence Cloud** - Free to 10 users, then priced by calculator - Space-level permissions and enterprise documentation controls - Bundled with Jira in most organisations, doubling the seat count - Rovo AI metered by monthly credit allowance **Polaris** - $0 software with unlimited docs and unlimited people - Nested doc tree and block editor with markdown shortcuts and sub-pages - Versioned files, file review and comments included - AI workers that write into the docs rather than sitting beside them > **Do not migrate for the price alone** > > If your Confluence spaces carry audited permissions or regulatory documentation, the migration cost and the control loss will exceed anything you save. The stronger reason to look elsewhere is that Confluence stores what people wrote and never writes anything itself. ## Where to go next - [alternatives/confluence](https://www.polarishq.co/alternatives/confluence) - [cost/jira-pricing](https://www.polarishq.co/cost/jira-pricing) - [compare/notion-vs-confluence](https://www.polarishq.co/compare/notion-vs-confluence) - [replace/confluence-and-jira](https://www.polarishq.co/replace/confluence-and-jira) - [replace/confluence-and-notion](https://www.polarishq.co/replace/confluence-and-notion) - [cost/stack-cost-50-person-team](https://www.polarishq.co/cost/stack-cost-50-person-team) ## Questions people ask **How much does Confluence cost per user?** Atlassian does not publish a readable per-user list price. Every Atlassian page we checked on 21 August 2026 routed to an interactive calculator, and the rate itself depends on which seat band each user falls into. Run the calculator for your own seat count. **Is Confluence free?** Confluence Cloud has a Free plan for up to 10 users with 2GB of storage and community support. Above 10 users you move to a paid tier priced through Atlassian's calculator. **Does deactivating a Confluence user reduce the bill?** Not for the current cycle. Atlassian bills on the maximum number of seats assigned to a product at any point during the billing cycle, so removals take effect from the next period. Additions, by contrast, are prorated immediately. **Are Polaris docs really unlimited and free?** Yes. Docs, tasks, workstreams, team chat, connections, versioning and file review are all included at no cost with no seat count. Revenue comes only from AI worker output, metered at roughly two dollars per human-equivalent hour. **How would I audit a Polaris bill?** Open the worker's work log. Each job records what drove the estimate: searches run, characters of finished prose, checklist items ticked, comments addressed and files produced. The formula that turns those into hours is published, so any line can be recomputed and disputed. ## Related - https://www.polarishq.co/alternatives/confluence - https://www.polarishq.co/cost/jira-pricing - https://www.polarishq.co/compare/notion-vs-confluence - https://www.polarishq.co/replace/confluence-and-jira - https://www.polarishq.co/replace/jira-and-confluence-and-slack - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/cost/stack-cost-50-person-team --- --- title: "Trello pricing 2026: $5 Standard, $10 Premium, annual rates" description: "Trello's Free, Standard, Premium and Enterprise rates on both billing bases, the ten-board free limit, and the five-person stack arithmetic in full." url: https://www.polarishq.co/cost/trello-pricing section: Cost updated: 2026-08-21 --- # Trello pricing, checked August 2026 The cheapest tracker on this list, and the one teams outgrow fastest. ## The short answer Trello lists Standard at $5 per user per month on annual billing and $6 on monthly, and Premium at $10 annual and $12.50 monthly, checked on trello.com/pricing on 21 August 2026. Enterprise is $17.50 per user per month on annual. Free allows 10 boards and up to 10 collaborators per Workspace. Polaris charges nothing for software. - **Standard:** $5 annual / $6 monthly - **Premium:** $10 annual / $12.50 monthly - **Free limit:** 10 boards per Workspace - **Checked:** 21 August 2026 ## Trello plans on both billing bases Read from trello.com/pricing on 21 August 2026. | Plan | Annual billing | Monthly billing | What it adds | | --- | --- | --- | --- | | Free | $0 | $0 | Unlimited cards, 10 boards per Workspace, up to 10 collaborators, inbox, unlimited Power-Ups per board | | Standard | $5 / user / month | $6 / user / month | Unlimited boards, AI features, Planner, card mirroring, custom fields | | Premium | $10 / user / month | $12.50 / user / month | Calendar, Timeline, Table, Dashboard and Map views, workspace templates, admin controls | | Enterprise | $17.50 / user / month | Not published | Unlimited Workspaces, org-wide permissions, Atlassian Guard Standard, 24/7 admin support | ## The ten-board wall, not the price, is what moves teams Trello Free gives unlimited cards and caps you at 10 boards per Workspace. Boards multiply faster than anyone expects, because a board is how Trello models a project, a client, a quarter and a checklist all at once. The eleventh board is the upgrade trigger for most teams, and it arrives long before the tenth teammate. The second wall is views. Calendar, Timeline, Table and Dashboard sit on Premium at $10 per user per month, double the Standard rate, and they are the features a team asks for the moment more than one project runs in parallel. ## A five-person team on Trello and the rest of the stack Monthly-billing rates, since small teams rarely commit annually. The annual column shows what committing would save. | Line | Monthly rate | Seats | Per month | Annual-billing equivalent / month | | --- | --- | --- | --- | --- | | Trello Premium | $12.50 | 5 | $62.50 | $50.00 | | Slack Pro | $8.75 | 5 | $43.75 | $36.25 | | Notion Plus | $10.00 | 5 | $50.00 | Up to 20% less with yearly | | Total | | 5 | $156.25 | $136.25 | | Total per year | | 5 | $1,875.00 | $1,635.00 | | Polaris software | $0 | unlimited | $0.00 | $0.00 | ## Three things about Trello worth being honest about - **It is the easiest tool here to get people to actually use** — A board with three lists needs no training. That adoption advantage is real, and it is why Trello survives in companies that also pay for something heavier. - **It is an Atlassian product** — Enterprise bundles Atlassian Guard Standard, and Trello sits in the same account structure as Jira and Confluence. Consolidation conversations tend to pull all three together. - **It tracks, and that is the ceiling** — A card moving from Doing to Done is a record of a person's afternoon. Trello never produces the afternoon. ## Trello Premium against Polaris **Trello Premium** - $10 per user per month annual, $12.50 monthly - Board, calendar, timeline, table and dashboard views - Boards as the unit of everything, which is both the charm and the ceiling - Docs and team chat live in other subscriptions **Polaris** - $0 software, unlimited people, boards and workstreams - Lanes shared between list and board view, so you organise once - Focus lane pinned first: today, this week, next 30 days - Roughly $2 per human-equivalent hour an AI worker delivers > **Cheap is not the same as free** > > Trello Premium at $10 a seat is genuinely inexpensive. It is also the third line on an invoice that already carries a docs tool and a chat tool, and it is the one line that most teams could stop paying entirely without losing a place to put work. ## Where to go next - [alternatives/trello](https://www.polarishq.co/alternatives/trello) - [compare/trello-vs-asana](https://www.polarishq.co/compare/trello-vs-asana) - [compare/trello-vs-monday](https://www.polarishq.co/compare/trello-vs-monday) - [replace/trello-and-slack](https://www.polarishq.co/replace/trello-and-slack) - [replace/notion-and-trello](https://www.polarishq.co/replace/notion-and-trello) - [cost/stack-cost-5-person-team](https://www.polarishq.co/cost/stack-cost-5-person-team) ## Questions people ask **How much is Trello per user per month?** Standard is $5 per user per month on annual billing and $6 on monthly. Premium is $10 annual and $12.50 monthly. Enterprise is listed at $17.50 per user per month on annual billing, which works out to $210 per user per year. All were read from trello.com/pricing on 21 August 2026. **What are the limits on Trello's free plan?** Unlimited cards, but 10 boards per Workspace and up to 10 collaborators per Workspace. Boards are the constraint that bites first, because Trello uses a board for almost every kind of container. **Is Polaris free the way Trello is free?** No, it is broader. Trello Free is a capped tier that upsells you at 10 boards. Polaris software is free with no caps on people, tasks, workstreams or docs, and the money comes from a different place entirely: roughly two dollars per human-equivalent hour that an AI worker delivers. **How is a human-equivalent hour estimated?** By a published formula applied to what the job did: fifteen minutes of base pickup time, twelve minutes per web search, finished prose at about ninety characters a minute, eight minutes per acceptance-criteria item, ten minutes per comment and per file, five per PDF, clamped between five minutes and eight hours. It is an estimate from observable effort, and every input is on the work log. **What if an AI worker gets the task wrong?** Send it back. The worker delivers as a comment and ticks its own acceptance criteria, but a person closes and rates the task. Nothing is marked done by the machine, in any circumstance. ## Related - https://www.polarishq.co/alternatives/trello - https://www.polarishq.co/compare/trello-vs-asana - https://www.polarishq.co/compare/trello-vs-monday - https://www.polarishq.co/replace/trello-and-slack - https://www.polarishq.co/cost/stack-cost-5-person-team - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/glossary/focus-lane --- --- title: "What a 5-person team's tool stack costs: $1,875 a year" description: "Notion Plus, Slack Pro and Trello Premium for five people comes to $1,875 a year on monthly billing. Every rate, seat count and multiplication is shown." url: https://www.polarishq.co/cost/stack-cost-5-person-team section: Cost updated: 2026-08-21 --- # Tool stack cost for a 5-person team Three subscriptions, five people, and the arithmetic laid out so you can check it against your own invoices. ## The short answer A five-person team on Notion Plus, Slack Pro and Trello Premium pays $156.25 a month, or $1,875 a year, at the list rates checked on 21 August 2026 with monthly billing. Committing annually where the vendor publishes an annual rate brings that to about $1,635. Polaris replaces all three at $0 and bills only for work an AI worker delivers. - **Monthly billing:** $1,875 / year - **Annual billing:** About $1,635 / year - **Per person:** $375 / year - **Prices checked:** 21 August 2026 ## The stack, line by line Rates read from each vendor's pricing page on 21 August 2026. Notion displays a single per-member rate with a yearly toggle marked "Save up to 20% with yearly" and does not print the yearly figure, so its annual column is expressed as the discount rather than a number we would be inventing. | Tool and plan | Rate | Basis | Seats | Per month | Per year | | --- | --- | --- | --- | --- | --- | | Notion Plus | $10.00 / member / month | As displayed | 5 | $50.00 | $600.00 | | Slack Pro | $8.75 / active user / month | Monthly | 5 | $43.75 | $525.00 | | Trello Premium | $12.50 / user / month | Monthly | 5 | $62.50 | $750.00 | | Total, monthly billing | | | 5 | $156.25 | $1,875.00 | | Slack Pro, annual | $7.25 / active user / month | Annual | 5 | $36.25 | $435.00 | | Trello Premium, annual | $10.00 / user / month | Annual | 5 | $50.00 | $600.00 | | Total with annual commitments | | | 5 | $136.25 | $1,635.00 | | Polaris | $0 | No seats | unlimited | $0.00 | $0.00 | ## What $375 a person a year actually buys Three places to put work. Notion holds what you wrote down, Trello holds what you agreed to do, Slack holds the conversation where you decided it. None of the three does any of the work, and at five people that is fine, because five people can hold the whole context in their heads and the tools are just filing. The awkward part is that this is the cheapest the stack will ever be. At ten people the same three subscriptions cost more than twice this, because Notion Plus stops being enough and Slack's history limit stops being survivable. ## The same stack as you grow Comparable stacks from the other pages in this cluster, all at list prices checked 21 August 2026. - **$1,875** — 5 people. Notion Plus, Slack Pro, Trello Premium, monthly billing - **$4,650** — 10 people. Notion Business, Slack Pro, Linear Basic - **$15,447** — 25 people. Notion Business, Slack Business+, Asana Starter, monthly billing - **$37,200** — 50 people. Plus Claude Team and GitHub Copilot seats ## The illustrative Polaris side Polaris is in free public beta with no customers and no usage data. The figures below are arithmetic on a stated assumption, not measurements. - **The software line is $0, and that part is not an estimate** — Unlimited humans, tasks, workstreams, docs, connections and the Chief of Staff cost nothing. No seats, no tiers, no upgrade wall. - **If your workers delivered 20 human-hours a month** — That would be about $40 a month, or $480 a year, at roughly two dollars per human-equivalent hour. Against $1,875 for the three subscriptions, illustrative only. - **If your workers delivered 40 human-hours a month** — About $80 a month, or $960 a year. Still roughly half the stack it replaced, and this time the invoice names what got produced. - **If nobody assigns anything, the bill is zero** — That is not a promotion. It is how the meter works: nothing delivered, nothing billed. ## What changes at five people **Three subscriptions** - $1,875 a year before anyone buys an AI subscription - Context split across three products and three search boxes - Every new hire adds three seat charges at once - Nothing on any of the three invoices names an output **One free workspace** - Tasks, docs and team chat in a single product at $0 - Focus lane pinned first: today, this week, next 30 days - Slack still connects, so signals arrive as prefilled task suggestions - AI workers on the same member list as the humans > **The number a five-person team should actually check** > > Not the $1,875. Check how many of those five seats are on all three products and how many are on one. Small teams routinely pay full seat price across a whole stack for someone who opens exactly one of them. ## Where to go next - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [cost/slack-pricing](https://www.polarishq.co/cost/slack-pricing) - [cost/trello-pricing](https://www.polarishq.co/cost/trello-pricing) - [for/startups](https://www.polarishq.co/for/startups) - [for/solo-founders](https://www.polarishq.co/for/solo-founders) ## Questions people ask **How is the $1,875 calculated?** Notion Plus at $10 per member per month, Slack Pro at $8.75 per active user per month on monthly billing, and Trello Premium at $12.50 per user per month on monthly billing, each multiplied by five seats and then by twelve. All three rates were read from the vendors' own pricing pages on 21 August 2026. **Would a different five-person stack cost less?** Yes. Trello Standard instead of Premium saves $6.50 per person per month, and staying on Slack Free saves the whole Slack line until the 90-day history limit bites. This page models the configuration most five-person teams actually end up on rather than the cheapest one available. **Is the Polaris comparison a real figure?** The $0 software line is real and unconditional. The usage figures are illustrative arithmetic on a stated assumption, because Polaris is in free public beta with no customers and no published usage data. Anyone presenting a Polaris monthly bill as a measurement is making it up. **How do I know the AI worker hours are not inflated?** Each job logs its own inputs, and the formula that turns them into hours is published: base pickup time, searches, characters of finished prose, checklist items, comments and files, clamped between five minutes and eight hours per session. You can recompute any line from the work log and challenge it. **What happens when an AI worker delivers something unusable?** You do not close the task. Agents deliver and humans close in Polaris, so the machine posts its work as a comment and ticks its acceptance criteria, and a person decides whether that is finished. Rejected work stays open and disputable. ## Related - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/cost/trello-pricing - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/for/startups - https://www.polarishq.co/for/solo-founders - https://www.polarishq.co/glossary/tool-sprawl --- --- title: "What a 10-person team's tool stack costs: $4,650 a year" description: "Notion Business, Slack Pro and Linear Basic for ten people costs $4,650 a year. Every rate, seat count and multiplication shown, with billing bases labelled." url: https://www.polarishq.co/cost/stack-cost-10-person-team section: Cost updated: 2026-08-21 --- # Tool stack cost for a 10-person team The size where the free tiers run out on all three products at roughly the same time. ## The short answer A ten-person team on Notion Business, Slack Pro and Linear Basic pays $387.50 a month, or $4,650 a year, at the list rates checked on 21 August 2026 with Slack on monthly billing. Moving Slack to annual brings it to $4,470. Polaris replaces all three at $0 for the software and bills only for work an AI worker delivers. - **Total:** $4,650 / year - **Per person:** $465 / year - **With annual Slack:** $4,470 / year - **Prices checked:** 21 August 2026 ## The stack, line by line Rates read from each vendor's own pricing page on 21 August 2026. Linear publishes only its yearly-billing rate, so that is what the Linear row uses and says. | Tool and plan | Rate | Basis | Seats | Per month | Per year | | --- | --- | --- | --- | --- | --- | | Notion Business | $20.00 / member / month | As displayed | 10 | $200.00 | $2,400.00 | | Slack Pro | $8.75 / active user / month | Monthly | 10 | $87.50 | $1,050.00 | | Linear Basic | $10.00 / user / month | Billed yearly | 10 | $100.00 | $1,200.00 | | Total | | | 10 | $387.50 | $4,650.00 | | Same stack, Slack on annual | $7.25 Slack rate | Mixed | 10 | $372.50 | $4,470.00 | | If Linear moves to Business | $16.00 / user / month | Billed yearly | 10 | $447.50 | $5,370.00 | | Polaris | $0 | No seats | unlimited | $0.00 | $0.00 | ## Why ten is the expensive number Three ceilings land within a few months of each other. Linear's free plan stops at 2 teams and 250 issues. Slack Free keeps 90 days of history, which becomes a problem the first time a decision from last quarter cannot be found. Notion's free tier limits blocks once a workspace has more than one member, and by ten people the Business features are what people are actually asking for. So the jump is not gradual. A team goes from paying nothing to paying about $4,650 a year inside one quarter, and nobody planned for it because each upgrade looked small on its own. ## Two seat counts that are usually wrong - **The people who only need one of the three** — A designer who lives in Figma, a founder who lives in Slack, an ops hire who lives in Notion. All three are billed on all three products in most companies, because seat audits happen at renewal and renewals are annual. - **Contractors and part-timers** — Slack gives you back the money when someone goes fully inactive for a 28-day window. Notion and Linear do not. A three-month contractor is a twelve-month seat unless somebody remembers. - **The upgrade you did not choose** — Linear Basic allows 5 teams. Crossing that pushes you to Business at $16, which on ten seats adds $720 a year for a limit rather than a feature. ## The illustrative Polaris side Polaris is in free public beta with no customers and no usage data. Every row below is arithmetic on a stated assumption, marked illustrative, and is not a measurement of anything. | Assumption | Rate | Per month | Per year | Versus $4,650 | | --- | --- | --- | --- | --- | | Software, all ten people | $0 | $0.00 | $0.00 | Real, not an estimate | | Illustrative: workers deliver 30 human-hours a month | ~$2 / human-hour | ~$60.00 | ~$720.00 | About 15% | | Illustrative: workers deliver 60 human-hours a month | ~$2 / human-hour | ~$120.00 | ~$1,440.00 | About 31% | | Illustrative: workers deliver 120 human-hours a month | ~$2 / human-hour | ~$240.00 | ~$2,880.00 | About 62% | | Nothing assigned to a worker | ~$2 / human-hour | $0.00 | $0.00 | Nothing delivered, nothing billed | ## The shape of the two bills **Three per-seat subscriptions** - Predictable, and predictably rising with every hire - $465 per person per year before any AI subscription - Three invoices, none of which names a deliverable - Seat cleanups only pay off at the next renewal **Free software, metered work** - $0 regardless of headcount, so hiring does not move the bill - Variable, and tied to what came back finished - Each hour attached to a named job with its inputs logged - You close and rate every delivery before it counts > **The honest downside of a usage meter** > > A per-seat bill is boring and predictable, and finance teams like that. A usage bill moves. What makes it defensible is not that it is always smaller, it is that every line names a job and can be recomputed from the work log. Predictability you cannot audit is worth less than variability you can. ## Where to go next - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [cost/stack-cost-5-person-team](https://www.polarishq.co/cost/stack-cost-5-person-team) - [cost/linear-pricing](https://www.polarishq.co/cost/linear-pricing) - [cost/notion-pricing](https://www.polarishq.co/cost/notion-pricing) - [replace/linear-and-notion-and-github](https://www.polarishq.co/replace/linear-and-notion-and-github) - [for/software-teams](https://www.polarishq.co/for/software-teams) ## Questions people ask **How is the $4,650 calculated?** Notion Business at $20 per member per month, Slack Pro at $8.75 per active user per month on monthly billing, and Linear Basic at $10 per user per month billed yearly, each times ten seats and then times twelve. All rates were read from the vendors' pricing pages on 21 August 2026. **Why is Linear shown at the yearly rate when Slack is shown monthly?** Because that is what each vendor publishes. Slack prints both its monthly and annual rates; Linear's page showed only the yearly-billing rate when checked. Mixing bases silently is how most comparison pages produce numbers that match nobody's invoice, so this page labels every row instead. **Does Polaris actually replace all three?** It covers all three jobs: tasks and boards, a nested doc tree with a block editor, and team chat, in one free workspace. Whether it replaces them for your team depends on what you use Linear's cycles and Notion's databases for. Slack, Notion and Linear are all in the connections catalog, so connecting first and cutting seats later is the low-risk path. **Are the Polaris usage numbers on this page real?** No, and they are labelled illustrative for that reason. Polaris is in free public beta with no customers and no usage data to publish. The $0 software line is real; every hour figure on this page is arithmetic on an assumption stated in the row. **How do I challenge an hour on the bill?** Open the job on the worker's work log. It records the searches run, the characters of finished prose delivered, the acceptance-criteria items ticked, the comments addressed and the files produced. The formula is published, so recompute the line and dispute it with the inputs in front of you. ## Related - https://www.polarishq.co/cost/stack-cost-5-person-team - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/replace/notion-and-linear-and-slack - https://www.polarishq.co/for/software-teams - https://www.polarishq.co/glossary/human-equivalent-hours --- --- title: "What a 25-person team's tool stack costs: $15,447 a year" description: "Notion Business, Slack Business+ and Asana Starter for twenty-five people comes to $15,447 a year on monthly billing, or $13,797 with annual commitments." url: https://www.polarishq.co/cost/stack-cost-25-person-team section: Cost updated: 2026-08-21 --- # Tool stack cost for a 25-person team The size where paying monthly instead of annually costs $1,650 a year on its own. ## The short answer A twenty-five person team on Notion Business, Slack Business+ and Asana Starter pays $1,287.25 a month, or $15,447 a year, at the list rates checked on 21 August 2026 with monthly billing. Committing annually where the rates are published brings it to $13,797, a difference of $1,650. Polaris replaces the three at $0 for the software. - **Monthly billing:** $15,447 / year - **Annual billing:** $13,797 / year - **Cost of paying monthly:** $1,650 / year - **Prices checked:** 21 August 2026 ## The stack on both billing bases Rates read from each vendor's own pricing page on 21 August 2026. Notion displays one per-member rate with a yearly toggle marked "Save up to 20% with yearly" and prints no yearly figure, so its rate is unchanged across both columns here rather than guessed. | Tool and plan | Monthly-billing rate | Annual-billing rate | Seats | Monthly total | Annual total | | --- | --- | --- | --- | --- | --- | | Notion Business | $20.00 | As displayed, up to 20% less | 25 | $500.00 | $500.00 | | Slack Business+ | $18.00 | $15.00 | 25 | $450.00 | $375.00 | | Asana Starter | $13.49 | $10.99 | 25 | $337.25 | $274.75 | | Total per month | | | 25 | $1,287.25 | $1,149.75 | | Total per year | | | 25 | $15,447.00 | $13,797.00 | | Per person per year | | | 25 | $617.88 | $551.88 | | Polaris software | $0 | $0 | unlimited | $0.00 | $0.00 | ## The $1,650 nobody negotiates Slack Business+ is $18 per active user per month on monthly billing and $15 on annual. Asana Starter is $13.49 against $10.99. Those two gaps alone are $1,650 a year on twenty-five seats, which is more than most teams spend on their entire design tooling, and it is spent purely on the option to cancel. Sometimes that option is worth buying, particularly during a reorganisation. What is not defensible is paying it by accident, which is the usual reason it appears on an invoice. ## Where the money leaks at twenty-five people - **Tier drift** — Slack Business+ at $18 rather than Pro at $8.75 is usually bought for SSO and compliance controls, then paid on every seat including the ones that never touch a compliance feature. It is the single largest line in this stack. - **Seats that outlived their people** — Slack credits you when someone goes inactive for a full 28-day window. Notion and Asana do not. At twenty-five people there are almost always two or three seats belonging to someone who left or moved. - **The fourth subscription** — This stack has three tools in it. Most twenty-five person companies also pay for a design tool, a CRM, and a growing pile of AI subscriptions. The cost-of-AI-subscriptions page in this cluster models that layer separately. ## The illustrative Polaris side Polaris is in free public beta with no customers and no usage data. Every row is arithmetic on the assumption stated in it, not a measurement. | Assumption | Per month | Per year | Versus $13,797 annual-billing stack | | --- | --- | --- | --- | | Software for all twenty-five people | $0.00 | $0.00 | Real, unconditional, not an estimate | | Illustrative: workers deliver 80 human-hours a month | ~$160.00 | ~$1,920.00 | About 14% | | Illustrative: workers deliver 160 human-hours a month | ~$320.00 | ~$3,840.00 | About 28% | | Illustrative: workers deliver 320 human-hours a month | ~$640.00 | ~$7,680.00 | About 56% | | Illustrative: one full-time-equivalent month of delivered work | ~$320.00 | ~$3,840.00 | 160 hours at roughly $2 an hour | ## Two ways to spend $13,797 **Three subscriptions for twenty-five people** - $551.88 per person per year, whether they log in or not - Three products, three search boxes, three permission models - A twenty-sixth hire adds three more seat charges - No invoice line corresponds to anything that got done **One free workspace plus metered work** - $0 for the software at twenty-five people or two hundred - Tasks, docs and team chat in one product with one permission model - AI workers on the same member list, assigned the same way - Each billed hour attached to a named job you can audit > **The comparison that actually matters** > > Not software against software. At twenty-five people the useful question is what $13,797 buys against what a contractor costs. At roughly two dollars per human-equivalent hour, that annual subscription budget is on the order of several thousand hours of delivered work. Whether those hours are worth as much as a contractor's is a real question, and it is the one we would rather be judged on. ## Where to go next - [cost/stack-cost-50-person-team](https://www.polarishq.co/cost/stack-cost-50-person-team) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) - [cost/asana-pricing](https://www.polarishq.co/cost/asana-pricing) - [cost/slack-pricing](https://www.polarishq.co/cost/slack-pricing) - [for/remote-teams](https://www.polarishq.co/for/remote-teams) ## Questions people ask **How is the $15,447 calculated?** Notion Business at $20 per member per month, Slack Business+ at $18 per active user per month on monthly billing, and Asana Starter at $13.49 per user per month on monthly billing, each times twenty-five seats and then times twelve. Rates were read from the vendors' pricing pages on 21 August 2026. **Why is the annual figure only $1,650 lower?** Because only two of the three vendors publish a separate annual rate. Slack drops from $18 to $15 and Asana from $13.49 to $10.99, which is $1,650 a year across twenty-five seats. Notion shows one rate with a yearly toggle marked up to twenty percent off but prints no yearly figure, so this page leaves the Notion line unchanged rather than inventing one. **Is the Polaris usage column real data?** No. Polaris is in free public beta with no customers and no published usage data, so every hour figure here is arithmetic on an assumption written into the row. The $0 software line is real and applies at any headcount. **How do I know the delivered hours are not inflated?** Because the estimate is mechanical and published. Fifteen minutes of base pickup time, twelve minutes per web search, finished prose at about ninety characters a minute, eight minutes per acceptance-criteria item, ten minutes per comment and per file, five per PDF, clamped between five minutes and eight hours per session. Each job's inputs are on the work log, so any line can be recomputed and challenged. **What if an AI worker delivers bad work at this scale?** The same thing as at any scale: you do not close the task. Agents deliver and humans close in Polaris. The worker posts its result as a comment and ticks its acceptance criteria; a person decides whether that counts as done and rates it. ## Related - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/cost/stack-cost-50-person-team - https://www.polarishq.co/cost/asana-pricing - https://www.polarishq.co/cost/slack-pricing - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/for/remote-teams - https://www.polarishq.co/glossary/tool-sprawl --- --- title: "What a 50-person team's tool stack costs: $37,200 a year" description: "Notion Business, Slack Business+, Linear Business plus Claude Team and Copilot seats for fifty people comes to $37,200 a year. Every multiplication shown." url: https://www.polarishq.co/cost/stack-cost-50-person-team section: Cost updated: 2026-08-21 --- # Tool stack cost for a 50-person team Three work tools and two AI subscriptions, which is where most companies of this size actually are. ## The short answer A fifty-person team on Notion Business, Slack Business+ and Linear Business pays $2,550 a month for work tools. Adding twenty Claude Team seats and fifteen GitHub Copilot Pro seats brings it to $3,100 a month, or $37,200 a year, at list rates checked 21 August 2026. Polaris covers the first three at $0 and meters AI work by the hour delivered. - **Work tools:** $30,600 / year - **AI subscriptions:** $6,600 / year - **Total:** $37,200 / year - **Per person:** $744 / year ## The full stack, line by line Rates read from each vendor's own pricing page on 21 August 2026. Slack and Linear rows use the annual-billing rate each vendor publishes; Notion shows one rate with a yearly toggle and no printed yearly figure. | Line | Rate | Basis | Seats | Per month | Per year | | --- | --- | --- | --- | --- | --- | | Notion Business | $20.00 / member / month | As displayed | 50 | $1,000.00 | $12,000.00 | | Slack Business+ | $15.00 / active user / month | Annual | 50 | $750.00 | $9,000.00 | | Linear Business | $16.00 / user / month | Billed yearly | 50 | $800.00 | $9,600.00 | | Work tools subtotal | | | 50 | $2,550.00 | $30,600.00 | | Claude Team, standard seat | $20.00 / seat / month | Annual | 20 | $400.00 | $4,800.00 | | GitHub Copilot Pro | $10.00 / user / month | Monthly | 15 | $150.00 | $1,800.00 | | AI subtotal | | | 35 of 50 | $550.00 | $6,600.00 | | Total | | | 50 | $3,100.00 | $37,200.00 | | Polaris software | $0 | No seats | unlimited | $0.00 | $0.00 | ## At fifty people the AI layer is a fifth of the bill Twenty Claude Team seats and fifteen Copilot Pro seats is a conservative configuration for a fifty-person company in 2026, and it comes to $6,600 a year. That is more than the entire tool stack of the five-person team in this cluster, and it buys access rather than output: a seat costs the same whether the person behind it ran two prompts this month or two hundred. It also stays on individual machines. Claude Code, Copilot and Cursor are extraordinary single-player tools, and none of them leaves an artefact your teammates can assign work to, review, or audit six months later. ## Four things that get expensive specifically at fifty - **Business tiers become mandatory, not optional** — SSO, SCIM, audit logs and granular permissions live on the higher tier of every product here. At fifty people that is a security requirement, so the $20 and $18 and $16 rates are not really choices. - **Partial AI rollout is a permanent argument** — Thirty-five of fifty people have an AI subscription in this model. Deciding who gets one is a recurring management task that exists only because the pricing is per seat. - **Nobody can audit what the AI produced** — Five figures a year on AI subscriptions, and no shared record of what came out of them. No work log, no deliverable attached to a task, no way to answer what the money bought. - **Every hire adds five lines** — Three work tools plus, increasingly, two AI subscriptions. The fifty-first person is roughly $744 a year before they open anything. ## The illustrative Polaris side Polaris is in free public beta with no customers and no usage data. Every hour figure below is arithmetic on the assumption stated in the row. | Assumption | Per month | Per year | Versus $37,200 | | --- | --- | --- | --- | | Software for all fifty people | $0.00 | $0.00 | Real, unconditional | | Illustrative: workers deliver 150 human-hours a month | ~$300.00 | ~$3,600.00 | About 10% | | Illustrative: workers deliver 300 human-hours a month | ~$600.00 | ~$7,200.00 | About 19% | | Illustrative: workers deliver 600 human-hours a month | ~$1,200.00 | ~$14,400.00 | About 39% | | Illustrative: 1,600 human-hours a month, ten FTE-months | ~$3,200.00 | ~$38,400.00 | Roughly the same money, for delivered work | ## What $37,200 buys, two ways **Five subscriptions** - Three places to put work and two AI tools that stay on laptops - $744 per person per year, rising with every hire - AI priced by access, so usage and cost are unrelated - No shared record of what any of the AI spend produced **One workspace, metered output** - Tasks, docs and team chat at $0 for all fifty - AI workers on the same member list as the humans, assigned identically - A real cloud machine wakes per task and keeps working after laptops close - Every billed hour attached to a job, with its inputs on the work log > **Say the uncomfortable version** > > At high volume a usage meter can cost as much as the subscriptions it replaced. The last row of the table above says so out loud. The difference is what the invoice describes: five figures of seat licences describes access, and five figures of metered hours describes work that arrived, was reviewed, and was closed by a person. ## Where to go next - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [cost/linear-pricing](https://www.polarishq.co/cost/linear-pricing) - [cloud-claude-code/claude-code-for-teams](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) - [for/software-teams](https://www.polarishq.co/for/software-teams) ## Questions people ask **How is the $37,200 calculated?** Notion Business at $20 per member per month on fifty seats, Slack Business+ at $15 annual on fifty, Linear Business at $16 billed yearly on fifty, Claude Team standard seats at $20 annual on twenty, and GitHub Copilot Pro at $10 on fifteen. That is $3,100 a month, times twelve. All rates were read from vendor pricing pages on 21 August 2026. **Why only thirty-five AI seats for fifty people?** Because that is what partial rollout looks like in practice, and it is the honest version. Giving all fifty people both subscriptions would put the AI line at $18,000 a year instead of $6,600. The model here is deliberately conservative. **Would Polaris really be cheaper at this size?** The software line would be zero, which is $30,600 of the $37,200. Whether the metered work costs more or less than $6,600 a year depends entirely on how much work you assign, and the last row of the illustrative table shows a volume where it costs about the same. We have no usage data and will not pretend otherwise. **Can fifty people share AI workers, or is it per person?** Shared. AI workers are members of the organisation, on the same roster as the humans, and anyone can assign a task to one. There are no AI seats to allocate, so the argument about who gets access does not exist. **How does a fifty-person company audit AI spend in Polaris?** Through the work log. Every job records the searches run, the finished prose delivered, the acceptance-criteria items ticked, the comments addressed and the files produced, and the formula that converts those into hours is published. Any line on the bill can be traced to a task somebody closed. ## Related - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/cost/confluence-pricing - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/for/software-teams --- --- title: "Per-seat vs usage pricing: what each model charges you for" description: "Per-seat pricing charges for places to put work. Usage pricing charges for work delivered. The trade-offs, the trust problem, and the honest case for each." url: https://www.polarishq.co/cost/per-seat-vs-usage-pricing section: Cost updated: 2026-08-21 --- # Per-seat pricing and usage pricing charge for different things One model prices access. The other prices output. Almost every argument about software cost is really about which of those you are buying. ## The short answer Per-seat pricing charges for the number of people who can open a product, so the bill tracks headcount and every new teammate costs more whether or not they use it. Usage pricing charges for units of work the product delivered, so the bill tracks output and an unused month costs nothing. Polaris prices software at $0 and meters delivered work at about $2 per human-equivalent hour. - **Per-seat unit:** A person with access - **Usage unit:** Work delivered - **Polaris software:** $0, no seats - **Polaris meter:** ~$2 per human-hour delivered ## A price is an answer to the question: what am I buying a unit of? Per-seat software sells you a licence for a person. The product's job is to be somewhere work can be recorded, and the vendor's revenue grows when more people are allowed in. That is a coherent model and it built most of the software industry. It also means the invoice describes your org chart, not your output. Usage pricing sells you a unit of something happening. Cloud infrastructure has worked this way for twenty years and nobody finds it strange: you pay for compute that ran, storage that was held, requests that were served. The awkward part has always been that the unit is a machine unit, so a finance team has to translate gigabyte-months into anything they care about. What changes with AI workers is that the unit can finally be a work unit. Not tokens, not credits, not seats. An hour of the work a person would otherwise have done, estimated from what the job actually did, and itemised. Per-seat logic breaks in two places. The first is tool sprawl: once a company runs four products across one headcount, it is paying four times for the same people, and each product's per-seat rate is individually reasonable while the total is not. A twenty-five person team in this cluster pays $551.88 per person per year across three tools, and none of those three tools ever did any work. The second is AI. Charging per seat for an AI feature prices access to a thing whose whole value is variable output. A colleague who runs two prompts and a colleague who runs two hundred cost the same, which turns a productivity question into an allocation argument about who is allowed a licence. Companies of fifty are having that argument right now, and it exists only because of the pricing model. ## The same company under both models A twenty-five person team, ten of whom do meaningful work in the product and fifteen of whom mostly read. | Question | Per-seat pricing | Usage pricing | | --- | --- | --- | | What is the billable unit? | A person with access | Work that was delivered | | What happens when you hire? | The bill rises immediately | The bill does not move | | What happens in a quiet month? | You pay in full | You pay less, or nothing | | What happens to the fifteen readers? | Billed the same as the ten doers | Cost nothing, because they produced nothing to bill | | Can the invoice be tied to an outcome? | No line corresponds to anything produced | Every line names a job | | How predictable is next month? | Exactly predictable | A forecast you have to make | | What is the vendor rewarded for? | Getting more people licensed | Delivering more work that gets accepted | ## The honest case for per-seat pricing This is not a strawman we are about to knock over. Per-seat won for good reasons and still wins in specific situations. - **Finance can plan it in one line** — Headcount times rate times twelve. A CFO can forecast next year's software spend in a spreadsheet cell, and there is real value in a number that never surprises anybody. - **Procurement is built for it** — Annual seat contracts fit purchase orders, approval thresholds and renewal calendars. Usage pricing regularly fails a procurement process not because it costs more but because nobody knows what number to put on the form. - **It aligns when the product really is a place** — If what you are buying is genuinely a shared record that people read and write, then people is the right unit. A wiki priced by the word would be absurd. - **It cannot run away from you** — The worst case of a per-seat bill is known in advance. The worst case of an unbounded meter is not, and that fear is legitimate rather than irrational. ## What each model is actually optimising **Per seat** - Vendor grows by increasing licensed headcount - Customer saves by restricting who gets access - The two incentives point in opposite directions - Nobody in the loop is measured on work produced **Per unit of delivered work** - Vendor grows only when more delivered work is accepted - Customer saves by assigning less, or by rejecting bad work - Both sides are measured on the same thing - Bad output is a revenue problem for the vendor, not just a support ticket ## The three fears of a usage meter, and what answers them Usage pricing fails on trust more often than on arithmetic. These are the objections in the order buyers raise them. 1. **It will run up a bill while I am not looking** — The unit has to be legible before anything else matters. In Polaris a single agent session is clamped to a maximum of eight human-equivalent hours, which at roughly two dollars an hour is about sixteen dollars for the largest job the system can bill. Work also arrives as a delivery you see, on a task you assigned, not as background consumption. 2. **I will pay for work that is wrong** — Agents deliver and humans close. An AI worker posts its result as a comment, ticks its own acceptance-criteria checklist, and stops. Nothing is ever marked done by the machine. You close and rate the task, and work you reject is work you can dispute. 3. **I cannot predict what this costs per month** — This one is real and we are not going to pretend it away. Polaris is in free public beta with no customers and no usage data, so there are no cohort ranges to quote. What exists instead is auditability: the formula is published, every job logs its inputs, and you can recompute any line yourself. > **The trade we are proposing** > > Give up a number that never moves, in exchange for a number you can take apart. A per-seat invoice is perfectly predictable and completely opaque about what it bought. A metered invoice moves, and every line on it points at a job somebody closed. We think that is the better deal, and we would rather argue about it than hide it. ## Where to go next - [glossary/per-seat-pricing](https://www.polarishq.co/glossary/per-seat-pricing) - [glossary/usage-based-pricing](https://www.polarishq.co/glossary/usage-based-pricing) - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) ## Questions people ask **What is the difference between per-seat and usage-based pricing?** Per-seat pricing charges a fixed rate for each person licensed to use a product, so the bill follows headcount. Usage-based pricing charges for units of something the product did, so the bill follows output. The first is predictable and unrelated to value delivered; the second is variable and directly related to it. **Is usage pricing always cheaper?** No. At high volume a meter can cost as much as the subscriptions it replaced, and the fifty-person stack page in this cluster shows a volume where it does. The argument for usage pricing is not that it is always smaller. It is that every line corresponds to something that arrived. **What is Polaris's billable unit?** A human-equivalent hour of work delivered by an AI worker, priced at roughly two dollars. The software itself is free with unlimited humans, tasks, workstreams and docs. Nothing delivered means nothing billed, in any month. **How do I know the hours are not inflated?** The formula is published and mechanical: base pickup time, twelve minutes per web search, finished prose at about ninety characters a minute, eight minutes per acceptance-criteria item, ten minutes per comment addressed and per file produced, five per PDF, clamped between five minutes and eight hours per session. Every input is recorded on the job's work log, so any line can be recomputed and challenged. **Why not just charge per token or per credit?** Because neither is a unit anyone can evaluate. A credit is an invented currency, and a token count tells a buyer nothing about whether the work was worth doing. An hour of human-equivalent effort is a unit a manager can compare against a salary, a contractor rate, or the time they would have spent themselves. **What happens if an AI worker does bad work?** You close nothing and rate nothing. The delivery sits on the task as a comment, and the task stays open. Because a human closes every task in Polaris, the acceptance decision is always yours, and the work log shows exactly what the hour estimate was built from if you want to argue about the line. ## Related - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/cost/stack-cost-50-person-team - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/alternatives --- --- title: "What AI subscriptions cost a team in 2026, checked August" description: "Claude, ChatGPT, GitHub Copilot and Cursor list prices checked August 2026, what a ten-person team pays, and why per-seat AI billing has no link to output." url: https://www.polarishq.co/cost/cost-of-ai-subscriptions section: Cost updated: 2026-08-21 --- # The cost of AI subscriptions for a team Five figures a year at fifty people, and no shared record of what any of it produced. ## The short answer Claude Pro lists at $20 a month, Claude Team at $25 per seat monthly or $20 annual, ChatGPT Business at $25 monthly or $20 annual, GitHub Copilot Pro at $10 a month, and Cursor Pro at $20 with Teams at $40 per user, checked 21 August 2026. A mixed ten-person team commonly pays about $4,440 a year. Polaris charges per hour delivered instead. - **Claude Team:** $25 monthly / $20 annual - **ChatGPT Business:** $25 monthly / $20 annual - **Copilot Pro:** $10 / month - **Checked:** 21 August 2026 ## AI subscription list prices Read from each vendor's own pricing page on 21 August 2026. Cursor's page was only partially readable, so this table quotes only the two rates that were unambiguous on it. | Product and plan | Monthly billing | Annual billing | Notes | | --- | --- | --- | --- | | Claude Free | $0 | $0 | Includes Claude Code | | Claude Pro | $20 / month | $17 / month, $200 upfront | Individual | | Claude Max | From $100 / month | Not published as a separate rate | 5x or 20x Pro usage | | Claude Team, standard seat | $25 / seat / month | $20 / seat / month | Teams of 2 to 150 | | Claude Team, premium seat | $125 / seat / month | $100 / seat / month | More usage than a standard seat | | ChatGPT Plus | $20 / month | Not offered | Individual | | ChatGPT Business, standard seat | $25 / user / month | $20 / user / month | From 2 users | | ChatGPT Business, premium seat | $125 / user / month | $100 / user / month | 5x standard usage | | GitHub Copilot Free | $0 | n/a | 2,000 completions a month | | GitHub Copilot Pro | $10 / month | Not published on the plans page | Unlimited completions | | GitHub Copilot Pro+ | $39 / month | Not published on the plans page | Premium models | | GitHub Copilot Max | $100 / month | Not published on the plans page | High-volume agent workflows | | Cursor Pro | $20 / month | Toggle present, rate not readable | Individual | | Cursor Teams | $40 / user / month | Toggle present, rate not readable | Team plan | ## A ten-person team's AI bill A realistic mixed configuration rather than a worst case: not everyone gets everything, and the annual rate is used wherever a vendor publishes one. | Line | Rate | Seats | Per month | Per year | | --- | --- | --- | --- | --- | | ChatGPT Business, annual | $20 / user / month | 6 | $120.00 | $1,440.00 | | Claude Team standard, annual | $20 / seat / month | 4 | $80.00 | $960.00 | | GitHub Copilot Pro | $10 / user / month | 5 | $50.00 | $600.00 | | Cursor Teams | $40 / user / month | 3 | $120.00 | $1,440.00 | | Total | | 10 people, 18 seats | $370.00 | $4,440.00 | | Per person per year | | 10 | $37.00 | $444.00 | ## Eighteen seats for ten people That is the shape of AI spending in 2026 and it is nobody's fault. Different tools are better at different work, so engineers end up with two subscriptions and everyone else has one, and the seat count exceeds the headcount. Each individual purchase was sensible. What the $4,440 does not buy is a record. There is no shared log of what came out of those eighteen seats, no artefact a colleague can pick up, and no way to answer at renewal what the money produced. The output lives in chat histories on individual accounts, and when someone leaves it goes with them. ## Four properties of per-seat AI that are worth naming - **Cost and usage are unrelated** — A seat costs the same whether it ran two prompts this month or two hundred. That is the defining property of the model, and it means no AI budget conversation can ever be about output. - **Allocation becomes a management task** — Because seats are scarce by price rather than by capacity, someone has to decide who gets one. Companies of fifty are running that argument quarterly, and it exists purely because of how the tools are priced. - **The work stays single-player** — Claude Code, Copilot and Cursor are excellent and they are local. They stop when the laptop closes, teammates cannot see or assign work to them, and there is no shared roster and no audit trail. - **Premium seats double the unit twice** — Both Claude Team and ChatGPT Business now sell a premium seat at $100 per seat per month on annual billing, five times the standard seat. The same access question comes back one tier up. ## Eighteen AI seats against AI workers **Per-seat AI subscriptions** - $4,440 a year for a ten-person team, checked 21 August 2026 - Priced by access, so a quiet month costs the same as a busy one - Output lives in individual chat histories - Nothing to assign, review, or hand to a colleague **AI workers in Polaris** - $0 for the workspace, no AI seats to allocate - Roughly $2 per human-equivalent hour actually delivered - A worker is a member of the org that anyone can assign a task to - Deliverables arrive as comments with files, on a task a person closes > **An illustrative comparison, and it is only that** > > If that ten-person team's AI workers delivered 100 human-equivalent hours in a month, the metered cost would be about $200, against $370 a month for eighteen seats. Polaris is in free public beta with no customers and no usage data, so treat the 100 hours as an assumption we chose, not a benchmark we measured. ## Where to go next - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cloud-claude-code/claude-code-for-teams](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) - [cloud-claude-code/cloud-agents-vs-local-agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) - [cost/stack-cost-50-person-team](https://www.polarishq.co/cost/stack-cost-50-person-team) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) ## Questions people ask **How much do AI subscriptions cost per person?** The common team rates checked on 21 August 2026 are $20 per seat per month on annual billing for both Claude Team and ChatGPT Business, $25 on monthly billing for each, $10 a month for GitHub Copilot Pro and $40 per user per month for Cursor Teams. Premium seats on Claude Team and ChatGPT Business are $100 per seat per month on annual billing. **Why does a ten-person team end up with eighteen AI seats?** Because the tools are good at different things. Engineers typically carry a coding assistant plus a general assistant, and everyone else carries one. Seat count exceeding headcount is the normal outcome of per-seat AI pricing, not a procurement failure. **Does Polaris replace these subscriptions?** It covers a different job. Claude Code, Copilot and Cursor are single-player tools that run on your machine; Polaris runs AI workers on cloud machines that keep working after you close your laptop, on tasks your teammates can see and assign. Many teams will keep a personal assistant subscription and still move delegated work into Polaris. **Is there a per-seat charge for AI workers in Polaris?** No. There are no AI seats and no access tiers, so there is nothing to allocate and no argument about who gets one. The only charge is roughly two dollars per human-equivalent hour that a worker delivers, itemised on that worker's log. **What if the delivered work is not good?** You do not close the task. In Polaris the machine never marks anything done: it posts its work as a comment, ticks its acceptance-criteria checklist and stops. A person closes and rates the task, and until they do, the delivery is open to dispute with the work log alongside it. ## Related - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/stack-cost-50-person-team - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/for/software-teams --- --- title: "What an AI worker costs: the $2-per-hour formula, in full" description: "The exact formula Polaris uses to estimate human-equivalent hours from observable effort, three worked examples, and what the estimate cannot tell you." url: https://www.polarishq.co/cost/what-an-ai-worker-costs section: Cost updated: 2026-08-21 --- # What an AI worker costs, and how the hours are counted The whole formula is on this page, including the parts that make it an estimate rather than a measurement. ## The short answer A Polaris AI worker costs about $2 per human-equivalent hour delivered, and nothing else. Hours are estimated from observable effort: a base of 15 minutes to pick up a task, 12 minutes per web search, finished prose at about 90 characters a minute, 8 minutes per checklist item ticked, 10 minutes per comment and per file, 5 per PDF, clamped to between 5 minutes and 8 hours per session. - **Rate:** ~$2 per human-hour delivered - **Session ceiling:** 8 hours, about $16 - **Software:** $0, unlimited people - **Billed when:** Work is delivered ## The formula, term by term This is the whole thing. Each term is a fixed rate applied to something the job observably did, and each input is written to the job's work log. | Term | Rate | What it represents | | --- | --- | --- | | Base pickup | 15 minutes, or 10 on a review job | What a person burns reading the task, opening the context and starting | | Web searches | 12 minutes each | Search, read and synthesise one source | | Finished prose | 90 characters per minute | Written output including the thinking behind it, not typing speed | | Acceptance-criteria items ticked | 8 minutes each | Fixed overhead per checklist item completed | | Comments addressed | 10 minutes each | Fixed overhead per comment answered on the task | | Files produced | 10 minutes each | Fixed overhead per file attached or document written | | PDFs | 5 minutes each | Additional overhead for a rendered document | | Clamp | Minimum 5 minutes, maximum 8 hours per session | No single agent session can bill more than eight human-equivalent hours | ## Three worked examples Same formula, three different jobs. You can check each of these with a calculator. 1. **A competitor research brief: 2.8 hours, about $5.57** — Base 15 minutes, plus 4 web searches at 12 minutes each for 48, plus 5,400 characters of finished prose at 90 a minute for 60, plus 3 acceptance-criteria items at 8 each for 24, plus 1 comment addressed for 10, plus 1 file produced for 10. That totals 167 minutes, which is 2.8 hours, which is about $5.57 at two dollars an hour. 2. **A document review: 1.1 hours, about $2.27** — Review jobs start from a base of 10 rather than 15. One web search for 12, 1,800 characters of finished prose for 20, two checklist items for 16 and one comment for 10 brings the total to 68 minutes, which is 1.1 hours, or about $2.27. 3. **A very large session: capped at 8 hours, about $16** — Twenty searches for 240 minutes, 40,000 characters of finished prose for 444, ten checklist items for 80 and a base of 15 comes to 779 minutes. The clamp cuts it to 480, so the job bills eight human-equivalent hours regardless. The ceiling exists so a single runaway session cannot produce a surprising invoice. ## What this formula is not It is not a measurement of how long a human would have taken. It is a rule applied to observable outputs, and the rates in it were chosen by us. Twelve minutes per search is a defensible estimate of search, read and synthesise; it is not derived from a study, and a fast researcher would beat it. It is also deliberately blunt. The five-minute floor almost never binds, because the base alone is ten or fifteen minutes. The eight-hour ceiling does bind on long sessions, and when it does, Polaris bills less than the formula produced rather than more. The reason to publish a simple formula rather than a clever one is that a simple formula can be argued with. Every input is on the work log: how many searches ran, how many characters came back, how many criteria were ticked. You can recompute any line on your bill and dispute it with the numbers in front of you. That is the actual product claim here, more than the two dollars. ## The comparison that made us pick this unit One competitor pricing brief run through the same delivery path, from the Polaris pricing work. A single example, not a benchmark. - **~7 min** — Wall-clock time to deliver. One agent session, real cloud machine - **~1h 24m** — Human-equivalent hours logged. 1.4 hours by the formula above - **$2.80** — What it billed. 1.4 hours at roughly two dollars - **~$70** — Same brief from a $50/hour analyst. At 1h 24m of that analyst's time ## What you get, and what you owe **Included at $0** - Unlimited humans, tasks, workstreams and docs - The Chief of Staff in every organisation from the first sign-in - Connections, file review, versioning, activity and the work log - Hiring an AI worker, which takes about sixty seconds in chat **Billed at ~$2 per hour** - Only human-equivalent hours a worker actually delivered - Nothing at all in a month where nothing was delivered - Each hour attached to a named job with its inputs logged - Capped at eight human-equivalent hours per session > **Agents deliver, humans close** > > The machine ticks its own acceptance criteria and posts the work as a comment with any files it produced. It never marks the task done. That is a product rule, not a setting: closing and rating a task is a human action in Polaris, which means the person paying is always the person who decided the work was finished. ## Where to go next - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [cost/cost-of-ai-subscriptions](https://www.polarishq.co/cost/cost-of-ai-subscriptions) ## Questions people ask **How much does an AI worker cost in Polaris?** Roughly two dollars per human-equivalent hour it delivers, and nothing for the software, the seat, or the act of hiring it. A worker that is on your roster but has not delivered anything this month costs nothing this month. **How do I know the hours are not inflated?** Recompute them. The formula is published in full on this page and every input is recorded on the job's work log: searches run, characters of finished prose, acceptance-criteria items ticked, comments addressed, files and PDFs produced. If a line does not match the log, it is challengeable, and that is the point of keeping the formula this simple. **Is a human-equivalent hour a real hour?** No, and calling it human-equivalent is meant to say so. It is an estimate of how long the same output would have taken a person, produced by a fixed rule from observable effort. The work itself lands far faster than the hours it logs, and that gap is the whole reason the unit is worth buying. **What is the most a single task can cost?** One agent session is clamped at eight human-equivalent hours, which is about sixteen dollars at roughly two dollars an hour. That ceiling is in the formula itself, so an unusually long session bills less than its raw effort rather than more. **What happens if an AI worker does bad work?** You do not close the task, and you rate it. The worker delivers as a comment and ticks its checklist, but only a person closes a task in Polaris. An unclosed delivery is an open dispute, with the work log sitting next to it showing exactly what the estimate was built from. **Does Polaris bill for failed or abandoned jobs?** Hours are logged against delivered work, job by job. Nothing delivered means nothing billed, which is why the pricing has no minimum, no platform fee and no seat charge to fall back on. ## Related - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/cost-of-ai-subscriptions - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail --- --- title: "Cloud AI agents for teams who already use Claude Code" description: "Fifteen pages on running AI agents on cloud machines instead of your laptop: the job queue, the session loop, the audit trail, and what Polaris actually is." url: https://www.polarishq.co/cloud-claude-code section: Cloud agents updated: 2026-08-21 --- # Cloud agents that keep working after you close the laptop For the person whose agent is brilliant, local, single-player, and dead the moment the lid goes down. ## The short answer Cloud agents are AI workers that execute on a server rather than on your laptop, so work continues when your machine sleeps. Polaris runs one: a queue table called agent_jobs, a worker process on Fly.io that claims a job, compiles the worker identity, runs a tool loop with live web search, and posts the result as a comment on the task. Humans close the task, never the machine. - **Pages in this cluster:** 15 - **Queue table:** agent_jobs - **Runtime host:** Fly.io, region sin - **Software price:** $0, billed per hour delivered ## The ceiling every local agent hits A terminal agent on your own machine is the best single-player tool most engineers have ever had. It reads your repository, runs your tests, uses your credentials, and iterates in the same second you think of something. Nothing in this cluster argues against that. The ceiling is not intelligence. It is the process. The agent lives inside a session on one laptop, owned by one person. Close the lid and it stops mid-thought. Your colleague cannot see it, cannot assign it anything, and cannot review what it produced without you pasting the transcript somewhere. There is no roster, no queue, no record of who asked for what. These fifteen pages are about moving that execution onto a machine that nobody has to keep awake, and putting a team around it: a shared list of workers, a queue with a contract, an append-only activity log, files with version history, and a rule that the machine never marks its own work finished. > **Polaris is not Claude Code** > > Claude Code is Anthropic's coding agent, and it runs in your terminal on your machine. Polaris is a separate product, built by superstack.digital, with no affiliation with or endorsement by Anthropic. Polaris runs its own agent runtime on cloud machines and calls the Anthropic API to do the thinking. If what you want is Claude Code, use Claude Code. If what you want is the capability people describe when they say "Claude Code in the cloud for my team", that is what this page is about. ## The job lifecycle, end to end Every page in this cluster refers back to these seven steps. They are the actual sequence in runtime/worker.mjs and the migrations behind it. 1. **Assignment writes a row** — A Postgres trigger on the tasks table fires when owner_id changes to a member whose kind is agent. It inserts one row into agent_jobs with status pending. A unique index allows only one live job per task, so double-assigning cannot spawn two machines on the same work. 2. **A worker claims it** — The runtime process polls every five seconds for the oldest pending job, then updates it to running with a WHERE clause that still requires status pending. If another process got there first the update returns nothing and this one moves on. The claim is the lock. 3. **Identity gets compiled** — The runtime reads the worker's own instructions, the full text of every SKILL.md-style playbook attached to that worker, the org's document tree, the task title, description, labels, bucket, checklist items and the last ten comments. Those become one system prompt and one task brief. 4. **The tool loop runs** — Up to ten rounds against the Anthropic Messages API, with server-side web search plus five tools the runtime implements itself: tick_checklist, post_progress, attach_file, create_doc and deliver. Every round's text and every search query is written to job_events, which the app streams live. 5. **Acceptance criteria get ticked** — The checklist on the task is handed to the worker as its acceptance criteria. Calling tick_checklist marks one item done and stamps the agent as its owner, so the box moves in everyone's browser through Realtime while the session is still running. 6. **Delivery is a comment** — The session ends when the worker calls deliver, which inserts a comment on the task authored by the agent's member row. Files land as real attachments in storage; written deliverables land as pages in Docs. The task status is set to in_progress, never done. 7. **The bill is written next to the work** — On success the job row stores human_minutes and the raw effort counters it was computed from: searches run, characters written, checklist items ticked, comments addressed, files produced. The estimate and its inputs sit on the same row, which is what makes a line on the invoice arguable. ## Where the work lives in each model Same agent quality in both columns. The difference is custody. | | Agent on your laptop | Agent on a Polaris machine | | --- | --- | --- | | Runs while you sleep | No, the session dies with the process | Yes, the queue drains on the server | | Who can assign it work | You | Anyone in the org, from the task | | Where the output goes | Your terminal scrollback | A comment, a file, or a Docs page | | Record of what happened | Shell history, if you kept it | activity_events plus a per-step job_events stream | | Credentials | Your local environment | Org-wide, server-side, no read path to the browser | | Repository and shell access | Full | None today, web search and writing only | | Who marks the work finished | You, informally | A human closes and rates it, by design | ## What this cluster covers Four questions, roughly, split across fifteen pages. - **Getting execution off your machine** — What actually keeps running when the laptop closes, what a queue contract guarantees, how long one session can last, and what happens when it fails halfway. - **Putting a team around the agent** — A shared roster where humans and agents sit in one members table, assignment that works identically for both, and an audit trail that answers who asked for this. - **Configuring what the worker is** — Skill files you can read and edit, org-wide tool credentials that the browser cannot read back, and instructions that compile into the prompt at job time. - **Getting work back you can use** — Deliverables as comments, attachments with version history, documents in the shared tree, and a review loop where commenting on the work wakes the worker again. ## Every page in this cluster - [Running an AI agent in the cloud instead of on your laptop](https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud) — Three approaches, one of which is ours, described precisely enough that you can tell which one you actually want. - [What keeps running after you close the lid](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) — The honest version: nothing on your laptop survives, so the work has to not be on your laptop. - [Turning a single-player agent into a team member](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) — The change is not a better prompt. It is putting the worker in the same table as the people. - [Sharing the agent, not the transcript](https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates) — There are two different things people mean by sharing, and only one of them survives contact with a second person. - [Queueing work at night and reading it in the morning](https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks) — The realistic version of overnight work, including the part where there is no scheduler. - [Running your AI workers from a phone](https://www.polarishq.co/cloud-claude-code/claude-code-from-your-phone) — A terminal agent needs a terminal. A queued job needs a text field and a network. - [Cloud agents versus local agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) — Most teams end up running both. The useful question is which work belongs where. - [Assigning a task to an AI worker](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) — There is no prompt box. The task is the prompt, and the checklist is the contract. - [The audit trail behind every agent session](https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail) — If you cannot reconstruct what happened three weeks later, you do not have an audit trail. You have a feeling. - [Skill files, and why a worker's capability should be readable](https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers) — The difference between a prompt and a playbook is that one of them is a document your colleague can edit. - [How an AI worker gets tool access](https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access) — The interesting part of tool access is not the list. It is where the credential lives and who can read it. - [How long an agent session can run, and what happens when it ends](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) — Every agent runtime has bounds. The useful thing a vendor can do is tell you what they are. - [When the deliverable is a file](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) — A chat reply is not a deliverable if the thing you needed was a document somebody can open. - [Agents deliver, humans close](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) — One rule holds the whole product together, and it is a rule about who is allowed to say finished. ## Related reading - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [integrations/web-search](https://www.polarishq.co/integrations/web-search) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [for/software-teams](https://www.polarishq.co/for/software-teams) ## Questions people ask **Is Polaris an official cloud version of Claude Code?** No. Claude Code is Anthropic's product and Polaris has no affiliation with Anthropic. Polaris is a separate workspace, built by superstack.digital, with its own agent runtime that calls the Anthropic API. Product names on this site belong to their respective owners and are used to describe compatibility of purpose, not partnership. **Can a Polaris worker check out my repository and run my tests?** Not today. The runtime gives a worker live web search and five tools for posting progress, ticking acceptance criteria, attaching files, drafting documents and delivering. There is no shell, no git checkout and no test runner in the session. For repository work, a local coding agent remains the right tool and Polaris is the place the resulting task, review and record live. **What happens if the runtime machine dies mid-job?** The job row stays in running with its claimed_at timestamp and an incremented attempts counter. Failures are caught and rewritten to pending for a retry while attempts is under two, then to error with the message stored on the row. Any work the model had already produced as comments, files or ticked checklist items is already committed to the database and survives. **Do I need my own Anthropic API key?** An organisation can store its own key, which is written to a table with no select policy so it never travels back to a browser. The runtime looks up the org key first and falls back to the key configured on the machine. Owners and admins are the only members who can write it. **How is any of this priced?** The software is free with no seats and no tiers. Billing is roughly two dollars per human-equivalent hour that a worker delivers, estimated by an open formula from observable effort and logged job by job. Nothing delivered means nothing billed. ## Related - https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/for/software-teams --- --- title: "How to run a coding-style AI agent in the cloud" description: "The three real ways to get an AI agent off your laptop, what each one costs you in setup and custody, and exactly what the Polaris cloud runtime does per job." url: https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud section: Cloud agents updated: 2026-08-21 --- # Running an AI agent in the cloud instead of on your laptop Three approaches, one of which is ours, described precisely enough that you can tell which one you actually want. ## The short answer Running an AI agent in the cloud means moving execution from your laptop to a server process that survives your session. Three approaches exist: keep your terminal agent alive on a rented box under tmux, drive an agent from CI on a trigger, or use a hosted runtime that owns the queue. Polaris takes the third approach, with a Postgres job queue and a worker process on Fly.io. - **Setup for the hosted path:** Assign a task - **Session deadline:** 8 minutes per job - **Shell access:** None ## What people mean by this question The search is usually not about hosting. It is about custody. Someone has a terminal agent that works well, and the constraint they have run into is that the agent belongs to one laptop and one person. They want the same quality of work to happen on a machine they do not have to babysit, and they want a colleague to be able to reach it. Claude Code is Anthropic's coding agent, and it runs in your terminal on your machine. Polaris is a separate product, built by superstack.digital, with no affiliation with or endorsement by Anthropic. Polaris runs its own agent runtime on cloud machines and calls the Anthropic API to do the thinking. If what you want is Claude Code, use Claude Code. If what you want is the capability people describe when they say "Claude Code in the cloud for my team", that is what this page is about. Anthropic ships Claude Code and changes what it offers on its own schedule. Check Anthropic's documentation for what is current there. The rest of this page is about the three architectures available to you regardless of which vendor's model is doing the thinking. ## The three architectures Pick by what you need to be true, not by what sounds modern. | | Terminal agent on a rented box | Agent driven from CI | Hosted runtime with a queue | | --- | --- | --- | --- | | Setup | Provision, install, keep a session alive | Write a workflow file and a trigger | Assign a task to a worker | | Repository and shell | Full | Full inside the runner | None | | Who can start work | Whoever holds the SSH key | Whoever can push or dispatch | Anyone in the org, from the task | | Where results land | That shell | Logs and artifacts | A comment, a file, a Docs page | | Ops burden | Yours | Yours | None on your side | | Fit | Solo engineer with a repo to work | Deterministic, repeatable pipelines | Teams handing work back and forth | ## What one hosted job does, in order This is the sequence in runtime/worker.mjs, not a marketing summary of it. 1. **The queue row appears** — Assigning a task to a member whose kind is agent fires a Postgres trigger that inserts a row into agent_jobs with status pending. Copilot members are excluded, and tasks already marked done are skipped. 2. **The runtime claims it** — A loop on the Fly machine polls every five seconds, takes the oldest pending row, and flips it to running with a conditional update that fails harmlessly if another process claimed it first. 3. **Context is assembled** — The worker's instructions, every attached skill document in full, the org's docs tree, the task's checklist, labels, bucket, due date and last ten comments are read in a handful of parallel queries and rendered into a system prompt and a brief. 4. **The loop runs with real tools** — Up to ten rounds against the Anthropic Messages API. Web search runs server-side with a cap of eight uses per session. Five further tools are implemented by the runtime and answered locally, whatever the stop reason. 5. **The session ends** — The worker calls deliver, which writes the final comment. If it never does, the runtime nudges once and then salvages the last block of text as the delivery, so a session that ran out of rounds still hands something back. ## What the hosted runtime does not do Read this before you decide. It is short and it matters. - **No shell** — There is no terminal in the session. The worker cannot run a build, a test suite or an arbitrary command. - **No repository checkout** — Nothing is cloned. Code work stays with your local agent, which is genuinely better at it. - **No scheduler** — Work starts because a human assigned a task or commented on one. There is no cron in the product today. - **No unbounded session** — One job is capped at ten model rounds and an eight-minute deadline. Long work is decomposed across sessions, on purpose. > **The swappable part** > > The contract is the agent_jobs table, not the machine. The queue is plain Postgres with a status enum, an attempts counter and a unique index permitting one live job per task. The Fly process is one consumer of that contract, and a different runtime could replace it without a single change in the app. ## Go deeper - [cloud-claude-code/cloud-agents-vs-local-agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) - [cloud-claude-code/claude-code-when-laptop-is-closed](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) ## Questions people ask **Can I point Polaris at my own machine instead of a hosted one?** The queue contract makes that possible in principle, because agent_jobs is an ordinary Postgres table and the runtime is one process reading it with a service-role key. Polaris does not ship a self-hosting path or documentation for it today, so treat this as an architectural property rather than a supported feature. **Which model runs the session?** Each worker can carry its own model setting, and the runtime falls back to a current Claude Sonnet model when the worker is set to inherit. The call is a plain Anthropic Messages API request with a tools array, made server-side from the Fly machine. **Does the agent see my whole workspace?** It sees the task it was assigned, that task's checklist and last ten comments, the bucket the task sits in, the titles of up to a hundred and twenty document pages so it knows where to file things, and its own skills and instructions. It does not receive the contents of other tasks or documents. **How fast does work start after I assign it?** The trigger writes the queue row inside the same transaction as the assignment, and the runtime polls every five seconds, so a free machine picks the job up in seconds. If a job is already running, the next one waits its turn in created_at order. **What does a session cost?** The software is free. A finished job stores a human-equivalent minute estimate on its own row, billed at roughly two dollars per hour, computed from counters such as searches run and characters written. A short research task typically lands in single-digit dollars, and a job that delivers nothing is not billed. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/integrations/web-search --- --- title: "Keep an AI agent working when your laptop is closed" description: "A local agent dies with its process. Here is exactly which pieces of a Polaris job survive a closed laptop, and what happens to work that was half finished." url: https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed section: Cloud agents updated: 2026-08-21 --- # What keeps running after you close the lid The honest version: nothing on your laptop survives, so the work has to not be on your laptop. ## The short answer An AI agent running in a terminal stops when the process stops, and closing a laptop stops the process. Work only continues if execution lives elsewhere. In Polaris the task, the queue row and the runtime process are all server-side, so a job that was claimed before you shut the lid keeps running and posts its comment, its files and its ticked checklist items whether or not any browser is open. - **What survives locally:** Nothing - **Poll interval:** 5 seconds - **Retries per job:** Up to 2 attempts ## Two failure stories Same task, same model, different place of execution. **Local session, lid closes** - The process is suspended or killed with the session - Partial output exists only in scrollback - Nothing was written anywhere a colleague can reach - Restarting means re-establishing the whole context by hand - Nobody else knew the work was in flight **Queued job, lid closes** - The claimed row keeps running on the Fly machine - Progress comments and ticked items are already in Postgres - Attachments are already in storage with a version number - Realtime replays the state to any browser that opens later - The activity feed shows who assigned it and when ## Why the browser is not part of the machinery The Polaris frontend is a static export talking to Postgres directly, with row-level security deciding what each member can read. It subscribes to changes on tasks, comments, jobs and the per-step event stream. It is a viewer. The runtime is a separate long-lived process on Fly.io holding a service-role key. It never talks to a browser. It reads its next job from the same database the app reads, does the work, and writes the results back. Close every tab in your company and the loop continues, because no step in it needs a client. ## What happens to a job while you are asleep 1. **Claimed work finishes** — A job already flipped to running continues through its rounds until it calls deliver or hits the eight-minute deadline. Comments, ticks, docs and files are written as they happen, not at the end. 2. **Queued work waits its turn** — Pending rows sit in created_at order. The runtime takes the oldest each cycle, so a backlog assigned at midnight drains one job at a time rather than all at once. 3. **A failure retries itself** — An exception writes an error event, then sets the row back to pending if it has been attempted fewer than twice. The third failure parks it in error with the message stored on the row. 4. **Nothing gets closed** — The machine sets the task to in_progress and stops there. A delivered task is waiting for you in the morning, not silently marked done. ## What still stops when you stop - **Anything you queued but never assigned** — Work starts on assignment. A task sitting in a bucket with no owner sits there all night. - **A worker waiting on a question** — Workers are instructed to deliver a clear statement of what they need rather than guess. That delivery is a comment you answer in the morning, which then wakes the worker again. - **Anything needing a credential nobody added** — Connections are authorised once, org-wide, by an owner or admin. A missing credential is not something a machine can resolve at three in the morning. > **The test that matters** > > Assign one task, shut the laptop, and open the task on your phone an hour later. If a comment, a ticked checklist and an attached file are waiting, the work genuinely happened somewhere other than your machine. That is the only demonstration of this claim worth trusting, including ours. ## Related reading - [cloud-claude-code/claude-code-overnight-tasks](https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) - [cloud-claude-code/claude-code-from-your-phone](https://www.polarishq.co/cloud-claude-code/claude-code-from-your-phone) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) ## Questions people ask **Does my browser need to stay open for the agent to keep working?** No. The runtime is a server process holding its own database credential, and it never communicates with a browser. The app subscribes to the same tables to display what is happening, so closing it changes the view and nothing else. **What if the job was mid-search when I closed the lid?** Web search runs server-side as part of the model call, so it was never happening on your machine in the first place. The session continues through its remaining rounds on the Fly machine. **Is partial work lost when a job fails?** No. Progress comments, ticked checklist items, drafted documents and uploaded files are each committed to Postgres or storage the moment the tool call succeeds. A failure later in the session leaves all of that intact and adds an error event describing what went wrong. **Can I see what happened while I was away?** Yes. Every session writes an ordered event stream with kinds such as session, thought, search, tool, progress, delivered, error and done. Opening the task later replays the stream in sequence alongside the comments and files it produced. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/cloud-claude-code/claude-code-from-your-phone - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/agent-runtime --- --- title: "AI coding agents for teams, not just one laptop" description: "Terminal agents are single-player by construction. Here is the data model that makes an AI worker assignable, visible and reviewable by a whole team." url: https://www.polarishq.co/cloud-claude-code/claude-code-for-teams section: Cloud agents updated: 2026-08-21 --- # Turning a single-player agent into a team member The change is not a better prompt. It is putting the worker in the same table as the people. ## The short answer Making an AI agent work for a team requires three things a terminal session cannot provide: a shared identity the whole org can address, a queue that accepts work from anyone, and a record of what was asked and delivered. Polaris stores humans and AI workers in one members table distinguished by a kind column, so assignment, mentions, activity logging and permissions behave identically for both. - **Roster table:** members - **Distinguishing column:** kind: human or agent - **Seat cost per teammate:** $0, human or agent ## Why single-player is a structural property A terminal agent is single-player for a reason that has nothing to do with the model. Its identity is your shell, its memory is your session, its permissions are your credentials, and its output is your scrollback. Every one of those is scoped to one person by construction. Sharing it therefore means copying: pasting a transcript into Slack, exporting a file, retelling the reasoning in standup. That copy is where the record breaks. Nobody can later ask which version of the brief produced this, or who approved the result. ## What the same-table decision actually buys Agents are members. That single choice removes most of the special cases. - **Assignment is one dropdown** — The owner field on a task points at a member row. Choosing a person and choosing a worker are the same interaction, and the only difference downstream is that a trigger notices kind equals agent and writes a queue row. - **Permissions are one policy** — Row-level security asks whether you belong to the org and whether you can see the task. Workers act through a service-role key on the server and are recorded as the actor via their member id, so the audit trail names them like it names anyone else. - **Activity is one feed** — Created, assigned, moved, started, completed, reopened, commented, attached: the same event kinds are written by the same triggers whether the actor is a person or a machine. - **Mentions work across kinds** — A comment on an agent-owned task wakes its owner. Writing another worker's first name in a comment wakes that worker too, which is how a second pair of eyes gets pulled onto a task. - **Ratings live on the work** — A thumbs up or down is stored once per task per rater against the worker's member id, so quality is attached to a named teammate rather than to a vibe. ## The same task, two owners What differs when a task is owned by a person versus a worker. | Step | Human owner | AI worker owner | | --- | --- | --- | | Assignment | owner_id set to their member row | owner_id set to their member row | | What happens next | They see it in Focus | A trigger inserts a pending job row | | Execution | Wherever they work | A claimed session on the runtime machine | | Progress signal | They move the task | Progress comments and a live event stream | | Delivery | They mark it done | A delivery comment, status stays in_progress | | Closing | Themselves | A human closes and rates it | ## The naming question, stated plainly Claude Code is Anthropic's coding agent, and it runs in your terminal on your machine. Polaris is a separate product, built by superstack.digital, with no affiliation with or endorsement by Anthropic. Polaris runs its own agent runtime on cloud machines and calls the Anthropic API to do the thinking. If what you want is Claude Code, use Claude Code. If what you want is the capability people describe when they say "Claude Code in the cloud for my team", that is what this page is about. ## What a team pays for this - **$0** — Per seat, human or AI. No tiers, no per-user pricing - **~$2** — Per human-equivalent hour delivered. Estimated by an open formula, logged per job - **0** — Charged for work not delivered. A failed job produces no billable minutes ## Related reading - [cloud-claude-code/share-claude-code-with-teammates](https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates) - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [glossary/ai-teammate](https://www.polarishq.co/glossary/ai-teammate) - [for/software-teams](https://www.polarishq.co/for/software-teams) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) ## Questions people ask **Do AI workers count as seats on the bill?** No. There is no seat charge for anyone in Polaris, human or AI. The software is free with unlimited members, tasks, workstreams and documents, and revenue comes only from delivered agent hours at roughly two dollars per human-equivalent hour. **Can two people assign work to the same worker at once?** Yes, and both tasks become queue rows. A unique index permits only one live job per task, so the same task cannot be double-started, while separate tasks queue in creation order and are worked one after another. **Can a worker create tasks for humans?** Agents deliver into the task they were assigned rather than filing new work autonomously. The Chief of Staff copilot is the surface that turns conversation and incoming signals into prefilled task suggestions, and each suggestion waits for a human click before it becomes a task. **How does a teammate know an agent is working right now?** The worker's member row carries an agent_status column that flips to working when a job is claimed and back to idle when it finishes, and the app subscribes to that table. Alongside it, the per-step event stream for the running job is visible while the session is live. **Does everyone in the org see every agent's work?** Visibility follows normal task permissions. Job rows and their event streams are readable by org members subject to the same check that governs the task itself, and private buckets restrict that check further. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/glossary/ai-teammate - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/for/software-teams - https://www.polarishq.co/cost/per-seat-vs-usage-pricing --- --- title: "How to share an AI agent's work with your teammates" description: "Pasting a terminal transcript into Slack is not sharing. Here is what a shared AI worker looks like: one roster, one queue, deliverables with version history." url: https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates section: Cloud agents updated: 2026-08-21 --- # Sharing the agent, not the transcript There are two different things people mean by sharing, and only one of them survives contact with a second person. ## The short answer Sharing an AI agent with teammates means giving them a way to assign it work and inspect its output without going through you. Copying a transcript does not do that. Polaris shares the worker itself: one roster row anyone can assign to, one queue that accepts work from any member, deliverables posted as comments with attached files, and an append-only record of who asked for what. - **What gets shared:** The worker, not the session - **Deliverable format:** Comment, file, or Docs page - **Re-brief method:** Comment on the task ## Two definitions of sharing **Sharing a transcript** - A snapshot, already stale when pasted - Your colleague cannot ask it a follow-up - The brief that produced it is not attached - Files arrive as uploads with no version history - You remain the only route to the agent **Sharing the worker** - A roster entry any member can assign a task to - A colleague comments and the worker wakes again - The brief, checklist and comments live on the task - Files carry versions, with the old one preserved - The worker answers to the team, not to you ## How a teammate uses a worker you configured No handover, no credentials, no explanation of your prompt. 1. **They open a task and pick the worker** — The worker appears in the same owner picker as every human. Assignment writes a queue row through the same trigger, whoever did the assigning. 2. **They write acceptance criteria** — Checklist items on the task are handed to the worker as its acceptance criteria, so the brief is structured rather than folkloric. 3. **They watch the session or walk away** — The per-step event stream renders live for anyone with access to the task. Nobody needs to keep it open for the work to finish. 4. **They read the delivery in place** — The result is a comment authored by the worker, plus any files attached to the task and any pages drafted into the shared Docs tree. 5. **They push back in the thread** — A comment from anyone other than the worker itself queues a follow-up session, with the latest human comment treated as the new brief. ## What travels with the work Everything a second person needs in order to trust the output. - **The instructions the worker ran under** — Instructions and attached skill documents are org-visible records, not a prompt hidden in someone's shell history. - **The searches it ran** — Each web search is written to the session event stream with the query text, so a reviewer can see what it looked at rather than inferring it. - **The acceptance criteria it ticked** — Ticked items record the worker as the owner of the tick, which makes an unticked box a visible gap rather than a silent one. - **The effort counters behind the bill** — Searches, characters written, items ticked, comments addressed and files produced are stored on the job row next to the estimate they generated. > **The mention trick** > > Writing another worker's first name in a comment on a task queues a session for that worker as well. A researcher delivers a brief, someone writes a line asking the writer to turn it into a draft, and a second worker picks up the same task with the full comment history already in its brief. ## Related reading - [cloud-claude-code/claude-code-for-teams](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [cloud-claude-code/ai-agents-that-produce-files](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [integrations/slack](https://www.polarishq.co/integrations/slack) ## Questions people ask **Can my teammate change how the worker behaves?** Yes, if they have access. Instructions and the skill library are org-wide records readable and editable by members, so tuning a worker is a normal edit rather than a request routed through whoever created it. Connection credentials are the exception and are restricted to owners and admins. **How does a colleague ask for a revision?** They comment on the task. A comment from anyone other than the worker itself inserts a follow-up job, and the runtime is instructed to treat the latest human comment as the brief, answer questions in-thread and avoid redoing finished work that nobody asked about. **Can I keep some agent work private?** Buckets can be private, and job rows and their event streams are readable only where the underlying task is visible to you. Work in a private bucket stays inside it rather than appearing in the org-wide activity feed for everyone. **Does sharing a worker mean sharing my API key?** No. An organisation stores one Anthropic key in a table with no select policy, writable only by owners and admins and readable only by server-side code. Members use the worker without ever holding or seeing the key. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/glossary/ai-teammate - https://www.polarishq.co/integrations/slack --- --- title: "Running AI agent tasks overnight while you sleep" description: "What an overnight AI agent queue can and cannot do: how jobs drain in order, why there is no cron in Polaris yet, and how to write a brief you can leave alone." url: https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks section: Cloud agents updated: 2026-08-21 --- # Queueing work at night and reading it in the morning The realistic version of overnight work, including the part where there is no scheduler. ## The short answer Overnight AI agent work means queueing tasks before you stop for the day and reading the deliveries when you return. In Polaris the queue is a Postgres table drained by a runtime process that polls every five seconds and works the oldest pending job first. Work starts because a task was assigned or commented on. There is no cron scheduler in the product today. - **Trigger for work:** Assignment or a comment - **Scheduler:** Not in the product today - **Queue order:** Oldest pending first ## Say the limitation first Polaris has no scheduler. You cannot tell a worker to run something every night at two. Work begins when a human assigns a task to a worker, or when someone comments on a task a worker already owns. That is the whole set of triggers, and the architecture notes name scheduling as a deliberate exclusion rather than an oversight. What you can do is stack the queue. Assign six tasks at six in the evening and the runtime works them in creation order while nobody is watching, writing progress, files and deliveries into the database as it goes. In the morning six tasks sit in progress with a delivery comment each, waiting for someone to close them. ## How to set up an evening batch Fifteen minutes of setup buys a queue that runs itself. 1. **Split the work into task-sized units** — One session is capped at ten model rounds and eight minutes, so a task should be something a careful person could finish in one sitting. Three narrow tasks beat one enormous one, and they run one after another anyway. 2. **Write the checklist before you assign** — Checklist items are handed over as acceptance criteria and the worker ticks them as it goes. An empty checklist means nothing to verify against in the morning. 3. **Say where the output belongs** — Written deliverables are drafted into the Docs tree by default and the worker chooses a parent page whose topic matches, or files under Unsorted. Naming the destination in the description removes the guess. 4. **Assign in the order you want them worked** — Pending jobs are taken oldest first, so the sequence you assign in is the sequence you get. 5. **Leave. Genuinely leave** — No browser needs to stay open. The runtime holds its own database credential and writes every result server-side. ## What you find in the morning - **A delivery comment per task** — The complete deliverable posted as a comment authored by the worker, with plain source URLs where it researched. - **Documents in the shared tree** — Briefs, reports and plans drafted as pages with real block structure, filed under a parent page or under Unsorted. - **Files where a file was the point** — Data as CSV or JSON, board-ready documents typeset as PDF, each attached to the task with a size, a type and an uploader. - **A replayable session** — The ordered event stream for each job, from the opening session line through searches and tool calls to the delivery. - **An hours line per job** — The human-equivalent estimate stored with the effort counters that produced it, ready to be challenged. - **Nothing marked done** — Every task sits in progress. Closing and rating stays with the person who owns the outcome. > **Where an overnight batch goes wrong** > > The common failure is a task too vague to attempt. Workers are instructed to deliver an explanation of exactly what they need rather than guess at scope, so an ambiguous brief returns a question, not a draft. That is the correct behaviour and it is still a wasted night if you left six of them. ## Related reading - [cloud-claude-code/claude-code-when-laptop-is-closed](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) ## Questions people ask **Can I schedule a recurring agent task?** Not today. Polaris starts work from assignment or from a comment on an agent-owned task, and recurring schedules are listed as a deliberate non-goal in the current architecture. Assigning a task each evening is the manual equivalent. **How many jobs run at the same time?** The runtime claims one job per cycle and works it to completion before taking the next, so a batch drains sequentially rather than in parallel. Each individual session is capped at eight minutes, which keeps a stuck job from blocking a queue all night. **What if a job fails at three in the morning?** The error is written to the session stream, the worker's status returns to idle, and the row goes back to pending for another attempt if it has been tried fewer than twice. After that it is parked with the error message on the job row, and anything already committed during the session remains on the task. **Will I be billed for a night that produced nothing?** Human-equivalent minutes are written only when a job finishes successfully. A job that ends in error carries no estimate and nothing to bill. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/glossary/acceptance-criteria --- --- title: "Assign and check AI agent work from your phone" description: "If execution is server-side, a phone is enough to start work and read the result. What the Polaris mobile app does today, and what it honestly does not." url: https://www.polarishq.co/cloud-claude-code/claude-code-from-your-phone section: Cloud agents updated: 2026-08-21 --- # Running your AI workers from a phone A terminal agent needs a terminal. A queued job needs a text field and a network. ## The short answer Assigning AI work from a phone is possible when execution happens on a server rather than on a device. Polaris ships a Flutter mobile app, live now as an installable web app, built around the copilot with voice input and spoken replies. Assigning a task from it writes the same queue row a desktop assignment writes, and the delivery comment appears on both. - **Mobile app:** Flutter, installable PWA - **Input:** Typing or voice - **App Store status:** Prepped, not released ## Why this works at all Nothing about a Polaris job runs on the device that started it. Assignment writes a row in Postgres, a trigger queues the job, and a process on a Fly machine claims it. The client is a viewer of a database, so any client that can write one row and subscribe to a few tables is sufficient. That is the practical difference from a terminal agent. Reaching a shell from a phone means SSH, a keyboard you do not want, and a session that dies when the connection drops. Reaching a queue means typing a sentence. ## What the mobile app does today Described from the shipped app, not from a roadmap. - **The copilot sits at the centre** — The Chief of Staff is the primary surface rather than a list of projects, which suits a screen where you have one thumb and one intention. - **Voice in, voice back** — Spoken input and spoken replies are part of the interface, so briefing a worker on a walk is a normal use rather than an accessibility afterthought. - **The same workspace** — It is the same database and the same organisation as the desktop app, with the same tasks, buckets, comments and workers. - **Installable now** — The app is live as an installable web app you can add to a home screen without an app store. ## What a phone is good and bad at here | Action | On a phone | Better on a desktop | | --- | --- | --- | | Assigning a task to a worker | Fine | | | Briefing by voice | Better than typing | | | Reading a delivery comment | Fine | | | Watching a live session stream | Fine | | | Writing a detailed checklist | | Yes | | Reviewing a long document with comments | | Yes | | Configuring connections and credentials | | Yes | > **The honest status** > > The mobile app is built and running as an installable web app with its store bundle prepared. It has not shipped through the App Store, and the TestFlight build is waiting on an App Store Connect key. Polaris as a whole is in free public beta with no customers to point at. ## Related reading - [cloud-claude-code/claude-code-when-laptop-is-closed](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) ## Questions people ask **Can I run a coding agent on my phone with this?** No. The Polaris runtime has no shell and no repository checkout, on any device. What a phone gives you is the ability to assign research, writing, analysis and review work to a worker and to read what came back. **Do I need to install anything?** No. The mobile app is live as an installable web app, so a browser and an add-to-home-screen is the whole installation. A native store release is prepared but not published. **Does voice input change what the worker receives?** No. Voice is an input method on the client. What reaches the queue is the same task title, description, checklist and comment text a keyboard would have produced. **Will I be notified when a delivery arrives?** The app subscribes to task changes, comments and job events in realtime, so a delivery appears while the app is open. Push notifications to a closed phone are not part of the product today. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/for/solo-founders --- --- title: "Cloud agents vs local agents: an honest comparison" description: "Local agents own your filesystem and your credentials. Cloud agents own durability and shared visibility. A row-by-row comparison, including where local wins." url: https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents section: Cloud agents updated: 2026-08-21 --- # Cloud agents versus local agents Most teams end up running both. The useful question is which work belongs where. ## The short answer A local agent runs in a process on your own machine with your filesystem, your shell and your credentials, and stops when that process stops. A cloud agent runs on a server, keeps working when your device sleeps, and writes its output somewhere a team can read. Local agents win on repository work and iteration speed. Cloud agents win on durability, shared access and record-keeping. - **Local strength:** Filesystem and shell - **Cloud strength:** Durability and shared access - **Common answer:** Run both ## Row by row Written from the Polaris runtime on the cloud side. Other cloud runtimes make different trade-offs. | Property | Local agent | Cloud agent (Polaris runtime) | | --- | --- | --- | | Execution host | Your machine | A Fly.io machine, region sin, shared CPU with 512MB | | Survives a closed lid | No | Yes | | Filesystem access | Your whole disk | None | | Shell and build tools | Yes | No | | Repository checkout | Yes | No | | Live web search | Depends on the tool | Server-side, capped at 8 uses per session | | Who can start work | The person at the keyboard | Any member, by assigning a task | | Concurrency control | Whatever you remember | One live job per task, enforced by a unique index | | Output destination | Terminal, and files you already own | Task comment, attachment with versions, Docs page | | Record of the run | Scrollback | Ordered event stream plus an activity feed | | Credential custody | Your local environment | Org-wide, server-side, no read path to a browser | | Session bound | Your patience | 10 model rounds, 8-minute deadline, 2 retry attempts | | Cost model | Your own API bill or plan | Free software, ~$2 per human-equivalent hour delivered | ## Work that belongs on a local agent This is not a concession. It is most engineering work. - **Anything touching the repository** — Refactors, test runs, migrations, dependency work. A cloud runtime with no checkout cannot compete and should not pretend to. - **Tight iteration loops** — When you want to see a result, adjust one word and try again, latency and immediacy beat everything else. - **Work with local-only secrets** — Things that depend on your machine's environment, VPN or device-bound credentials. - **Exploration you will throw away** — If the artifact is understanding rather than a deliverable, a durable record adds nothing. ## Work that belongs on a cloud agent - **Research with sources** — Live search, cross-checking, and a brief that ends with the URLs it actually opened, delivered where a team can quote it. - **Anything a second person must read** — If the output has an audience beyond you, it needs a home with a version history rather than a paste. - **Work that outlives your session** — Long compilations, monitoring sweeps, batches you queue in the evening. - **Work someone else should be able to start** — A colleague assigning a task at midnight should not require you to be awake. - **Anything that will be argued about later** — Ticked acceptance criteria, an ordered event stream and stored effort counters exist precisely for the argument. ## Names, precisely Claude Code is Anthropic's coding agent, and it runs in your terminal on your machine. Polaris is a separate product, built by superstack.digital, with no affiliation with or endorsement by Anthropic. Polaris runs its own agent runtime on cloud machines and calls the Anthropic API to do the thinking. If what you want is Claude Code, use Claude Code. If what you want is the capability people describe when they say "Claude Code in the cloud for my team", that is what this page is about. So the comparison on this page is architectural. It is local execution against server execution, not one vendor's product against another's. > **The split most teams land on** > > Code stays local, because that is where the repository and the test suite are. Research, writing, analysis, competitive sweeps, document review and anything with an audience move to the queue, because that is where durability and a record matter more than a shell. ## Related reading - [cloud-claude-code/run-claude-code-in-the-cloud](https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) - [cloud-claude-code/give-an-ai-agent-tool-access](https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/cloud-development-environment](https://www.polarishq.co/glossary/cloud-development-environment) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) ## Questions people ask **Is a cloud agent slower than a local one?** Startup is slower and iteration is much slower, because you are writing a task rather than typing into a live session. Throughput on unattended work is higher, because the session continues without you. Choose by which of those two matters for the piece of work in front of you. **Can a cloud agent read my codebase?** The Polaris runtime cannot, because it performs no checkout and has no filesystem access to your machine. GitHub is in the connection catalog as an org-wide credential, so repository access is a credential question rather than a checkout question, and code execution is not part of the session either way. **Which is safer?** They fail differently. A local agent runs with whatever your shell can reach, which is broad and hard to audit after the fact. A cloud agent runs with credentials an admin authorised org-wide, stored in tables with no client read path, and writes an ordered record of every step it took. **Do I have to pick one?** No, and most people who search for this end up running both. The practical division is that anything requiring a shell stays local and anything requiring an audience or durability moves to the queue. **What limits a single cloud session?** Ten model rounds, an eight-minute deadline, eight web searches, a two-megabyte cap on any single attached file, and up to a hundred and twenty blocks in a drafted document. Longer work is decomposed into more tasks rather than one longer session. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/cloud-development-environment - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/integrations/github --- --- title: "How to assign work to an AI agent properly" description: "Assignment is the whole interface: pick the worker, write acceptance criteria, walk away. What a good agent brief contains and what happens when you save it." url: https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent section: Cloud agents updated: 2026-08-21 --- # Assigning a task to an AI worker There is no prompt box. The task is the prompt, and the checklist is the contract. ## The short answer Assigning work to an AI agent in Polaris means setting a task's owner to a member whose kind is agent. A Postgres trigger immediately inserts a job into the queue, the runtime claims it within seconds, and the task's title, description, labels, checklist and last ten comments become the brief. The checklist doubles as acceptance criteria, which the worker ticks as it satisfies each item. - **The interface:** The owner field on a task - **The contract:** The checklist - **Time to first activity:** Seconds ## Why there is no prompt box A chat box produces a request that exists only in one conversation. A task produces a record with an owner, a bucket, a due date, labels, a description, a checklist and a comment thread, all of which persist and all of which a second person can read. So Polaris uses the task as the unit of instruction. Assignment is the send button. Everything you would have typed into a prompt already has a field, and the fields survive the session that consumed them. ## What happens the moment you assign 1. **The trigger fires** — An after-insert-or-update trigger on the tasks table checks whether the new owner is a member with kind agent that is not the copilot, and whether the task is not already done. If so it inserts a pending job row carrying the org, the task and the agent. 2. **The index protects you** — A unique index over task_id, limited to jobs in pending or running, means a second assignment while one is live is quietly ignored rather than starting a second machine on the same work. 3. **The runtime claims it** — Within one five-second poll, the job flips to running, the worker's status flips to working, and the task moves to in progress. Everyone watching sees all three at once through Realtime. 4. **The brief is rendered** — Title, bucket, due date, labels, description, every checklist item with its current state, and the last ten comments with their authors, are formatted into one message. The worker's instructions and skills become the system prompt. 5. **It states a plan** — Workers are instructed to post one short progress comment early, saying what they intend to do. That comment is the earliest point at which you can stop a session heading the wrong way. ## What makes a brief a worker can actually execute The failure mode is vagueness, and the fix is boring. - **A title that names the deliverable** — Not review pricing, but a comparison table of five competitors' published pricing with sources. The worker is told to deliver something concrete and specific, and the title sets what concrete means. - **Checklist items that can be verified** — Each item is an acceptance criterion the worker ticks by name. An item like do a good job cannot be ticked honestly, and an item like include the effective date for each price can. - **A destination for the output** — Written deliverables go into the Docs tree by default, under a parent page whose topic matches or under Unsorted. Naming the page you want removes the guess. - **The reason, in one line** — A worker that knows a brief is for a board meeting on Thursday makes different choices about length and format than one that does not. - **Scope small enough for one sitting** — A session is capped at ten rounds and eight minutes. Work that exceeds that should be several tasks, which also gives you several checkpoints. > **The rule that never bends** > > The machine sets the task to in progress and never to done. Delivery and completion are different events with different owners: the worker delivers, and the person who owns the outcome reads it, closes it and rates it. ## Related reading - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [cloud-claude-code/skill-files-for-ai-workers](https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers) - [cloud-claude-code/claude-code-overnight-tasks](https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) ## Questions people ask **What if I assign a task with no checklist?** The brief says the checklist is empty and the worker proceeds on the title and description alone. It still delivers, but there is nothing structured to verify the result against, which makes review slower and disagreements harder to settle. **Can I change the task after assigning it?** Editing the task does not restart a running session, because the brief was rendered at claim time. Commenting does: a comment from anyone other than the worker queues a follow-up session in which the latest human comment is treated as the new brief. **How do I stop a session I regret?** Sessions are short by design, with a hard eight-minute deadline and a ten-round cap, so the practical answer is to let it finish and then redirect it with a comment. There is no kill switch on a running job in the product today. **Can I assign the same task to a person and an agent?** A task has one owner. The pattern that works is to assign the worker, let it deliver into the comment thread, and leave the human as the person who reviews and closes. Mentioning another worker by first name in a comment pulls it onto the same task without changing the owner. **Does the worker see the whole bucket?** It sees the bucket's name for context, not its other tasks. Its window is the task it was assigned, that task's checklist and recent comments, its own instructions and skills, and the titles of the org's document pages so it knows where to file written work. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers - https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/ai-worker --- --- title: "What an AI agent audit trail should record" description: "Six records make agent work auditable: the activity feed, session events, comments, file versions, doc snapshots and the effort counters behind the bill." url: https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail section: Cloud agents updated: 2026-08-21 --- # The audit trail behind every agent session If you cannot reconstruct what happened three weeks later, you do not have an audit trail. You have a feeling. ## The short answer An AI agent audit trail records who asked for the work, what the agent did, what it produced and who accepted it. Polaris writes six separate records: an append-only activity feed, a per-step session event stream, the comment thread, file version history, document snapshots, and the effort counters stored beside each job's human-hour estimate. Members can read these; only triggers and server-side code can write them. - **Records kept:** 6 - **Session event kinds:** 8 - **Client write access:** None ## The six records Each answers a different question, which is why one of them is not enough. - **Activity feed** — An append-only table written by database triggers, covering created, assigned, unassigned, moved, started, completed, reopened, urgent on and off, due changed, commented, attached and deleted. It answers who did what to this task. - **Session event stream** — Per job, an ordered sequence numbered from one, with kinds session, thought, search, tool, progress, delivered, error and done. Each search event carries the query text. It answers what the machine actually did. - **The comment thread** — Progress notes and the final delivery are comments authored by the agent's member row, sitting in the same thread as human replies. It answers what was said and by whom. - **File version history** — Each revision of an attachment preserves the previous path, size, version number and author before the new one is written. It answers what this file looked like before. - **Document snapshots** — Updating a document in place first stores a full snapshot of the title and every block. It answers what this page said last week. - **Effort counters and the estimate** — A finished job stores its human-minute figure alongside the raw counters it was computed from. It answers why this line on the bill says what it says. ## Reconstructing a session after the fact The order to read them in when something looks wrong. 1. **Start at the activity feed** — Find the assigned event. It names the actor, the worker and the moment, which fixes who set this in motion. 2. **Open the job row** — It carries the status, the attempts counter, the claim time, the finish time and, on failure, the error message that ended it. 3. **Replay the event stream** — Read the events in sequence number order. Search events show the exact queries, tool events show each call, and thought events show the model's own text between calls. 4. **Read the thread** — The early progress comment states the plan the worker announced. The delivery comment is the claimed result, with source URLs where it researched. 5. **Diff the artifacts** — Compare the current attachment against its stored versions, or the current document blocks against the snapshot taken before the last update. 6. **Check the arithmetic** — The effort counters on the job row multiply out through a published formula. If the search count does not match the search events, that is a discrepancy you can point at. ## Who can write what Job rows carry a read policy for org members and no client write path beyond creating a review job on something you can already see. Activity events and session events are readable by members and written by triggers and the runtime's service credential, never by a browser. The connection and API-key tables go further. They have no select policy at all, so a credential cannot be read back through the client API by anyone, including the admin who set it. What members can query is a function that lists which connections hold a working credential, without ever returning one. > **The point of writing the counters down** > > Any usage-priced product can put a number on an invoice. Storing the searches, characters, ticks, comments and files that produced the number, on the same row as the number, is what makes the line arguable. An audit trail you cannot use to dispute a charge is decoration. ## Related reading - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [cloud-claude-code/ai-agents-that-produce-files](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) ## Questions people ask **Can an agent delete its own history?** No. The activity and session event tables are written through triggers and the runtime's server-side credential, and neither exposes a delete path to a member client. Deleting the task removes its rows by cascade, which is a human action recorded as a delete event on the feed. **How long is the session stream kept?** Events live as ordinary rows tied to their job, so they persist as long as the job does, and job rows persist until the task is deleted. Individual event bodies are truncated at fifteen hundred characters, so a very long block of model text is stored clipped rather than in full. **Does the audit trail record what the model was thinking?** It records the text the model produced between tool calls, stored as thought events, plus every search query and every tool call with a short summary. It is a record of observable actions and outputs, not of internal state. **Is any of this exportable?** Everything described here is ordinary Postgres in your own project's database, queryable with SQL. Polaris does not ship a one-click audit export in the product today. **Who can see an agent's work in a private bucket?** Job rows and their event streams are readable only when the underlying task is visible to you, and private buckets narrow that check. Work done inside a private bucket does not surface in the org-wide activity view for members without access. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "Skill files: how to give an AI worker a playbook" description: "A skill file is a markdown playbook an AI worker loads at job time. How Polaris stores them, what the starter library contains, and how they reach the prompt." url: https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers section: Cloud agents updated: 2026-08-21 --- # Skill files, and why a worker's capability should be readable The difference between a prompt and a playbook is that one of them is a document your colleague can edit. ## The short answer A skill file is a markdown playbook that tells an AI worker how to do one kind of work: when the method applies, the steps to follow, and the shape of the output. Polaris stores each as a row with a name, a description and a markdown body, mirroring the SKILL.md format. Every skill attached to a worker is pasted into the system prompt in full at the start of each job. - **Format:** Name, description, markdown body - **Attached via:** A list of skill ids on the worker - **Loaded:** In full, at the start of every job ## What a skill file is for Instructions describe who a worker is. A skill describes how a particular job gets done. Splitting them means one worker can carry several methods, and one method can be shared by several workers without being retyped. Polaris keeps skills in an org-wide library. A worker row holds a list of skill ids, and at job time the runtime fetches those rows and pastes each one into the system prompt under a heading with its name and description, followed by its complete body. There is no summarisation step and no retrieval ranking: an attached skill is present in full, every time. ## How a skill reaches the model 1. **Someone writes it** — A name, a one-line description of when it applies, and a markdown body. The shape that works is a when-to-use section, a numbered method, and an output section describing the deliverable. 2. **It joins the org library** — Skills are org-scoped rows readable and writable by members, so a method one person refines is available to every worker in the organisation. 3. **It gets attached to a worker** — The worker's row carries an array of skill ids. Attaching and detaching is an edit to that array, not a rewrite of the worker. 4. **A job compiles it in** — At claim time the runtime reads every attached skill and renders them under a Your skills heading in the system prompt, above the operating rules for the session. 5. **The worker follows it** — The method in the file shapes the session directly, because it is sitting in the same prompt as the task brief and the acceptance criteria. ## The starter library, as shipped Six skills are seeded into every organisation. These are the real names and the real intent. - **web-research** — Decompose a question into three to five queries first, prefer primary sources, cross-check load-bearing numbers across at least two, stop when new sources stop changing the answer, and end with a sources list that only cites what was opened. - **competitive-analysis** — Build and confirm the competitor set before going deep, then capture offering, pricing model, positioning, one strength and one exploitable weakness for each, normalised into one table. Facts are separated from the analyst's read. - **summarization** — Identify the reader and the decision before writing, extract claims and numbers exactly rather than paraphrasing figures, and cut anything nobody would act on. - **copywriting** — Restate the audience and the single action, draft three angles with a marked recommendation, keep sentences short and concrete, and cut filler. - **content-strategy** — Anchor on the business goal, propose three to five pillars with the audience question each answers, map to a realistic cadence, and flag what should not be published. - **social-posts** — Write three hook options first, respect platform norms without being generic, one idea per post, and hand back the rejected hooks labelled so a human can swap them. ## Hidden prompt versus readable playbook **A prompt in someone's config** - Lives on one machine, in one person's setup - Improvements are invisible to everyone else - Nobody can review why the output has this shape - Leaves with the person who wrote it **A skill in the library** - A row any member can open and edit - One edit changes every worker carrying it - The method behind an output is inspectable - Survives the person who wrote it > **Where these come from** > > The format mirrors the SKILL.md convention used across the Claude skills ecosystem: frontmatter-style name and description, markdown instructions in the body. Polaris stores them as database rows rather than files so a team can edit them in the app, and the architecture notes name reuse of that ecosystem as the reason the runtime was built this way. ## Related reading - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) - [cloud-claude-code/give-an-ai-agent-tool-access](https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access) - [glossary/skill-file](https://www.polarishq.co/glossary/skill-file) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) ## Questions people ask **How is a skill different from the worker's instructions?** Instructions define the worker as a whole: its name, mission and standing rules, such as never marking a task done. A skill defines a method for one kind of work and can be attached to several workers. Both end up in the same system prompt, with instructions first and skills underneath. **Can I edit a skill file?** Yes. Skills are org-scoped rows that any member of the organisation can read and edit, and the change applies to every worker carrying that skill on its next job. There is no hidden or vendor-locked copy of the text. **Do skills consume context on every job?** Yes, because attached skills are pasted in full into the system prompt at the start of each session. That is a deliberate trade: a worker carrying three focused playbooks behaves more predictably than one carrying ten sprawling ones, so attach the methods that job actually needs. **Can a worker gain a skill during a session?** No. The skill set is compiled once, when the job is claimed. Adding a skill applies from the next session onward, which keeps a running session's behaviour reproducible from its inputs. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/ai-workers/research-analyst --- --- title: "Giving an AI agent access to your tools, safely" description: "Fourteen connections, one credential per org, each verified live against the provider before storage, with no read path back to any browser." url: https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access section: Cloud agents updated: 2026-08-21 --- # How an AI worker gets tool access The interesting part of tool access is not the list. It is where the credential lives and who can read it. ## The short answer Giving an AI agent tool access means storing a credential the agent's server-side runtime can use and the client never sees. Polaris ships a fixed catalog of fourteen connections, validates each credential live against the provider's own API before storing it, and keeps it in a table with no select policy, so no browser can read it back. Web search needs no credential and runs in every session. - **Connections in the catalog:** 14 - **Credential scope:** One per organisation - **Client read access:** None ## The catalog, in full Fixed list. Nothing outside it is connectable today. - **Communication** — Slack for channels, messages and task signals. WhatsApp for chats and customer messages. Gmail for reading and sending mail. - **Knowledge and design** — Notion for pages and databases. Google Drive for documents, sheets and files. Figma for design files and comments. - **Engineering** — GitHub for repositories, issues and pull requests. Linear for engineering issues. Supabase for a product database and its auth. - **Commercial** — Stripe for payments and invoices. HubSpot for CRM contacts and deals. Instagram for posts and engagement data. - **Time and the open web** — Google Calendar for events and scheduling. Web search for open research, which is the one entry needing no credential at all. ## What connecting actually does 1. **An owner or admin starts it** — Only members with the owner or admin role can write a connection credential. Row-level security enforces that on insert, update and delete. 2. **The credential is checked against the provider** — A server-side function calls the provider's own API with the token before anything is stored. Slack is checked with an auth test, GitHub by fetching the user, Notion by reading the integration, Linear with a viewer query, Stripe by fetching the account. 3. **A failure is reported honestly** — If the provider rejects the token the connection does not become connected. The error names the provider and what it said, rather than optimistically saving and failing later inside a job. 4. **Storage has no way out** — Verified credentials go into a table with insert, update and delete policies for admins and no select policy for anyone. The client API cannot return the secret, including to the person who pasted it. 5. **Status without exposure** — Members can call a function that lists which connections hold a live credential. It returns names, never secrets, which is how the interface shows a connection as connected without ever handling the token. ## What runs inside the session today Being precise about this matters more than sounding capable. Inside a task session the runtime exposes live web search, capped at eight uses, plus five tools it implements itself: post a progress comment, tick one acceptance-criteria item, attach a file, draft a document into the shared tree, and deliver. A review session swaps in tools for updating a file or document in place and resolving individual comments, with web search capped at five. Connections are the credential layer underneath that. Each is authorised once, org-wide, and stored where only server-side code can reach it, and each worker carries its own list of the connections it is meant to use. If a page tells you a worker is already posting to your Slack channels unattended, check it against the runtime rather than the marketing. > **The question worth asking any vendor** > > Ask whether a token you paste can ever be read back through their client API. In Polaris the answer is no by construction, because the table holding credentials has no select policy at all. That is a schema-level fact you can verify in the migration, not a promise in a security page. ## How each credential is obtained Where the human goes to get the token, per provider type. | Connection | Credential type | Where it comes from | | --- | --- | --- | | Slack | Bot token | Your Slack app, OAuth and Permissions | | GitHub | Fine-grained personal access token | GitHub developer settings, scoped to the repos needed | | Notion | Internal integration secret | Notion integrations, with pages shared to it | | Linear | Personal API key | Linear API settings | | Stripe | Restricted key | Stripe dashboard, read-only scopes recommended | | Figma | Personal access token | Figma account settings | | HubSpot | Private app token | HubSpot settings, integrations | | Gmail, Drive, Calendar | OAuth | Google, through the authorisation flow | | Web search | None | Available in every session | ## Related reading - [integrations/web-search](https://www.polarishq.co/integrations/web-search) - [integrations/slack](https://www.polarishq.co/integrations/slack) - [integrations/github](https://www.polarishq.co/integrations/github) - [cloud-claude-code/skill-files-for-ai-workers](https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers) - [glossary/model-context-protocol](https://www.polarishq.co/glossary/model-context-protocol) ## Questions people ask **Can I connect a tool that is not in the catalog?** No. The catalog of fourteen is fixed in the product today and there is no custom connection form. If the tool you need is missing, the honest answer is that Polaris cannot reach it yet. **Is the credential per worker or per organisation?** The credential is stored once per organisation and per connection, so one Slack token serves every worker in the org. Each worker separately carries the list of connections it is intended to use, which is a configuration of scope rather than a second credential. **What stops a worker using a connection it should not?** Two things: the connection list on the worker, and the scope of the credential itself. The stronger control is the second one. A read-only restricted key cannot be talked into writing, whatever a prompt says, which is why the catalog hints recommend restricted scopes where the provider offers them. **Does the agent need my Anthropic key too?** An organisation can store its own Anthropic key, in a separate table with the same write-only trust model and no select policy. The runtime looks up the org key when it claims a job and falls back to the key configured on the machine. **How do I revoke access?** Delete the connection credential, which owners and admins can do, and revoke the token at the provider. Because the credential is stored once per organisation rather than copied per worker, there is one place to remove it. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/github - https://www.polarishq.co/glossary/model-context-protocol --- --- title: "Long-running AI agent tasks: limits and failure modes" description: "Ten rounds, an eight-minute deadline, two retries and a salvage path. The real bounds on a Polaris agent session and how to structure work that exceeds them." url: https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks section: Cloud agents updated: 2026-08-21 --- # How long an agent session can run, and what happens when it ends Every agent runtime has bounds. The useful thing a vendor can do is tell you what they are. ## The short answer A Polaris agent session is bounded by three limits: a maximum of ten model rounds, a hard eight-minute deadline checked before each round, and eight web searches. A failed job returns to the queue while it has been attempted fewer than twice, then parks with its error message stored. Work that exceeds one session is split across tasks, or continued by commenting, which queues a fresh session. - **Rounds per session:** 10 - **Session deadline:** 8 minutes - **Attempts before parking:** 2 ## Why the bounds exist An unbounded agent loop has two failure modes and both are expensive. It can spin, repeating variations of the same call while a bill accrues. Or it can wander far enough from the brief that the output needs more review than doing the work would have. Short sessions with a hard stop make both survivable. The trade is real and worth stating. You cannot hand a Polaris worker an eight-hour job and walk away. You hand it a job a careful person could finish in a sitting, and you get a checkpoint at the end of each one. ## Every limit in the session Read from the runtime, not estimated. | Limit | Value | What happens at the edge | | --- | --- | --- | | Model rounds | 10 per session | The loop exits and the salvage path runs | | Wall clock | 8 minutes, checked before each round | The job throws a timeout and is retried or parked | | Web searches | 8 per task session, 5 per review session | Further searches are unavailable to the model | | Retries | Up to 2 attempts per job | The third failure sets the job to error with its message | | Poll interval | 5 seconds | Sets how quickly a free machine picks up the next job | | Live jobs per task | 1, enforced by a unique index | A second assignment is ignored rather than duplicated | | Attached file size | 2MB of text per file | The tool returns an error the model can react to | | Document length | 120 blocks, 4000 characters each | Extra blocks are dropped rather than failing the job | ## What happens when a session runs out The end of a session is engineered, not left to chance. 1. **The nudge** — If a round produces no tool call and nothing has been delivered, the runtime injects one message telling the worker to finish now by calling deliver. This happens exactly once per session. 2. **The salvage** — If the loop still ends without a delivery, the runtime walks the message history backwards, takes the most recent block of model text, and posts it as the delivery comment. Work never evaporates because the loop ran out. 3. **The failure path** — An exception at any point writes an error event to the session stream, returns the worker to idle, and sets the job back to pending for another attempt if it has been tried fewer than twice. 4. **The park** — After the second failed attempt the job is set to error with the message truncated onto the row. It stops retrying, and the task keeps everything the failed sessions had already committed. 5. **The continuation** — Commenting on the task queues a fresh session in which the latest human comment is the brief. That is the supported way to carry work across sessions. ## How to structure work that is genuinely long The decomposition is the skill. These four patterns cover most of it. - **One deliverable per task** — A competitor scan, a pricing table and a recommendation memo are three tasks. Each fits a session, each gets its own acceptance criteria, and each gives you a place to intervene. - **Chain by comment** — Deliver the research, then comment asking for the memo built on it. The follow-up session receives the previous comments as part of its brief, so context carries without being retyped. - **Let the document accumulate** — Written work is drafted into the shared Docs tree, and a review session updates a page in place while snapshotting the previous version. A long document can grow across sessions with its history intact. - **Use the checklist as a progress marker** — Ticked items persist between sessions. A follow-up session sees which criteria are already satisfied and is instructed not to redo finished work unasked. > **What this runtime is not built for** > > Multi-hour autonomous execution, background monitoring and scheduled runs are not what this is. There is no cron, no long-lived agent process per worker, and no session that outlives eight minutes. If your work needs an agent running for hours unattended, the honest answer is that this runtime is the wrong shape for it today. ## Related reading - [cloud-claude-code/claude-code-overnight-tasks](https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks) - [cloud-claude-code/cloud-agents-vs-local-agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) ## Questions people ask **What counts as a round?** One request to the model and the tool results returned to it. A round where the model calls three tools still counts as one. Server-side web search can pause a turn mid-round, and the runtime continues that turn rather than spending another round on it. **Does a retry start from scratch?** Yes. A retried job builds its context again from the current state of the task, which now includes any progress comments, ticked items, documents and files the failed attempt already committed. The worker is instructed not to redo work that is visibly finished. **Can several jobs run at the same time?** The runtime claims one job per loop cycle and works it to completion before claiming another, so a single machine processes the queue sequentially. The claim is written as a conditional update, so additional machines could be added against the same queue without two of them taking the same row. **What if the model keeps calling tools and never delivers?** The round cap ends the loop, the single nudge asks for a delivery first, and the salvage path posts the last text as the delivery if the nudge did not work. The eight-minute deadline is checked before each round and stops the session regardless. **Am I billed for a session that hit its limit?** Only completed jobs record human-equivalent minutes, and a job that ends in error records none. A session that hit the round cap but still delivered through the salvage path counts as delivered, with its effort counters reflecting what it actually produced. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/autonomous-agent --- --- title: "AI agents that produce real files, not just chat" description: "A Polaris worker decides whether work deserves a file, names it, picks the format and uploads it to the task. PDFs are typeset server-side, with versions kept." url: https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files section: Cloud agents updated: 2026-08-21 --- # When the deliverable is a file A chat reply is not a deliverable if the thing you needed was a document somebody can open. ## The short answer AI agents that produce files write actual artifacts rather than pasting text into a reply. A Polaris worker chooses whether work deserves a file, names it, picks the format from PDF, Markdown, CSV, JSON, HTML or plain text, and uploads it to the task like a human would. PDFs are typeset server-side into a real binary. Written deliverables default instead to pages in the shared document tree. - **File formats:** pdf, md, csv, json, html, txt - **Size limit:** 2MB of text per file - **Default for prose:** A page in Docs ## Two destinations, chosen on purpose Written deliverables go to Docs by default. Briefs, reports, plans and research become pages in the org's document tree, with real block structure: headings, paragraphs, bullets, to-dos, quotes, callouts, code and dividers. The worker picks a parent page whose topic fits, or the page files under Unsorted, and the delivery comment references it so a reader can follow the link. Files are for when a file format is the point. Tabular data as CSV, structured output as JSON, a board-ready document as PDF. The runtime instructs workers that files complement the document rather than replacing it, which is why a research session usually returns a page and a session producing a dataset usually returns a file. ## What happens when a worker attaches a file 1. **The worker decides and names it** — Format and filename are the worker's call, made from what the work needs. The name is sanitised to safe characters before anything else happens. 2. **PDFs get typeset** — A PDF is not text renamed. Markdown-ish content is laid out server-side with a coloured header band carrying the organisation name, real heading and bullet styling, monospaced table rows, and a footer line naming the worker, the product and the date. 3. **It goes to storage** — Bytes are uploaded under a path scoped by organisation, then task, then a timestamp and the filename, with the correct content type for the extension. 4. **A row makes it real** — An attachment record stores the name, path, byte size, MIME type, version number and the worker's member id as the uploader. It is the same record a human upload creates. 5. **The feed notices** — A trigger writes an attached event to the activity feed naming the file and its uploader, so a file appearing on a task is visible to the team without anyone announcing it. 6. **The delivery describes it** — When files were produced, the delivery comment is a tight summary of them rather than a repetition of their contents. ## What the runtime enforces The guardrails around file production, stated exactly. - **Empty files are rejected** — A file with no content returns an error to the model rather than creating an empty attachment. - **Two megabytes of text per file** — Above that the tool returns an error the worker can react to, usually by splitting the output. - **A failed PDF render fails loudly** — If typesetting throws, the tool returns the render error rather than silently uploading raw text with a pdf extension. - **Documents are capped** — A drafted page takes up to a hundred and twenty blocks, each up to four thousand characters, with block kinds validated against a known list and anything unrecognised falling back to a paragraph. - **Versions are preserved** — Every attachment carries a version number, and a revision stores the previous path, size and author before the new bytes are written. > **The detail that gives it away** > > The footer on a generated PDF reads that it was prepared by the named worker, marked as an AI worker, with the date. A deliverable that a human might forward to a client says on its face what produced it. That was a choice, and it is the sort of choice worth checking for in any tool that generates documents on your behalf. ## Related reading - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [cloud-claude-code/ai-agent-audit-trail](https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail) - [cloud-claude-code/share-claude-code-with-teammates](https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [ai-workers/technical-writer](https://www.polarishq.co/ai-workers/technical-writer) ## Questions people ask **Does the agent decide whether to produce a file, or do I?** The worker decides, from the shape of the work and its instructions. You influence it by naming the artifact you expect in the task title or description, which is more reliable than hoping. Written deliverables default to a document page unless a file format is genuinely the point. **Are the PDFs real PDFs?** Yes. They are generated as binary PDFs server-side with a typeset layout, not text files with a changed extension. Headings, bullets and table-ish rows each get their own styling, and the document carries a header band and a footer attributing it to the worker. **Can a worker revise a file it already delivered?** Yes, through a review session. Comment on the file, send the open comments to the worker, and it writes a complete new version while the previous one is preserved with its version number, path, size and author intact. **Where do the files actually live?** In your project's own storage bucket, under a path scoped by organisation and task, with a database row recording name, size, type, version and uploader. They are ordinary attachments, reachable the same way a human upload is. **Can it produce a spreadsheet or a slide deck?** CSV and JSON cover structured data, and PDF covers a document meant to be read as laid out. Native spreadsheet and presentation formats are not produced by the runtime today. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates - https://www.polarishq.co/cloud-claude-code/claude-code-overnight-tasks - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/ai-workers/technical-writer --- --- title: "Reviewing and approving AI agent work before it counts" description: "The machine never marks its own work done. How review works in Polaris: comment on the file, send the batch, get a new version with each comment resolved." url: https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work section: Cloud agents updated: 2026-08-21 --- # Agents deliver, humans close One rule holds the whole product together, and it is a rule about who is allowed to say finished. ## The short answer Reviewing AI agent work means a person, not the machine, decides whether it is finished. In Polaris a worker sets the task to in progress and delivers a comment; it can never set a task to done. A reviewer reads the delivery, comments on the file or document, sends the open comments back as a review job, and closes the task with a rating once satisfied. - **Who marks a task done:** A human, always - **Review job kind:** review - **Rating:** One per task, per reviewer ## The rule, and why it is structural Workers are told in their operating rules that they never mark a task done and hand it back to the person who owns the outcome. That is not only a prompt. The runtime sets task status to in progress at claim time and touches it again only to deliver, so there is no code path by which a session marks itself complete. The reason is that delivery and acceptance are different judgements. A worker can tell whether it produced something. Only the person who needed it can tell whether it was the right thing. Collapsing the two is how teams end up with a board full of closed tasks and no confidence in any of them. ## The review loop A batch loop, deliberately. You finish commenting before anything wakes up. 1. **Read the delivery in place** — The deliverable is a comment on the task, next to the ticked acceptance criteria, any attached files and any drafted document pages. An unticked checklist item is the first thing to look at. 2. **Comment on the artifact, not around it** — Comments attach to a file or a document page, and can be anchored to a quoted passage, a specific block or a table cell, so feedback points at the thing it is about. 3. **Send the whole batch** — When you have finished commenting, send the open comments to a worker. That creates a review job, and a unique index permits only one live review per file or page, so the same batch cannot be started twice. 4. **The worker works all of them in one pass** — It is instructed to address every comment it can, apply the changes, and resolve each comment it handled. Anything it genuinely cannot address stays open and is explained in the summary. 5. **A new version is written** — A file becomes a new version with the previous path, size and author preserved. A document is updated in place after a full snapshot of the old title and blocks is stored. 6. **A human closes and rates it** — Marking the task done is a human action recorded on the activity feed, and a thumbs up or down is stored once per task per reviewer against the worker's member row. ## Two ways to send work back Pick by whether the problem is in the artifact or in the brief. **A task comment** - Best when the brief itself needs to change - Queues a fresh task session for the owner - The latest human comment becomes the brief - The worker is told not to redo finished work unasked - Naming another worker pulls them in too **A review session** - Best when the artifact is close but wrong in places - Comments anchor to a passage, a block or a cell - Every open comment is worked in a single pass - Each addressed comment is resolved by the worker - The previous version is snapshotted first ## What a reviewer should check In this order, because it gets faster with practice. - **The unticked boxes** — Acceptance criteria the worker did not tick are its own admission that something is unfinished. Start there rather than reading the prose first. - **The sources** — Research deliveries end with plain URLs. The skill that governs research says never cite a source you did not open, and the session's search events show what was actually queried. - **The numbers** — Figures are the thing to verify by hand. A summarisation method that says extract numbers exactly is a rule the reviewer should hold it to. - **The hours line** — The job stores its human-minute estimate beside the counters that produced it. If the search count on the bill does not match the search events in the stream, that is a specific, answerable question. > **Why the loop is batched** > > A comment on a task wakes the worker immediately, which is right for a conversation and wrong for a document review, where you want to read the whole thing before anything responds. Review sessions are explicitly sent, so you can leave fifteen comments across a report and only then hand the batch over. ## Related reading - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) - [cloud-claude-code/ai-agents-that-produce-files](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) - [cloud-claude-code/ai-agent-audit-trail](https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail) - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) ## Questions people ask **Can an AI worker ever close a task?** No. The runtime sets a task to in progress when it claims a job and never sets it to done, and workers are instructed in their operating rules to hand the outcome back to the person who owns it. Closing a task is a human action, recorded on the activity feed as a completed event with the actor named. **What happens to comments the worker could not address?** They stay open. The review rules tell the worker to leave a comment unresolved only when it genuinely cannot act on it, and to explain why in the delivery summary. An unresolved comment with a stated reason is a better outcome than a resolved one with a guess behind it. **Can I get the previous version of a file back?** The previous version's path, size, version number and author are stored before a new one is written, and document updates store a full snapshot of the old title and blocks. The history exists in your own database, though a one-click restore is not in the product today. **What does a rating do?** A thumbs up or down is recorded once per task per reviewer against the worker's member row, which builds a per-worker quality record rather than a per-session one. It is a signal for the humans deciding what to assign next. **Does asking for changes cost more?** A follow-up or review session is a job like any other and records its own human-equivalent minutes, computed largely from comments resolved and content rewritten. A short round of corrections produces a correspondingly short line, and a session that fails records nothing. ## Related - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/delivery-comment --- --- title: "Polaris Integrations: 14 Tools an AI Worker Can Use" description: "The fourteen connections in the Polaris catalog, what each one gives an AI worker, how a credential is verified and stored, and who is allowed to authorize it." url: https://www.polarishq.co/integrations section: Integrations updated: 2026-08-21 --- # Every tool a Polaris worker can be given One catalog, one credential per tool per organisation, authorized by an owner and used by every worker who carries it. ## The short answer Polaris connections are the tool access you give an AI worker. The catalog holds fourteen entries: Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and web search. Thirteen need a credential, which an owner or admin authorizes once for the whole organisation. Nine are checked live against the provider before anything is stored. Web search needs nothing. - **Connections in the catalog:** 14 - **Need a credential:** 13 of 14 - **Verified live before storing:** 9 - **Who can authorize:** Owners and admins ## What a connection actually is A worker in Polaris is a member row, a set of instructions, a few skill files and a list of connections. The connection list is the part that decides what the worker can reach outside the workspace. Everything else is what it knows; connections are what it can touch. The list is fixed. Fourteen tools, chosen because they are where small teams already keep their work, and no free-text field where somebody can paste an arbitrary endpoint. When the copilot interviews you about a new hire, the tool question offers names from this catalog and nothing else. Two consequences follow. A worker cannot acquire access you did not grant, and the set of credentials your organisation holds is short enough to read in one screen. ## The catalog Auth mode and credential type come straight from the connection catalog in the product. Live check means Polaris calls the provider with the credential and refuses to store it if the provider says no. | Connection | Credential | Live check | What it covers | | --- | --- | --- | --- | | Slack | One-click OAuth, or a bot token | Yes | Channels the bot was invited to, and posting | | Notion | Internal integration secret | Yes | Pages shared with the integration | | Linear | Personal API key | Yes | Issues visible to the key's owner | | GitHub | Fine-grained access token | Yes | Repos, issues and pull requests in the token's scope | | Stripe | Restricted key | Yes | Payments and invoices at the key's scopes | | Figma | Personal access token | Yes | Design files and comments | | HubSpot | Private-app token | Yes | CRM contacts and deals | | WhatsApp | Meta system-user token | Yes | Chats on the WhatsApp Business app | | Instagram | Long-lived page token | Yes | Posts and engagement data | | Supabase | Service-role or restricted key | No live check yet | A project database and its auth | | Gmail | Google OAuth | Flow not live | Reading and drafting email | | Google Drive | Google OAuth | Flow not live | Docs, sheets and files | | Google Calendar | Google OAuth | Flow not live | Events and scheduling | | Web search | None | Not applicable | Open web research inside every session | ## How authorizing works The same four steps for every credentialled tool in the catalog. 1. **Open the connection** — Click a pending connection on a worker card, or open it from Configure. Members see the panel; only owners and admins get the input, and everyone else is told to ask one of them. 2. **Hand over the credential** — Paste the token, or for Slack click through the provider's own consent screen and copy nothing at all. The field is a password field and the value never lands in a form log. 3. **Polaris calls the provider** — For the nine connections with a live check, Polaris makes a real API call with the credential. A rejected token produces the provider's own error text and nothing is written. 4. **Stored once, org-wide** — The verified credential is stored server-side against your organisation, and every worker in the org carrying that connection flips from pending to connected in the same moment. Revoking deletes it and flips them all back. ## What the credential can and cannot reach **Held server-side** - One row per organisation per tool, written by the edge function that verified it - Read by the server when a worker uses the tool, never sent to the browser - Replaceable by pasting a new token over the old one - Deletable in one click, org-wide, by an owner or admin **Bounded by the provider, not by us** - A worker sees exactly what the token sees, no more and no less - Scoping happens where you create the token, in the provider's own console - A read-only key stays read-only inside Polaris - Revoking at the provider kills the access even if the row still exists > **What we do not claim** > > Polaris is in free public beta. There is no SOC 2 report, no certification, no data-residency guarantee and no compliance programme to point at, and inventing one would be worse than the gap itself. What exists is this: credentials are verified against the provider before storage, held server-side, scoped by the token you chose, and removable org-wide in one click. ## Every connection page - [Connect Slack to Polaris](https://www.polarishq.co/integrations/slack) — Six scopes, no access to direct messages, and a bot that only reads the channels somebody invited it into. - [Connect Notion to Polaris](https://www.polarishq.co/integrations/notion) — Notion decides what Polaris can see, page by page, because an internal integration only reaches what you explicitly share with it. - [Connect Linear to Polaris](https://www.polarishq.co/integrations/linear) — A Linear API key carries one person's visibility, so the account you make it on decides what every worker can see. - [Connect GitHub to Polaris](https://www.polarishq.co/integrations/github) — Fine-grained tokens let you hand over three repositories instead of an account, which is the whole reason to use them here. - [Connect Gmail to Polaris](https://www.polarishq.co/integrations/gmail) — The honest version: Gmail is in the catalog, the Google sign-in flow has not shipped, and the product says pending rather than pretending. - [Connect Google Calendar to Polaris](https://www.polarishq.co/integrations/google-calendar) — Scheduling is the one thing an assistant is asked for first, and it is the one connection still waiting on Google sign-in. - [Connect Google Drive to Polaris](https://www.polarishq.co/integrations/google-drive) — Files a worker produces already arrive as attachments on the task. Files your team already keeps in Drive are the part still waiting. - [Connect Figma to Polaris](https://www.polarishq.co/integrations/figma) — A Figma token belongs to a person, and Polaris shows you which person it verified, which is the detail worth checking before you store it. - [Connect HubSpot to Polaris](https://www.polarishq.co/integrations/hubspot) — Private-app tokens are scoped where you create them, which makes HubSpot one of the easier connections to grant narrowly. - [Connect Stripe to Polaris](https://www.polarishq.co/integrations/stripe) — This is the one connection where the key you choose matters more than anything on this page. - [Connect Supabase to Polaris](https://www.polarishq.co/integrations/supabase) — Supabase is the connection where a careless key choice does the most damage, and the only one Polaris cannot yet check for you. - [Connect WhatsApp to Polaris](https://www.polarishq.co/integrations/whatsapp) — For a lot of companies WhatsApp is the support desk, the sales channel and the customer record, and none of it is written down anywhere else. - [Connect Instagram to Polaris](https://www.polarishq.co/integrations/instagram) — Engagement data is the only honest input to a content calendar, and it is the input most calendars are written without. - [Web search in Polaris](https://www.polarishq.co/integrations/web-search) — Nothing to authorize, nothing to store, and every query a worker runs appears in the activity feed while it works. ## Related reading - [ai-workers](https://www.polarishq.co/ai-workers) - [cloud-claude-code/give-an-ai-agent-tool-access](https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [glossary/skill-file](https://www.polarishq.co/glossary/skill-file) - [use-cases](https://www.polarishq.co/use-cases) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) ## Questions people ask **Can I connect a tool that is not in the Polaris catalog?** Not today. The catalog is a fixed list of fourteen and the hiring flow only offers names from it. When somebody asks for something outside the list, the copilot can search the MCP market to see what exists, but nothing outside the catalog can be authorized in the product. **Who is allowed to authorize a connection?** Only the owner or an admin of the organisation. The check happens on the server, not in the interface, so a member who opens the panel is told to ask an owner rather than being shown a disabled input they could work around. **Do I have to connect the same tool again for every new worker?** No. A credential belongs to the organisation, not to a worker. Authorize Slack once and every worker that carries the Slack connection, including ones hired months later, is connected the moment they are created. **Can anyone read the token back out of Polaris?** The browser cannot. Credentials are written by an edge function into a table the client never selects from, and the interface only ever shows the label the provider returned, such as the workspace name or the account login. Reading it back would mean going through the server, which no product surface does. **What happens to a worker when I revoke a connection?** The stored credential is deleted and every worker in the organisation carrying that connection returns to pending in the same action. The worker keeps its instructions and skill files, so re-authorizing later puts it straight back to work. **Which connections are not fully live yet?** The three Google connections, Gmail, Drive and Calendar, are OAuth and that flow has not shipped, so they stay pending and store nothing. Supabase accepts and stores a key but has no live verification call yet, and the product labels it as unvalidated rather than pretending otherwise. ## Related - https://www.polarishq.co/ai-workers - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/integrations/web-search - https://www.polarishq.co/integrations/notion - https://www.polarishq.co/glossary/ai-worker --- --- title: "Polaris Slack Integration: Messages Into Tasks" description: "What connecting Slack gives an AI worker: the channels it was invited to, posting as Polaris, and threads arriving in the Inbox as prefilled task suggestions." url: https://www.polarishq.co/integrations/slack section: Integrations updated: 2026-08-21 --- # Connect Slack to Polaris Six scopes, no access to direct messages, and a bot that only reads the channels somebody invited it into. ## The short answer Connecting Slack gives a Polaris worker the channels the bot has been invited to. The copilot can list channels, read their history and post messages, so a thread becomes a prefilled task suggestion in the Inbox with bucket, lane, labels and owner already set, waiting for one click. Slack is authorized once per organisation by an owner and stored server-side. - **Authorization:** One click, or a bot token - **Scopes requested:** 6 - **Direct messages:** Out of scope - **Verified live:** Against Slack before storing ## What Slack gives a worker Most teams already run their day in Slack, which means most teams already lose work in Slack. A decision gets made in a thread on Thursday, three people read it, nobody writes it down, and on Monday it does not exist. With Slack connected, the copilot can list the channels the bot is in, read their recent messages and post back. That is the raw material for the Inbox: a message that looks like work becomes a task suggestion with a bucket, a lane, labels and an owner already filled in. You approve it or you do not. Signals become suggestions, never silent tasks, because a workspace that creates tasks behind your back stops being trusted within a week. Posting matters as much as reading. A worker that can write to a channel can put its delivery where the team already looks, instead of leaving it in a tool somebody has to remember to open. ## Connecting Slack 1. **Click connect directly** — The button opens Slack's own consent screen with the six scopes listed. Approve there and you land back in Polaris connected, with no token to copy anywhere. 2. **Or paste a bot token instead** — The advanced path takes a bot token beginning xoxb- from your own Slack app, under OAuth and Permissions. Same result, more steps, useful when your workspace restricts which apps can be installed. 3. **Polaris checks it against Slack** — Either path ends with a call to Slack that confirms the token works and returns the workspace name, which is what the interface then shows you. A bad token is refused with Slack's own error. 4. **Invite the bot where you want it read** — Slack only exposes channels the bot is a member of. Run the invite command for Polaris in each channel that matters. This is the step people forget, and the symptom is a worker that says it can see nothing. ## The six scopes, and what each one buys This is the exact scope list the Polaris Slack app requests. Nothing outside it is available to a worker. | Scope | What it allows | | --- | --- | | channels:read | See which public channels exist | | channels:history | Read messages in public channels the bot joined | | groups:read | See private channels the bot was invited to | | groups:history | Read messages in those private channels | | chat:write | Post messages as Polaris | | users:read | Resolve user IDs into names, so a suggestion can name an owner | > **The one-time setup nobody mentions** > > One-click connect needs a Slack app to exist for the deployment, with its client ID and secret configured. A published manifest in the repo creates that app with exactly the six scopes above, and until an owner has done it once the direct button returns a not-configured error rather than a consent screen. The bot-token path works regardless. ## What Slack cannot do here Worth reading before you connect, because it is shorter than the list of things people assume. - **It cannot read your DMs** — No direct-message scope is requested, so private conversations and group DMs are outside what any worker can reach, whatever anybody asks it to do. - **It cannot join channels by itself** — Access follows invitations. A channel nobody invited the bot into is invisible, including public ones it can see the name of. - **It cannot delete or edit your history** — The write scope covers posting. Nothing in the scope list allows removing or rewriting what your team already said. - **It cannot download files** — No file scope is requested, so attachments in a channel are not part of what comes back. ## Workers and work this connection feeds - [ai-workers/project-coordinator](https://www.polarishq.co/ai-workers/project-coordinator) - [ai-workers/support-specialist](https://www.polarishq.co/ai-workers/support-specialist) - [ai-workers/executive-assistant](https://www.polarishq.co/ai-workers/executive-assistant) - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) - [use-cases/executive/meeting-follow-ups](https://www.polarishq.co/use-cases/executive/meeting-follow-ups) - [use-cases/customer-support/escalation-tracking](https://www.polarishq.co/use-cases/customer-support/escalation-tracking) ## Questions people ask **Can a Polaris worker read my Slack direct messages?** No. The Slack app requests six scopes and none of them cover direct messages or group DMs. It reads public channels it has joined and private channels it was invited to, and that is the whole surface. **Do I need to create my own Slack app?** Only if the deployment has not had one configured, or if your workspace policy requires apps to be internally owned. The repository ships a Slack app manifest with the exact six scopes, so creating one is a paste rather than a form-filling exercise, and the resulting bot token goes in through the advanced path. **Why can my worker not see a channel I know exists?** Almost always because the bot is not in it. Slack exposes history only for channels the app has joined, so the fix is to invite Polaris in that channel. Private channels additionally need the invite to come from somebody already in them. **Does connecting Slack create tasks automatically?** No. Messages become suggestions in the Inbox with the fields prefilled, and a person approves them. That boundary is deliberate: a workspace that silently manufactures tasks from chat becomes noise faster than the chat it was meant to tame. **Where does the Slack token live once it is authorized?** Server-side, in a row belonging to your organisation, written by the function that verified it with Slack. The interface shows the workspace name that Slack returned. The browser never receives the token itself. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/alternatives/slack - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/glossary/ai-worker --- --- title: "Polaris Notion Integration: Import Pages, Give Access" description: "Connect Notion with an internal integration secret to give AI workers the pages you share, and import whole page trees into Polaris Docs with nesting kept." url: https://www.polarishq.co/integrations/notion section: Integrations updated: 2026-08-21 --- # Connect Notion to Polaris Notion decides what Polaris can see, page by page, because an internal integration only reaches what you explicitly share with it. ## The short answer Connecting Notion gives a Polaris worker the pages you share with your Notion integration, and nothing else. With the credential stored, Polaris can import whole page trees into Docs, where child pages become nested doc pages and blocks keep their structure. Notion authorizes with an internal integration secret that Polaris verifies against the Notion API before storing it. - **Credential:** Internal integration secret - **Access model:** Per-page sharing, chosen by you - **Verified live:** Against the Notion API ## The access model is the feature Notion internal integrations do not get a workspace. They get whatever pages a human has shared with them, one at a time, through the same connection menu people use to share with colleagues. That makes Notion the easiest connection in the catalog to reason about: if you did not share it, no worker can read it. It also makes the failure mode obvious. A worker that reports an empty Notion is almost never broken. Somebody created the integration and never shared a page with it. ## Connecting Notion 1. **Create an internal integration** — In Notion's integrations settings, create one for your workspace. The secret it hands back starts with ntn_ or secret_. 2. **Share the pages it should see** — Open each page or database you want reachable and share it with the integration by name. Sharing a parent covers its children, which is usually the move. 3. **Paste the secret into Polaris** — An owner or admin pastes it into the connect panel. Polaris calls the Notion users endpoint with it and stores it only if Notion answers, showing you the integration or workspace name it got back. ## Importing Notion into Polaris Docs Beyond giving a worker read access, the stored Notion credential drives a real importer that moves content into the Polaris Docs tree. - **Pick from what the integration can see** — The importer lists the pages the integration has been granted, so the picker is a direct reflection of what you shared and a decent audit of it. - **Page trees stay trees** — Child pages come across as nested doc pages rather than being flattened into one long document, so a wiki keeps the shape people navigate by. - **Blocks keep their kind** — Headings, bullets, numbered lists, to-dos, quotes, code, callouts, dividers and toggles map onto the equivalent Polaris block, and nesting becomes indentation. - **The result is a Polaris doc** — Once imported, pages are versioned, commentable and reviewable like anything else in Docs, and a worker can draft into them. ## Connect, or move Both are reasonable. The wrong answer is deciding by accident. **Keep Notion, connect it** - The team's wiki habits stay where they are - Workers read the prior art before they write anything - You keep Notion's database views, which Polaris does not have - Nothing to migrate, nothing to break **Import into Docs** - One place for docs and the tasks that reference them - Versioned files, comments and file review on the same pages - No per-seat bill for people who only read - Workers can draft directly into the tree instead of pasting ## Workers and work this connection feeds - [ai-workers/technical-writer](https://www.polarishq.co/ai-workers/technical-writer) - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [ai-workers/ops-coordinator](https://www.polarishq.co/ai-workers/ops-coordinator) - [use-cases/engineering/technical-documentation](https://www.polarishq.co/use-cases/engineering/technical-documentation) - [use-cases/operations/process-documentation](https://www.polarishq.co/use-cases/operations/process-documentation) - [use-cases/hr/policy-documentation](https://www.polarishq.co/use-cases/hr/policy-documentation) ## Questions people ask **Can a Polaris worker see my whole Notion workspace?** No, and Notion is the reason rather than Polaris. An internal integration reaches only the pages a human has shared with it. Grant one page and that is the extent of it; grant a top-level page and its children come with it. **What does the Notion import bring across?** Page trees, with child pages as nested doc pages and blocks mapped to their Polaris equivalents, including headings, bullets, to-dos, quotes, code, callouts and toggles. Nesting becomes indentation. Notion database views are not part of it. **Is the import a one-way copy or a live sync?** It is an import. Pages are read from Notion and written into Polaris Docs as real docs, after which the two are independent. Editing the Polaris copy does not touch Notion and vice versa. **Where does the integration secret go?** It is verified against the Notion API and then stored server-side against your organisation, one row for the whole org. Every worker carrying the Notion connection uses it, and revoking removes it for all of them at once. **Can a worker write back into Notion?** Drafting in Polaris is what the product does today: a worker creates versioned docs in the Docs tree and delivers them as comments on the task. Treat Notion as the source you connect and Docs as the place work lands. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/alternatives/notion - https://www.polarishq.co/replace/notion-and-slack - https://www.polarishq.co/cost/notion-pricing - https://www.polarishq.co/integrations/slack --- --- title: "Polaris Linear Integration for AI Workers" description: "Give an AI worker Linear issues with a personal API key, verified live before storage, and keep engineering in Linear while workers do the writing around it." url: https://www.polarishq.co/integrations/linear section: Integrations updated: 2026-08-21 --- # Connect Linear to Polaris A Linear API key carries one person's visibility, so the account you make it on decides what every worker can see. ## The short answer Connecting Linear gives a Polaris worker access to your Linear issues through a personal API key, authorized once for the organisation. Polaris verifies the key by asking Linear who it belongs to and refuses to store a key Linear rejects. The key inherits its owner's visibility, so create it on an account whose access matches what workers should read. - **Credential:** Personal API key - **Verified live:** Linear returns the key owner - **Visibility:** Whatever the key's owner can see ## Who the key is, matters more than what it can do Linear issues an API key against a person. When Polaris validates one, Linear answers with that person's name, which is exactly what the connect panel then displays back to you. That single detail should drive the decision: a key made on the CTO's account gives workers the CTO's view of every team. For most teams the right move is a key from an account scoped to the teams the workers actually support. Not because Polaris will misbehave, but because access you never granted is access you never have to reason about again. ## What workers do with Linear connected Linear stays the engineering tracker. The connection is about the work that surrounds an issue rather than replacing the board. - **Read the issue before writing about it** — A worker drafting release notes or a postmortem starts from what the issues actually say, instead of from what somebody remembered in a meeting. - **Turn a backlog into prose** — Cycle summaries, changelog drafts and status write-ups are the kind of work that always slips because it is nobody's favourite hour of the week. - **Cross-reference the tracker with the conversation** — With Slack connected as well, a worker can line up what was said in a channel against what exists as an issue, and file the gaps as suggestions in the Inbox. - **Carry context between tools** — The same worker can read a Linear issue and produce a customer-facing note in Docs, which is the handoff that usually costs a person half a day. ## Keep Linear, add workers Engineering teams that like Linear should keep Linear. The two sit beside each other. **Stays in Linear** - Cycles, estimates and the issue graph - Engineering triage and the keyboard-driven workflow - Git branch and pull-request linking - Whatever your team has already automated there **Moves to Polaris** - The writing nobody does: release notes, postmortems, docs - Work owned by people who do not live in an engineering tracker - AI workers assigned tasks the same way a person is - The bill, which stops being per seat > **Revoking is one click and it is org-wide** > > Deleting the stored key removes it for the whole organisation and returns every worker carrying the Linear connection to pending in the same action. If the key belonged to somebody who has left, revoke in Polaris and revoke in Linear. The second one is the one that actually stops the access. ## Workers and work this connection feeds - [ai-workers/qa-engineer](https://www.polarishq.co/ai-workers/qa-engineer) - [ai-workers/project-coordinator](https://www.polarishq.co/ai-workers/project-coordinator) - [ai-workers/technical-writer](https://www.polarishq.co/ai-workers/technical-writer) - [use-cases/product/release-notes](https://www.polarishq.co/use-cases/product/release-notes) - [use-cases/engineering/bug-triage](https://www.polarishq.co/use-cases/engineering/bug-triage) - [use-cases/engineering/sprint-planning](https://www.polarishq.co/use-cases/engineering/sprint-planning) ## Questions people ask **What kind of Linear credential does Polaris need?** A personal API key from Linear's API settings, the kind that starts with lin_api_. Polaris validates it by running a small query that asks Linear who the key belongs to, and shows you that name once it is stored. **Does connecting Linear sync my issues into Polaris tasks?** No. Polaris tasks and Linear issues stay separate. The connection is read access for workers, not a two-way mirror, which is deliberate: a half-working sync between two trackers is worse than either tracker alone. **Can I limit which Linear teams a worker sees?** Through Linear, yes. The key inherits the visibility of the account that created it, so a key made on a member with access to two teams gives workers those two teams. Polaris has no separate filter on top of that. **Do I still need Linear if I use Polaris?** That depends on whether cycles and the issue graph are load-bearing for your team. Plenty of teams keep both, connecting Linear for engineering and running everything else in Polaris. The alternatives page for Linear makes the case on both sides. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/qa-engineer - https://www.polarishq.co/use-cases/engineering/sprint-planning - https://www.polarishq.co/use-cases/product/release-notes - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/cost/linear-pricing - https://www.polarishq.co/integrations/github --- --- title: "Polaris GitHub Integration: Scoped Repo Access" description: "Give AI workers repository access with a fine-grained GitHub token scoped to the repos they need, verified live against GitHub before Polaris stores anything." url: https://www.polarishq.co/integrations/github section: Integrations updated: 2026-08-21 --- # Connect GitHub to Polaris Fine-grained tokens let you hand over three repositories instead of an account, which is the whole reason to use them here. ## The short answer Connecting GitHub gives a Polaris worker the repositories, issues and pull requests inside a fine-grained access token's scope. Polaris verifies the token against GitHub before storing it and shows the account login it came back with. Scope the token to the repositories workers need and choose read-only permissions where writing is not part of the job. - **Credential:** Fine-grained access token - **Verified live:** GitHub returns the account login - **Scope control:** Per repository, at GitHub ## Scope the token, not the trust GitHub's fine-grained tokens are the reason this connection is comfortable to grant. You pick the repositories, you pick the permissions, and the token cannot reach anything you left out. A worker that only writes documentation does not need write access to production infrastructure, and a token is where that decision gets enforced. Polaris checks the token by asking GitHub whose it is, then shows you that login in the connect panel. If the login surprises you, that is the check working. ## What the connection gives a worker Everything below is bounded by the token you created. Narrow the token and this list narrows with it. - **The repositories you named** — Not the organisation, not every repo the creator can see. A fine-grained token lists its repositories explicitly and that list is the worker's world. - **Issues and pull requests** — The raw material for triage write-ups, release notes and the summary of what changed this week that nobody has time to assemble. - **Whatever the permissions allow** — Read-only stays read-only inside Polaris. There is no privilege escalation path in the product, because the credential is used as given. - **Context to write from** — A technical writer working from real code and real pull-request discussion produces documentation that matches the repository rather than the intention. ## Scoping decisions worth making before you paste Polaris cannot make these for you. GitHub is where they are enforced. | Decision | Safe default | When to widen it | | --- | --- | --- | | Which repositories | The specific repos workers are briefed on | Never widen speculatively; add repos when a task needs one | | Read or write | Read-only | Only if you intend workers to open issues or pull requests | | Which account | A machine account or a member with matching access | Avoid keys made on an admin account entirely | | Expiry | The shortest expiry your workflow tolerates | Longer only where re-authorizing is genuinely disruptive | > **Two places to revoke** > > Revoking in Polaris deletes the stored token and puts every worker carrying the GitHub connection back to pending. It does not tell GitHub anything. A token that leaked has to be revoked at GitHub as well, and that is the revocation that actually ends the access. ## Workers and work this connection feeds - [ai-workers/technical-writer](https://www.polarishq.co/ai-workers/technical-writer) - [ai-workers/qa-engineer](https://www.polarishq.co/ai-workers/qa-engineer) - [ai-workers/project-coordinator](https://www.polarishq.co/ai-workers/project-coordinator) - [use-cases/engineering/code-review-workflow](https://www.polarishq.co/use-cases/engineering/code-review-workflow) - [use-cases/engineering/technical-documentation](https://www.polarishq.co/use-cases/engineering/technical-documentation) - [use-cases/engineering/incident-postmortems](https://www.polarishq.co/use-cases/engineering/incident-postmortems) ## Questions people ask **Does Polaris need admin access to my GitHub organisation?** No. A fine-grained personal access token scoped to specific repositories is what the catalog asks for, and it is the right answer. There is no GitHub App install, no organisation-wide permission request and no reason to hand over more than the repositories in play. **Can an AI worker push code or merge a pull request?** Only if the token you created allows it. Permissions are set at GitHub, and Polaris uses the credential exactly as issued. A read-only token stays read-only no matter what a worker is asked to do. **How does Polaris verify a GitHub token?** It calls GitHub's user endpoint with the token before storing anything. A rejected token produces GitHub's status code as an error and nothing is written. A valid one comes back with the account login, which the connect panel shows you. **Is GitHub a good connection for a non-engineering worker?** Sometimes. A technical writer or a project coordinator reading issues and pull requests is doing legitimate work with it. In those cases give read-only access to the specific repositories they write about, and nothing else. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/ai-workers/qa-engineer - https://www.polarishq.co/use-cases/engineering/code-review-workflow - https://www.polarishq.co/replace/linear-and-notion-and-github - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access - https://www.polarishq.co/integrations/linear --- --- title: "Polaris Gmail Integration: Status and What It Grants" description: "Gmail is an OAuth connection in the Polaris catalog covering reading and sending email. The Google sign-in flow is not live, so the connection stays pending." url: https://www.polarishq.co/integrations/gmail section: Integrations updated: 2026-08-21 --- # Connect Gmail to Polaris The honest version: Gmail is in the catalog, the Google sign-in flow has not shipped, and the product says pending rather than pretending. ## The short answer Gmail sits in the Polaris connection catalog as an OAuth connection, meaning it is authorized by signing in with Google rather than by pasting a token. That flow is not live yet, so Gmail stays pending on a worker and no credential is stored. Workers can be hired with Gmail on their list today, and the writing they do lands in Docs until the flow ships. - **Auth mode:** Google OAuth - **Status:** Pending, flow not shipped - **Credential stored:** None ## Why this page starts with a limitation Three connections in the Polaris catalog authorize through Google rather than through a pasted token: Gmail, Drive and Calendar. Google sign-in is a different mechanism from the token flow that powers the other ten, and it arrives with the worker execution runtime rather than ahead of it. Until then, the product tells you the truth in the connect panel: the connection is pending, and nothing is faked. No placeholder credential is stored, no worker claims to have read your mail, and the connection badge on the worker card stays honest about the state it is in. We would rather publish this page saying so than publish a page describing a mailbox integration you cannot have this afternoon. ## What the Gmail connection covers The catalog entry describes Gmail as reading and sending email. That is the intended surface of the connection, and it is roadmap rather than a shipped capability. - **Reading** — A support or sales worker briefed on a mailbox can work from what actually arrived rather than from a summary somebody typed into a task. - **Drafting and sending** — Reply drafts are the obvious first job, and the interesting question is where the human sits in the loop when the medium is email rather than a task comment. - **Authorization scope** — Google's own consent screen defines what an OAuth grant covers. When the flow ships, that screen is where the boundary is set and displayed. ## Email work you can run today, without the connection Two of the three reasons people ask for Gmail have answers already. **Works now** - A worker drafts the email as a versioned doc, you copy it out and send it - Follow-up tracking as real tasks with owners and due dates - Slack connected instead, where a lot of the same signal already lives - Delivery arrives as a comment you can edit before anything goes out **Waits for the flow** - A worker reading an inbox directly - Sending from your address - Triage that starts the moment a message lands - Threading a reply into an existing conversation > **What pending means on a worker card** > > A worker hired with Gmail on its connection list is created normally and works normally on everything else. The Gmail badge reads pending, and it stays pending until the flow exists. Connections that are authorized flip to connected across the whole organisation the moment a credential is stored, so nothing about the pending state is stuck or broken. ## Workers that will want this connection - [ai-workers/support-specialist](https://www.polarishq.co/ai-workers/support-specialist) - [ai-workers/sdr](https://www.polarishq.co/ai-workers/sdr) - [ai-workers/executive-assistant](https://www.polarishq.co/ai-workers/executive-assistant) - [use-cases/customer-support/response-templates](https://www.polarishq.co/use-cases/customer-support/response-templates) - [use-cases/sales/lead-research](https://www.polarishq.co/use-cases/sales/lead-research) - [use-cases/executive/meeting-follow-ups](https://www.polarishq.co/use-cases/executive/meeting-follow-ups) ## Questions people ask **Can a Polaris worker read my Gmail today?** No. Gmail authorizes through Google sign-in and that flow has not shipped, so the connection stays pending and no credential exists to read anything with. The connect panel says exactly this rather than showing a connected state that would not be true. **Why not accept a Google app password or an API key instead?** Because that would trade a proper consent screen for a long-lived credential with a scope nobody reviewed. The ten token connections in the catalog use tokens because their providers issue scoped ones. Google's answer is OAuth, so Gmail waits for OAuth. **Should I still hire a worker with Gmail on its list?** If email is genuinely part of the role, yes. The worker is created and does everything else it was hired for, and the Gmail badge sits at pending. When a credential is stored for a connection, every worker in the organisation carrying it flips to connected at once. **What is the closest thing that works now?** Drafting. A worker writes the email as a doc in Polaris, versioned and commentable, and delivers it as a comment on the task. You read it, edit it and send it yourself, which for outbound and support replies is where most teams want the human anyway. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/use-cases/customer-support/response-templates --- --- title: "Polaris Google Calendar Integration: Current Status" description: "Google Calendar is an OAuth connection in the Polaris catalog for events and scheduling. The Google sign-in flow is not live, so the connection stays pending." url: https://www.polarishq.co/integrations/google-calendar section: Integrations updated: 2026-08-21 --- # Connect Google Calendar to Polaris Scheduling is the one thing an assistant is asked for first, and it is the one connection still waiting on Google sign-in. ## The short answer Google Calendar is an OAuth connection in the Polaris catalog, covering events and scheduling. Authorization happens through Google sign-in, and that flow has not shipped, so the connection stays pending on a worker and stores no credential. Time-bound work in Polaris runs on task due dates and the Focus lane's three horizons rather than on calendar events. - **Auth mode:** Google OAuth - **Status:** Pending, flow not shipped - **Works today instead:** Due dates and Focus horizons ## Where time lives in Polaris right now Polaris already has a model of time, and it is not a calendar. Focus is the home screen, and it sorts the non-negotiables across three horizons: today, this week, and the next thirty days. Tasks carry due dates and owners. That is the spine deadlines hang from. A calendar answers a different question. It says when a specific hour is spoken for, which is what an executive assistant needs before it can propose anything. Until the Google flow ships, a worker cannot see that layer, and the connect panel says so. ## What the Calendar connection covers The catalog describes it as events and scheduling. Roadmap, not shipped. - **Reading availability** — Knowing which hours are already committed is the precondition for every scheduling task anybody would assign a worker. - **Events** — The meeting is where most follow-up work is generated, and the follow-ups are the part that reliably goes missing. - **Consent at Google** — When the flow ships, the boundary is whatever Google's consent screen grants, shown to the person authorizing it. ## Running scheduling work without the connection Less elegant, entirely functional. 1. **Keep the commitment as a task** — A meeting that matters becomes a task with a due date and an owner, which puts it in Focus where the team already looks. 2. **Let the worker prepare, not book** — Agendas, pre-reads and the briefing note are the valuable half of a meeting. A worker delivers those as docs regardless of whether it can see a calendar. 3. **Capture the follow-ups where they were said** — With Slack connected, the thread after a meeting becomes prefilled task suggestions in the Inbox, which is the failure mode calendars never fixed anyway. > **No credential, no pretending** > > Nothing is stored for a pending connection. There is no placeholder row, no partial grant and no worker that will claim to have checked your availability. When the Google flow ships, authorizing it once will connect it for the entire organisation, the way every other connection in the catalog behaves. ## Workers that will want this connection - [ai-workers/executive-assistant](https://www.polarishq.co/ai-workers/executive-assistant) - [ai-workers/recruiter](https://www.polarishq.co/ai-workers/recruiter) - [ai-workers/project-coordinator](https://www.polarishq.co/ai-workers/project-coordinator) - [use-cases/hr/interview-scheduling](https://www.polarishq.co/use-cases/hr/interview-scheduling) - [use-cases/executive/meeting-follow-ups](https://www.polarishq.co/use-cases/executive/meeting-follow-ups) - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) ## Questions people ask **Can a Polaris worker schedule meetings for me?** Not yet. Google Calendar authorizes through Google sign-in and that flow has not shipped, so no worker can read availability or create events. Scheduling-adjacent work such as agendas, pre-reads and follow-up capture does run today. **How does Polaris handle deadlines without a calendar?** Through tasks and Focus. Every task carries an owner and a due date, and Focus arranges the non-negotiables across today, this week and the next thirty days. It is a commitment view rather than an hour-by-hour grid. **Is interview scheduling possible with a recruiter worker?** Partly. A recruiter worker can screen candidates, prepare the interview pack and track each stage as tasks. Proposing and booking the slot itself needs calendar access, so a person does that step for now. **Will connecting Calendar later require rehiring my workers?** No. Connections belong to the organisation, not to a worker. A worker already carrying the pending Calendar connection moves to connected the moment a credential exists, keeping its instructions, skill files and history. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/google-drive - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/use-cases/hr/interview-scheduling - https://www.polarishq.co/glossary/focus-lane --- --- title: "Polaris Google Drive Integration: Files and Status" description: "Google Drive is an OAuth connection in the Polaris catalog for docs, sheets and files. The Google sign-in flow is not live, so the connection stays pending." url: https://www.polarishq.co/integrations/google-drive section: Integrations updated: 2026-08-21 --- # Connect Google Drive to Polaris Files a worker produces already arrive as attachments on the task. Files your team already keeps in Drive are the part still waiting. ## The short answer Google Drive is an OAuth connection in the Polaris catalog covering docs, sheets and files. It authorizes through Google sign-in, a flow that has not shipped, so the connection stays pending and stores no credential. Files produced by a worker do not depend on it: deliverables arrive as attachments on the task comment, and written work lands in versioned Polaris docs. - **Auth mode:** Google OAuth - **Status:** Pending, flow not shipped - **Output files:** Attached to the task today ## Two different file problems People ask for a Drive connection for two unrelated reasons. The first is reading: the brand guidelines, the spreadsheet of last quarter's numbers, the folder of contracts. The second is output: where the thing a worker made ends up. Only the first one needs Drive. Output is already solved. A worker attaches files to the task it delivered, including generated PDFs, and written work is drafted into the Docs tree where it is versioned, commentable and reviewable. Nothing about that path routes through Google. ## Reading versus producing **Reading Drive, not available yet** - Working from a spreadsheet the finance team maintains - Pulling context out of an existing folder of material - Reading a contract your legal folder already holds - Anything that starts with a file somebody else made **Producing files, available now** - Deliverables attached to the task comment, PDFs included - Written work as versioned docs with comments and file review - A nested doc tree that behaves like the wiki it replaces - Content pasted into the task when the source is short enough ## What the Drive connection covers The catalog entry names docs, sheets and files. Roadmap, not shipped. - **Docs** — Prior art a worker should read before writing anything new, which is the difference between a first draft and a restatement. - **Sheets** — The numbers a finance or data worker is expected to reconcile against, in the format they are actually kept in. - **Files** — Everything else a team hoards in folders, where the scope of a grant matters more than the file count. > **The workaround that is not really a workaround** > > Paste the material into the task or into a doc. Task context is what a worker reads first, and a two-page brief pasted into the description beats a Drive grant covering four thousand files nobody audited. When Drive ships, the grant is still worth scoping deliberately. ## Workers that will want this connection - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [ai-workers/financial-analyst](https://www.polarishq.co/ai-workers/financial-analyst) - [ai-workers/content-writer](https://www.polarishq.co/ai-workers/content-writer) - [use-cases/finance/budget-reporting](https://www.polarishq.co/use-cases/finance/budget-reporting) - [use-cases/operations/sop-maintenance](https://www.polarishq.co/use-cases/operations/sop-maintenance) - [cloud-claude-code/ai-agents-that-produce-files](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) ## Questions people ask **Can a Polaris worker open a file from my Google Drive?** Not today. Drive authorizes through Google sign-in and that flow has not shipped, so the connection stays pending with no credential stored. Material a worker needs can be pasted into the task or into a Polaris doc in the meantime. **Where do the files a worker produces end up?** On the task. A delivery arrives as a comment with the work in it, and generated files, including PDFs, are attached to that comment. Written work is also drafted into the Docs tree, where it is versioned and open to comments. **Does Polaris replace Google Drive?** No, and it is not trying to. Polaris Docs is a nested block editor with versioning and review, which covers written work. A general file store for video, design source files and everything else your company keeps is a different product. **Will a Drive grant cover my whole account?** That is decided at Google's consent screen when the flow ships, not inside Polaris. Whatever the grant covers is what a worker can reach, which is why the scope is worth reading rather than clicking past. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/integrations/gmail - https://www.polarishq.co/integrations/google-calendar - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/use-cases/finance/budget-reporting --- --- title: "Polaris Figma Integration for Design Workers" description: "Connect Figma with a personal access token, verified live against Figma before storage, to give AI workers design files and comments to work from." url: https://www.polarishq.co/integrations/figma section: Integrations updated: 2026-08-21 --- # Connect Figma to Polaris A Figma token belongs to a person, and Polaris shows you which person it verified, which is the detail worth checking before you store it. ## The short answer Connecting Figma gives a Polaris worker access to design files and comments through a personal access token. Polaris verifies the token against Figma before storing it and displays the account email Figma returned. Because the token carries one person's access, the account you create it on decides the boundary, and the connection is authorized once for the whole organisation. - **Credential:** Personal access token - **Verified live:** Figma returns the account email - **Boundary:** The token owner's file access ## Design work that is not designing The reason to connect Figma is rarely to have an AI push pixels. It is the surrounding work: the review notes nobody wrote up, the handoff spec that lags the file by three days, the component that got renamed in one place and not in the other, the comment thread with four open questions in it. That work is text about a design, and it is the kind of task that sits in a backlog for a month because it belongs to whoever has the least going on that week. ## What the connection gives a worker Bounded by what the token's owner can open in Figma. Nothing broader. - **Design files** — The file is the source of truth for a handoff note or a review summary, and reading it beats working from a screenshot somebody pasted in a channel. - **Comments** — Comment threads are where design decisions actually get made and where they reliably get lost. A worker that can read them can turn them into tracked tasks. - **A basis for consistency checks** — Auditing whether a set of files still follows the same conventions is tedious for a person and unbothered work for a machine. - **Handoff context** — A design review or an asset handoff written from the file rather than from memory is the difference between a spec engineers trust and one they re-ask about. ## Connecting Figma 1. **Create a personal access token** — From Figma's settings. The value starts with figd_ and is shown once, so paste it straight into Polaris. 2. **Pick the account deliberately** — The token inherits that account's file access. If your team has a limited account for tooling, use it rather than the design lead's login. 3. **Authorize as an owner or admin** — Polaris calls Figma with the token, refuses it if Figma does, and shows you the account email it verified before storing anything. > **Check the email in the confirmation** > > After a successful authorization, the connect panel shows the account Figma identified. That single line answers the question a security review would ask: whose access did we just hand to every worker in the organisation carrying this connection? If the answer is somebody with more access than the work needs, revoke and issue a narrower token. ## Workers and work this connection feeds - [use-cases/design/design-review](https://www.polarishq.co/use-cases/design/design-review) - [use-cases/design/asset-handoff](https://www.polarishq.co/use-cases/design/asset-handoff) - [use-cases/design/design-system-maintenance](https://www.polarishq.co/use-cases/design/design-system-maintenance) - [use-cases/design/brand-consistency-audits](https://www.polarishq.co/use-cases/design/brand-consistency-audits) - [ai-workers/technical-writer](https://www.polarishq.co/ai-workers/technical-writer) - [ai-workers/project-coordinator](https://www.polarishq.co/ai-workers/project-coordinator) ## Questions people ask **Can an AI worker edit my Figma files?** Treat this connection as access to read design files and comments. The work it feeds is written output such as review notes, handoff specs and consistency audits, delivered as comments and docs in Polaris for a person to close. **Which Figma files does the connection cover?** Whatever the token's owner can open. Figma personal access tokens carry that person's access, so choosing the account is choosing the scope. Polaris shows you the account email it verified so the choice is visible rather than assumed. **How is the Figma token stored?** Server-side, one row per organisation, written only after Figma accepted it. The browser never reads it back, and revoking deletes it for every worker carrying the connection at once. **What happens when the token expires or is rotated?** Figma starts rejecting it and the worker loses access. Paste a replacement into the same connect panel and it overwrites the stored value after the same live check. There is no need to touch individual workers. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/use-cases/design/design-review - https://www.polarishq.co/use-cases/design/asset-handoff - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/integrations/slack --- --- title: "Polaris HubSpot Integration: CRM for AI Workers" description: "Connect HubSpot with a private-app token, verified live against your portal, so AI workers can work from real CRM contacts and deals instead of a stale export." url: https://www.polarishq.co/integrations/hubspot section: Integrations updated: 2026-08-21 --- # Connect HubSpot to Polaris Private-app tokens are scoped where you create them, which makes HubSpot one of the easier connections to grant narrowly. ## The short answer Connecting HubSpot gives a Polaris worker your CRM contacts and deals through a private-app access token. Polaris calls HubSpot to verify the token before storing it and identifies the portal it belongs to. Private apps are scoped in HubSpot's own settings, so a worker doing research reads without ever holding permission to change a deal stage. - **Credential:** Private-app access token - **Verified live:** Polaris identifies the portal - **Scoping:** Per private app, in HubSpot ## CRM hygiene is the job nobody does Every sales team has the same unpaid tax. Contacts with no company, deals sitting in a stage they left three weeks ago, notes that exist as a memory rather than a record, and a pipeline review that starts with twenty minutes of arguing about whether the data is real. It is exactly the shape of work an AI worker handles well: high volume, rule-driven, boring enough that a person postpones it, specific enough that acceptance criteria can say what good looks like. ## What the connection gives a worker Everything below is bounded by the scopes on the private app you created. - **Contacts** — The list a researcher enriches, a coordinator deduplicates, or an SDR reads before writing anything to anybody. - **Deals** — Pipeline state as it actually is, which is the only useful starting point for a weekly review or a stale-deal sweep. - **A real basis for outreach** — An SDR worker that reads the CRM writes a first line about the account instead of a first line about the industry. - **Reporting inputs** — Pipeline summaries and stage-movement write-ups become a delivered doc rather than a spreadsheet somebody rebuilds every Monday. ## Scoping the private app HubSpot is where the boundary is set. These are the decisions that matter before you paste anything. | Question | Conservative answer | | --- | --- | | Which objects? | Only the ones the worker is briefed on, usually contacts and deals | | Read or write? | Read-only unless a worker is explicitly meant to update records | | Which portal? | The production portal only if the work needs it; a sandbox for anything experimental | | Who creates it? | An admin who understands the scope screen, not whoever is closest | > **Customer data raises the stakes** > > A CRM holds other people's personal data, and connecting it means your obligations to those people now include this tool. Polaris stores the token server-side after verifying it, and workers reach exactly what the scopes allow. What Polaris cannot do is decide on your behalf whether that grant fits your privacy commitments. Read the scope screen with that in mind. ## Workers and work this connection feeds - [ai-workers/sdr](https://www.polarishq.co/ai-workers/sdr) - [ai-workers/customer-success-manager](https://www.polarishq.co/ai-workers/customer-success-manager) - [use-cases/sales/crm-hygiene](https://www.polarishq.co/use-cases/sales/crm-hygiene) - [use-cases/sales/pipeline-management](https://www.polarishq.co/use-cases/sales/pipeline-management) - [use-cases/sales/lead-research](https://www.polarishq.co/use-cases/sales/lead-research) - [use-cases/sales/proposal-writing](https://www.polarishq.co/use-cases/sales/proposal-writing) ## Questions people ask **What HubSpot credential does Polaris use?** A private-app access token, the kind beginning pat-, created under Settings and Integrations in HubSpot. Polaris verifies it against HubSpot's account details endpoint and identifies the portal before storing anything. **Can a worker change a deal or delete a contact?** Only if the private app you created has those scopes. HubSpot enforces the boundary, and Polaris uses the token as issued. A read-only private app produces a worker that can analyse the pipeline and cannot alter it. **Is connecting a CRM safe from a privacy standpoint?** That is a judgement about your own obligations rather than a claim we can make for you. Polaris verifies the token, stores it server-side, and gives workers exactly the scopes you granted. There is no certification or compliance programme behind it to lean on, and pretending otherwise would be the real risk. **Can I connect HubSpot without giving access to the whole portal?** Scope the private app to specific objects, which is the mechanism HubSpot provides. Polaris has no additional filter of its own, so the private-app configuration is the boundary that counts. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/ai-workers/customer-success-manager - https://www.polarishq.co/use-cases/sales/crm-hygiene - https://www.polarishq.co/use-cases/sales/pipeline-management - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/for/agencies --- --- title: "Polaris Stripe Integration: Read-Only by Design" description: "Connect Stripe with a restricted read-only key so AI workers can reconcile payments and invoices without ever holding permission to move money." url: https://www.polarishq.co/integrations/stripe section: Integrations updated: 2026-08-21 --- # Connect Stripe to Polaris This is the one connection where the key you choose matters more than anything on this page. ## The short answer Connecting Stripe gives a Polaris worker payments and invoice data through a restricted API key. Polaris verifies the key against Stripe before storing it and shows the account name it returned. Restricted keys with read-only scopes are the right credential here: a worker reconciling invoices needs to read charges, and nothing about that job requires the ability to move money. - **Credential:** Restricted key - **Recommended scopes:** Read-only - **Verified live:** Stripe returns the account name > **Use a restricted key, not a secret key** > > Stripe lets you create restricted keys with per-resource read permissions. A worker chasing unpaid invoices or categorising expenses needs read access to charges, invoices and customers, and nothing more. Polaris cannot inspect what your key is allowed to do, so this choice is yours alone and it is the one that decides the blast radius. ## What finance work actually needs Look at the finance tasks that go undone in a small company. Which invoices are late. What that unlabelled charge was. Whether last month closed cleanly. Whether the budget line matches the bank. Every one of those is a reading problem. Somebody opens a dashboard, exports something, matches it against something else and writes a paragraph. The writing is the deliverable and the reading is the tedium, which is a good description of the work AI workers are worth hiring for. None of it requires the ability to issue a refund, and a credential that can issue refunds is a credential you have to think about every time somebody joins the org. ## What the connection gives a worker Bounded by the restricted key's scopes. - **Payments** — The record of what came in, which is the starting point of every reconciliation anybody has ever put off. - **Invoices** — Status, amounts and dates, so a chase list is generated from Stripe rather than from a spreadsheet that was accurate a fortnight ago. - **Customer records** — Enough to attribute a charge to an account when the finance question is really an account question. - **Material for the monthly close** — A close checklist that arrives with the numbers already filled in is a different kind of Monday from one that starts empty. ## Two credentials, two different weeks **Restricted read-only key** - A worker can describe the state of your revenue - Nothing it does can alter a customer's money - A leaked key exposes data, which is bad and bounded - Security review takes one screen **A full secret key** - The worker can read exactly the same things - It also holds permissions no task needs - A leaked key becomes an incident rather than a problem - You now audit every worker that carries the connection ## Workers and work this connection feeds - [ai-workers/bookkeeper](https://www.polarishq.co/ai-workers/bookkeeper) - [ai-workers/financial-analyst](https://www.polarishq.co/ai-workers/financial-analyst) - [use-cases/finance/invoice-tracking](https://www.polarishq.co/use-cases/finance/invoice-tracking) - [use-cases/finance/expense-categorization](https://www.polarishq.co/use-cases/finance/expense-categorization) - [use-cases/finance/monthly-close-checklist](https://www.polarishq.co/use-cases/finance/monthly-close-checklist) - [use-cases/finance/budget-reporting](https://www.polarishq.co/use-cases/finance/budget-reporting) ## Questions people ask **Can an AI worker refund a customer or move money in Stripe?** Only if the key you stored permits it, which is why the catalog recommends a restricted read-only key. Stripe enforces the permissions and Polaris uses the credential exactly as issued. Create a key that cannot move money and no instruction to a worker can make it move money. **How does Polaris verify a Stripe key?** It calls Stripe's account endpoint before storing anything. A rejected key returns Stripe's status code as an error and nothing is written. A valid one comes back with the account display name, which the connect panel shows so you can confirm you connected the account you meant to. **Where is the Stripe key kept?** Server-side against your organisation, written by the function that verified it. The browser never receives it back. Revoking deletes the row and returns every worker carrying the Stripe connection to pending in the same action. **Should I connect a production Stripe account at all?** For invoice tracking and reconciliation, production data is the point and a read-only key is a proportionate way to reach it. For anything experimental, connect a test-mode key first. Polaris is in free public beta and has no compliance certification to offer, so decide with that on the table. **Can one organisation connect two Stripe accounts?** No. Credentials are stored one per tool per organisation, so a second Stripe key replaces the first rather than sitting beside it. Teams with several entities need separate organisations for now. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/ai-workers/financial-analyst - https://www.polarishq.co/use-cases/finance/invoice-tracking - https://www.polarishq.co/use-cases/finance/monthly-close-checklist - https://www.polarishq.co/integrations/hubspot - https://www.polarishq.co/use-cases/finance --- --- title: "Polaris Supabase Integration: Keys and Cautions" description: "Connect a Supabase project key to give AI workers your product database. The one connection with no live verification yet, and the one with the strongest keys." url: https://www.polarishq.co/integrations/supabase section: Integrations updated: 2026-08-21 --- # Connect Supabase to Polaris Supabase is the connection where a careless key choice does the most damage, and the only one Polaris cannot yet check for you. ## The short answer Connecting Supabase gives a Polaris worker a project's database and auth through a project key from your Supabase dashboard. Supabase is the one connection in the catalog with no live verification call yet, so the key is stored and labelled as unvalidated rather than confirmed. Prefer a restricted key over a service-role key, which bypasses row-level security entirely. - **Credential:** Service-role or restricted key - **Live check:** None yet, and labelled as such - **Risk to weigh:** Service-role keys bypass RLS > **Two things to know before you paste anything** > > First, a Supabase service-role key bypasses row-level security and reads every row in the project, which is a different category of access from every other connection in this catalog. Prefer a restricted key scoped to what a worker actually needs. Second, Polaris has no live validation call for Supabase yet, so a mistyped key is stored and the connection is labelled as not validated instead of failing loudly. ## Why the honest label matters Nine of the thirteen credentialled connections are checked against the provider before Polaris writes anything. Slack, GitHub, Notion, Linear, Stripe, Figma, HubSpot, WhatsApp and Instagram all get a real API call, and a bad token is refused with the provider's own error. Supabase does not have that check yet. The product says so in the stored label rather than showing a green state it has not earned. It is a small thing, and it is the difference between a connection you can reason about and one you have to test by watching a worker fail. ## What the connection gives a worker A product database, which is the most sensitive thing most companies own. - **Your project data** — Data-quality sweeps, metric definitions checked against reality, and the answer to a question that would otherwise become a ticket for whoever owns the schema. - **Auth records** — Signup and account state, useful for onboarding and support work and worth thinking twice about before you grant it. - **A basis for reporting** — Recurring reporting written from the source rather than from a dashboard export that drifted three weeks ago. ## Choosing the key Supabase issues several kinds. They are not interchangeable. | Key | Reaches | Use it here? | | --- | --- | --- | | Restricted key, scoped | Only what you granted it | Yes, this is the default worth taking | | Service-role key | Every row, ignoring row-level security | Only if a task genuinely requires it, and knowingly | | Publishable key | What an anonymous client can see | Fine when a worker only needs public data | ## Workers and work this connection feeds - [ai-workers/data-analyst](https://www.polarishq.co/ai-workers/data-analyst) - [ai-workers/qa-engineer](https://www.polarishq.co/ai-workers/qa-engineer) - [use-cases/data/data-quality-monitoring](https://www.polarishq.co/use-cases/data/data-quality-monitoring) - [use-cases/data/reporting-automation](https://www.polarishq.co/use-cases/data/reporting-automation) - [use-cases/data/metric-definitions](https://www.polarishq.co/use-cases/data/metric-definitions) - [use-cases/data/dashboard-requests](https://www.polarishq.co/use-cases/data/dashboard-requests) ## Questions people ask **Does Polaris check my Supabase key before storing it?** Not yet. Supabase is the one credentialled connection in the catalog without a live validation call, so the key is stored and the connection is labelled as not validated. Nine other connections do get a real API call and refuse a bad credential outright. **Should I give a worker a service-role key?** Usually not. Service-role keys ignore row-level security and read everything in the project. A restricted key scoped to the tables a worker is briefed on does the same job for almost every task, with a fraction of the exposure. **Is this how Polaris itself is built?** Yes, which is why the caution is specific rather than generic. Polaris runs on Supabase for Postgres, auth, realtime, row-level security and edge functions, and connection credentials are stored in that same database, written by an edge function and never read back into a browser. **Can a worker write to my database?** That depends entirely on the key you stored. Supabase enforces the permissions, and Polaris uses the credential as issued. If you do not want writes, do not issue a key that can write. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/data-analyst - https://www.polarishq.co/use-cases/data/data-quality-monitoring - https://www.polarishq.co/use-cases/data/reporting-automation - https://www.polarishq.co/use-cases/data - https://www.polarishq.co/integrations/github --- --- title: "Polaris WhatsApp Integration for Customer Messages" description: "Connect WhatsApp Business with a Meta system-user token so AI workers can work from customer chats, verified against Meta before Polaris stores anything." url: https://www.polarishq.co/integrations/whatsapp section: Integrations updated: 2026-08-21 --- # Connect WhatsApp to Polaris For a lot of companies WhatsApp is the support desk, the sales channel and the customer record, and none of it is written down anywhere else. ## The short answer Connecting WhatsApp gives a Polaris worker chats and customer messages from a WhatsApp Business app, authorized with a Meta system-user access token. Polaris calls Meta with the token before storing it and shows the identity Meta returned. Customer conversations that only exist in a phone become material a support worker can triage, summarise and turn into tracked tasks. - **Credential:** Meta system-user access token - **Set up at:** The Meta developer console - **Verified live:** Meta accepts the token ## The channel that never made it into a tool In large parts of the world, WhatsApp is where customers actually talk to a business. Bookings, complaints, reschedules, a photo of a broken part, a question about the price. Almost none of it reaches the systems the company runs on, because moving it there is manual and nobody has the hour. That gap is why this connection exists. A conversation that lives only in a phone cannot be triaged, cannot be counted and cannot be handed over when the person holding the phone goes on holiday. ## What the connection gives a worker Bounded by what the Meta app and its token permit. - **Customer chats** — The conversation as it happened, which is what triage needs. A summary written by whoever answered is not the same artefact. - **Material for tickets** — A support worker can turn a thread into a task with a bucket, an owner and acceptance criteria, so it stops depending on one person's memory. - **Patterns across conversations** — The same complaint appearing eleven times in a fortnight is a product finding, and it is invisible while each instance lives in a separate chat. - **A record that survives handover** — Once a conversation has produced a task and a doc, the next person on shift starts from the record rather than from scratch. ## Connecting WhatsApp 1. **Set up the WhatsApp Business app at Meta** — This connection sits on the Meta developer platform rather than on consumer WhatsApp, so a business app has to exist before any credential does. 2. **Create a system-user access token** — System users are Meta's mechanism for tokens that belong to a business rather than to an employee's personal account, which is the right shape for a tool credential. 3. **Authorize in Polaris** — An owner or admin pastes the token. Polaris calls Meta with it, refuses it if Meta does, and stores it org-wide once Meta answers. > **What the live check proves, and what it does not** > > Polaris asking Meta about a token confirms that Meta accepts it and returns an identity. It does not enumerate which permissions that token carries. So a token can verify successfully and still be scoped too narrowly for the job, or wider than you intended. The Meta app configuration is the real boundary, and it is worth reading rather than trusting the green tick. ## Workers and work this connection feeds - [ai-workers/support-specialist](https://www.polarishq.co/ai-workers/support-specialist) - [ai-workers/customer-success-manager](https://www.polarishq.co/ai-workers/customer-success-manager) - [use-cases/customer-support/ticket-triage](https://www.polarishq.co/use-cases/customer-support/ticket-triage) - [use-cases/customer-support/escalation-tracking](https://www.polarishq.co/use-cases/customer-support/escalation-tracking) - [use-cases/customer-support/customer-feedback-loops](https://www.polarishq.co/use-cases/customer-support/customer-feedback-loops) - [use-cases/customer-support/response-templates](https://www.polarishq.co/use-cases/customer-support/response-templates) ## Questions people ask **Does this connect my personal WhatsApp?** No. The credential is a Meta system-user access token for a WhatsApp Business app, created in the Meta developer console. Personal WhatsApp accounts are not part of this connection and there is no path in the product to add one. **How does Polaris verify the token?** By calling Meta's Graph API with it before storing anything. A token Meta rejects produces an error and nothing is written. A valid one returns an identity, which the connect panel shows you. **Can a worker reply to a customer on WhatsApp?** What a token permits is set in the Meta app, and Polaris uses the credential as issued. The workflow the product is built around is different anyway: a worker prepares the reply, delivers it as a comment on the task, and a person decides whether it goes out. **What about customer privacy?** Customer conversations are personal data, and connecting them means this tool falls inside whatever commitments you have made to those customers. Polaris verifies and stores the token server-side and gives workers what the Meta app allows. It carries no certification or compliance programme, so that judgement stays with you. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/customer-support - https://www.polarishq.co/integrations/instagram - https://www.polarishq.co/for/small-business --- --- title: "Polaris Instagram Integration for Social Workers" description: "Connect Instagram with a long-lived Meta page token so an AI worker can work from real posts and engagement data instead of a screenshot of last week." url: https://www.polarishq.co/integrations/instagram section: Integrations updated: 2026-08-21 --- # Connect Instagram to Polaris Engagement data is the only honest input to a content calendar, and it is the input most calendars are written without. ## The short answer Connecting Instagram gives a Polaris worker posts and engagement data through a long-lived page access token from a Meta app. Polaris verifies the token with Meta before storing it. A social worker can then plan and draft from what your audience actually responded to, rather than from an assumption about what performs, with a person approving anything that publishes. - **Credential:** Long-lived page access token - **Set up at:** The Meta developer console - **Verified live:** Meta accepts the token ## Most content calendars are written blind The usual process is a person opening a spreadsheet, remembering two posts that did well, and filling four weeks with variations of them. The data that would settle the argument is one screen away and nobody pulls it, because pulling it, tabulating it and interpreting it is an hour of work that produces no visible output. That hour is the job. Give a worker the engagement data and the calendar arrives with reasons attached, which also makes it arguable, which is what makes it better than the version nobody could challenge. ## What the connection gives a worker Bounded by what the Meta app and page token permit. - **Posts** — What you actually published and when, which is the baseline any recommendation has to beat. - **Engagement data** — How each post performed, so a content plan cites evidence rather than instinct dressed up as strategy. - **Input for planning** — A monthly calendar built from the last three months of results, delivered as a doc a person edits before anything is scheduled. - **A record in the workspace** — Social work stops living in one person's head and starts living as tasks with owners, dates and a delivery trail. ## Where the human stays in charge Social is a public channel, which changes where the approval boundary sits. **The worker prepares** - The performance read on recent posts - A calendar with a reason next to each slot - Draft copy and hooks as versioned docs - The delivery as a comment on the task **You decide** - What is in your brand voice and what is not - What publishes, and when - Which numbers are worth reacting to at all - Closing the task and rating the delivery > **Agents deliver, humans close** > > This applies everywhere in Polaris and it matters most on a public channel. A worker never marks its own task done. It hands back the calendar, the draft or the analysis as a comment on the task, a person closes and rates it, and nothing reaches an audience because a machine thought it was ready. ## Workers and work this connection feeds - [ai-workers/social-media-manager](https://www.polarishq.co/ai-workers/social-media-manager) - [ai-workers/content-writer](https://www.polarishq.co/ai-workers/content-writer) - [use-cases/marketing/social-media-scheduling](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) - [use-cases/marketing/content-calendar](https://www.polarishq.co/use-cases/marketing/content-calendar) - [use-cases/marketing/competitor-monitoring](https://www.polarishq.co/use-cases/marketing/competitor-monitoring) - [use-cases/marketing/campaign-planning](https://www.polarishq.co/use-cases/marketing/campaign-planning) ## Questions people ask **What credential does the Instagram connection need?** A long-lived page access token created through a Meta app in the developer console. Polaris calls Meta with it before storing anything and refuses a token Meta rejects. **Can a worker post to Instagram on my behalf?** What a token permits is decided in the Meta app, and Polaris uses it as issued. The product's own boundary is firmer than that: a worker delivers drafts and plans as comments and docs, and a person publishes. **Does the connection cover other Meta surfaces?** The catalog entry is Instagram, covering posts and engagement data, and WhatsApp is listed separately with its own credential. Both are set up through the Meta developer console, and both are authorized independently in Polaris. **Who in my organisation can connect it?** Only an owner or an admin. The check runs on the server, so a member who opens the panel is told to ask an owner. Once stored, the credential belongs to the organisation and every worker carrying the Instagram connection uses it. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/social-media-manager - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/use-cases/marketing/social-media-scheduling - https://www.polarishq.co/use-cases/marketing/content-calendar - https://www.polarishq.co/integrations/whatsapp - https://www.polarishq.co/use-cases/marketing --- --- title: "Web Search for AI Workers: No Credential Needed" description: "The one Polaris connection that needs no credential. Live web search runs inside the delivery session, logged query by query and priced by the formula." url: https://www.polarishq.co/integrations/web-search section: Integrations updated: 2026-08-21 --- # Web search in Polaris Nothing to authorize, nothing to store, and every query a worker runs appears in the activity feed while it works. ## The short answer Web search is the one connection in the Polaris catalog that needs no credential. A worker running a task on a cloud machine performs live searches during the session, each one recorded in the activity feed as it happens. Searches are capped per session and counted in the human-hours estimate at roughly twelve minutes each, so research effort shows up on the bill line by line. - **Credential:** None - **Runs:** Inside the delivery session - **Counted at:** About 12 minutes per search - **Per session:** Capped at a handful of searches ## Why this one is different Thirteen connections in the catalog exist to give a worker access to something your company owns. Web search gives it access to what everybody owns, which is why there is no token, no consent screen and no owner-only authorization step. Every worker has it. It is also the connection that changes the character of the output most. A brief written without search is a language model recalling; a brief written with it is a worker reading current pages and citing them. The second kind is the only kind worth paying for. ## What search does inside a session The mechanics, because they explain both the quality and the cost. - **It runs while the machine works** — Searches happen inside the tool loop on the cloud machine, not as a preparation step. A worker searches, reads, revises what it thought, and searches again. - **Every query is posted** — Each search is emitted to the activity feed with the query text, so you can watch what a worker looked for and judge whether it looked in sensible places. - **The session is capped** — A single run allows a handful of searches rather than an unbounded crawl. A question needing forty sources is several tasks, not one heroic session. - **It feeds the bill honestly** — The hours formula counts each search at about twelve minutes of human-equivalent effort, which is roughly what the same lookup and read costs a person. ## What five searches cost Worked from the published hours formula, not measured on a customer. Hours are human-equivalent and every line is written to the worker's work log. | Line | Human-equivalent minutes | | --- | --- | | Picking up the task | 15 | | 5 live web searches at 12 minutes each | 60 | | The finished prose, at 90 characters per minute | varies with length | | Search subtotal alone, at $2 per hour | $2.00 | > **What web search cannot reach** > > The open web and nothing else. Paywalled reports, gated PDFs, anything behind a login and anything your company keeps privately are outside it, and a worker will say so rather than fill the gap with a confident guess. If the answer lives in your own systems, that is what the other thirteen connections are for. ## Workers and work this connection feeds - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [ai-workers/competitive-analyst](https://www.polarishq.co/ai-workers/competitive-analyst) - [ai-workers/seo-specialist](https://www.polarishq.co/ai-workers/seo-specialist) - [use-cases/executive/strategic-research](https://www.polarishq.co/use-cases/executive/strategic-research) - [use-cases/marketing/competitor-monitoring](https://www.polarishq.co/use-cases/marketing/competitor-monitoring) - [use-cases/sales/lead-research](https://www.polarishq.co/use-cases/sales/lead-research) - [use-cases/product/competitive-tracking](https://www.polarishq.co/use-cases/product/competitive-tracking) ## Questions people ask **Do I have to enable web search for a worker?** It is in the catalog as a connection with no credential and no authorization step, so there is nothing to enable and nothing to store. Adding it to a worker's list is a statement about the role rather than a permission grant. **How do I know a worker actually searched?** Every query is posted to the activity feed as it runs, with the text of the search, and the count appears on the work log line that priced the job. A confident brief with no searches logged is a brief worth challenging, and the record to challenge it with is right there. **Does web search make a task more expensive?** Yes, by design and transparently. The hours formula counts each search at about twelve minutes of human-equivalent effort, so five searches add an hour, which is two dollars. That is what makes research work priceable rather than a mystery line on a bill. **Can a worker read a specific page I give it?** Put the URL in the task and say what you want from it. Task context is the first thing a worker reads, and pointing it at a page beats hoping a search surfaces the one you had in mind. **Why is the number of searches limited?** Because a session that searches indefinitely produces a worse answer and a bigger bill. The cap keeps a run bounded to minutes and keeps the hours estimate honest. Deep questions are better split into several tasks, each with its own acceptance criteria. ## Related - https://www.polarishq.co/integrations - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/use-cases/executive/strategic-research - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "Polaris for startups, agencies, remote teams and 9 more" description: "How Polaris fits twelve specific kinds of team, from agencies that resell hours to remote teams losing decisions to scrollback, including where it fits badly." url: https://www.polarishq.co/for section: Solutions updated: 2026-08-21 --- # Twelve kinds of team, twelve different arguments The case for Polaris is not the same case twice. Pick the team you actually run. ## The short answer Polaris fits teams that keep work in a tracker, notes in a doc tool and conversation in chat, and pay for all three by the seat. The workspace is free with no seat count. AI workers are hired in about sixty seconds, run on cloud machines, and bill roughly two dollars per human-hour of work they deliver. Nothing delivered, nothing billed. - **Audience pages:** 12 - **Software cost:** $0, no seats - **Billing unit:** ~$2 per human-hour delivered - **Status:** Free public beta ## Why these pages are not one page repeated An agency and a nonprofit both save money on seats. That is where the similarity ends. An agency cares because the gap between its bill rate and its delivery cost is the entire business. A nonprofit cares because per-seat pricing charges it for forty volunteers who log in twice a quarter. Same fact, two completely different arguments, and a page that makes the generic version of both persuades neither. So each page below picks the one thing that is most true for that kind of team, and argues that. Some of them argue that Polaris is a poor first move right now. Those pages are not softer than the others. They are the ones worth reading. ## The four arguments underneath the twelve pages Every page is a version of one of these, sharpened for a particular week of work. - **Cost that only exists when work exists** — Per-seat software bills for a place to put work, whether or not anyone puts anything there. Polaris bills for work that came back finished. Startups, nonprofits and small businesses feel this hardest because their headcount is lumpy or their budget is a grant line. - **An hourly cost line you can audit** — Human-equivalent hours are estimated by an open formula from observable effort and logged job by job. Agencies and consultancies live and die on the arithmetic between what an hour costs them and what they charge for it. - **Work that continues after the laptop closes** — A real cloud machine wakes per task. That matters most to distributed teams, where the useful hours are the ones when the other half of the team is asleep, and to founders who already run agents locally and lose them at midnight. - **One workspace instead of a tab per tool** — Tasks, docs and team chat in one product, with humans and AI workers in the same members table and the same assignment flow. Product, design and software teams notice this as the end of copying the same decision into three places. > **Read this before any page below** > > Polaris is in free public beta. There are no customers, no case studies and no traction numbers to show you. What exists is a working product, four recorded demos including an unstaged machine session, and a bill that stays at zero until an AI worker hands back finished work. Every page here is written against that reality, not around it. ## Every audience page - [Polaris for startups](https://www.polarishq.co/for/startups) — Your next hire takes about sixty seconds and costs nothing in the weeks it has nothing to do. - [Polaris for agencies](https://www.polarishq.co/for/agencies) — You already know what an hour of delivery sells for. This is a page about what it costs. - [Polaris for solo founders](https://www.polarishq.co/for/solo-founders) — Being the whole org chart is not a time problem. It is an accountability problem. - [Polaris for remote teams](https://www.polarishq.co/for/remote-teams) — The problem was never the meetings. It is what happens to a decision made while half the team is asleep. - [Polaris for software teams](https://www.polarishq.co/for/software-teams) — Nobody joined your team to write the release notes. Something still has to write them. - [Polaris for consultancies](https://www.polarishq.co/for/consultancies) — The scarce resource in a consultancy is not hours. It is partner attention, spent on work a partner should not be doing. - [Polaris for ecommerce brands](https://www.polarishq.co/for/ecommerce-brands) — You staff for peak and pay for it in February, or you staff for February and suffer in November. - [Polaris for nonprofits](https://www.polarishq.co/for/nonprofits) — Your org chart is six staff and forty people who help. Per-seat pricing was designed for the opposite shape. - [Polaris for product teams](https://www.polarishq.co/for/product-teams) — Look at a product manager's calendar and subtract the meetings. What is left is mostly reformatting. - [Polaris for small business](https://www.polarishq.co/for/small-business) — The jobs that never get done in a small business are not hard. There is simply nobody free to do them. - [Polaris for design studios](https://www.polarishq.co/for/design-studios) — Nobody has ever put a line item for asset renaming on an invoice, and every studio pays for it anyway. - [Polaris for technical founders](https://www.polarishq.co/for/technical-founders) — The agent that wrote half your week's work is single-player, local, and gone at midnight. ## The pages behind the arguments - [cost](https://www.polarishq.co/cost) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [ai-workers](https://www.polarishq.co/ai-workers) - [cloud-claude-code](https://www.polarishq.co/cloud-claude-code) - [use-cases](https://www.polarishq.co/use-cases) ## Questions people ask **Which of these twelve pages should I read if my team is a mix?** Read the one that describes how you get paid, not the one that describes what you build. A four-person studio that also resells hours has more in common with the agencies page than with the design studios page. The commercial shape of the team predicts the argument better than the craft does. **Does Polaris cost more for a bigger team?** No. The workspace is free with no seat count and no tier, so unlimited humans, tasks, workstreams and docs cost nothing at any headcount. The only bill is roughly two dollars per human-equivalent hour that an AI worker delivers, which tracks the work you asked for rather than the number of people who can log in. **Is there a kind of team Polaris is clearly wrong for today?** Yes, several. Organisations with a signed data processing agreement that forbids third-party handling of client or donor records should keep that data out until their own legal review is done. Teams in the middle of a peak trading season should not be moving their operating system that week. Both cases are covered in detail on the relevant pages. **Do we have to leave Slack, Notion or Linear to use Polaris?** No. Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp and Instagram are all in the connection catalog, so a worker can be given access to the tools you already run. Plenty of teams will connect rather than migrate, at least at first. **What is the first thing to actually do?** Sign in with an emailed code, tell the Chief of Staff what keeps slipping, and let it run the hiring interview. That produces a worker with a readable skill file in about a minute. Assign it one real task from this week and judge the delivery, because the delivery is the product. ## Related - https://www.polarishq.co/for/startups - https://www.polarishq.co/for/agencies - https://www.polarishq.co/for/remote-teams - https://www.polarishq.co/for/technical-founders - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/ai-workers - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/use-cases --- --- title: "Polaris for startups: output before you can afford payroll" description: "Startups run out of runway before they run out of work. Polaris keeps the workspace free at any headcount and bills only for hours an AI worker delivers." url: https://www.polarishq.co/for/startups section: Solutions updated: 2026-08-21 --- # Polaris for startups Your next hire takes about sixty seconds and costs nothing in the weeks it has nothing to do. ## The short answer Polaris gives a startup output without payroll. The workspace is free for unlimited people, so tasks, docs and team chat cost nothing at any headcount. AI workers are hired in chat in about sixty seconds, get an editable skill file, and bill roughly two dollars per human-hour they deliver. Nothing delivered, nothing billed, which keeps the cost variable while runway is the binding constraint. - **Software cost:** $0 at any headcount - **Time to hire a worker:** ~60 seconds - **Cost of an idle worker:** $0 - **Billing unit:** ~$2 per human-hour delivered ## The arithmetic that breaks first A startup does not fail because it lacks a project management tool. It fails because the list of things that must happen this quarter is four times longer than the number of people available to do them, and both available answers are slow. Raising takes months. Hiring takes weeks to source, weeks to close, weeks to become useful, and the cost starts on day one regardless. In the meantime the company is paying per seat for three or four products that hold work rather than do it. A tracker, a doc tool, a chat app, and by now a stack of individual AI subscriptions on personal cards that nobody can see or assign anything to. That is fixed cost buying storage, at exactly the stage where every fixed cost should be a variable one. ## What the cost line looks like instead Three numbers, all of them checkable inside the product. - **$0** — Workspace, forever. Unlimited humans, tasks, workstreams and docs. No seat count, no tier. - **~60s** — To hire an AI worker. A short interview in chat, then a worker card with a readable SKILL.md. - **~$2** — Per human-hour delivered. Estimated by an open formula, logged job by job, challengeable line by line. ## The first three workers most startups need Not the ones that sound impressive. The ones covering the work a two-person founding team keeps postponing. - **Research analyst** — The competitor teardown, the market sizing, the pricing scan that keeps getting pushed to next week because it takes a full day and nobody has one. It runs a live tool loop with real web search and delivers the finding as a comment with the file attached. - **Content writer** — Landing copy, the changelog nobody has written since March, the launch post, the docs page a customer asked for twice. Give it the skill file that carries your voice rules and it stops being a rewrite job. - **SDR** — First-touch outbound before you can justify a sales hire. Account research, list building and drafting, with a human closing every task, because agents deliver and people decide what actually gets sent. ## Two ways to add capacity in month four **Hire a person** - Weeks of sourcing before a first conversation - Salary starts on day one and does not pause in a slow month - Onboarding lives in someone's head and leaves when they do - Wrong hire is expensive to unwind and hard to talk about **Hire an AI worker** - A short interview in chat, every answer a click, roughly a minute - Costs nothing in a week where you assign it nothing - Its instructions are a SKILL.md file you can read, edit and reuse - Wrong brief is a file edit, not a conversation with a lawyer > **Where this is a bad idea right now** > > Polaris is a free public beta with no users behind it. Do not make it the only home of the one process your revenue depends on in week one. Put a real but survivable workstream on it first, judge two or three deliveries, and expand from evidence rather than from a page like this one. ## Next, for a startup - [cost/stack-cost-5-person-team](https://www.polarishq.co/cost/stack-cost-5-person-team) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [ai-workers/content-writer](https://www.polarishq.co/ai-workers/content-writer) - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) ## Questions people ask **We are three people. Is Polaris overkill?** The workspace costs nothing at three people, so the question is not price, it is whether you need a tracker, docs and team chat in one place yet. If your work currently lives in a shared note and one Slack channel, the honest reason to start is the AI workers, not the project management. Hire one worker, assign it something real, and let the workspace fill up around it. **How do we know the hours on the bill are not inflated?** Human-equivalent hours are estimated by an open formula from observable effort: base pickup time, searches at roughly twelve minutes each, finished prose at about ninety characters per minute, plus fixed overheads for checklist items, comments addressed and files produced, clamped between five minutes and eight hours per session. Every job is logged on the worker's work log, and you can challenge any line from the log itself. **What happens to our bill in a quiet month?** It goes to zero. The software is free with no seats, and billing only starts when an AI worker delivers something. A month where you assign nothing costs nothing, which is the opposite of how a per-seat stack behaves when a founder goes heads-down on product. **Does this replace a first marketing or ops hire?** It changes when you need one, not whether. A worker will produce the research, the drafts and the recurring reporting, but a human still owns the judgement, closes the task and rates the work. Treat it as buying back the hours that were going to be a contractor invoice, and hire the person when the judgement load, not the production load, is the thing overwhelming you. **Can our whole team use it if only one of us is technical?** Yes. Sign-in is passwordless, an emailed code with no password path anywhere in the product, and hiring a worker is a chat interview where every answer is a click. Nobody opens a terminal. The technical founder will care about the runtime and the skill files; everyone else will only ever see tasks, docs and chat. ## Related - https://www.polarishq.co/for/solo-founders - https://www.polarishq.co/for/technical-founders - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/sdr - https://www.polarishq.co/cost/stack-cost-5-person-team - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent --- --- title: "Polaris for agencies: an auditable cost per delivered hour" description: "Agencies resell hours. Polaris puts an itemised, auditable cost per human-hour delivered underneath the rate card, and gives each client its own workstream." url: https://www.polarishq.co/for/agencies section: Solutions updated: 2026-08-21 --- # Polaris for agencies You already know what an hour of delivery sells for. This is a page about what it costs. ## The short answer Agencies bill clients by the hour or by the retained month, so anything that changes the cost of an hour changes the margin directly. Polaris bills roughly two dollars per human-equivalent hour an AI worker delivers, itemised on a work log you can challenge line by line. The workspace itself is free, with a workstream per client and no per-seat charge for freelancers. - **Billing unit:** ~$2 per human-hour delivered - **Evidence:** Every hour itemised on the work log - **Client containers:** Unlimited workstreams - **Freelancer seats:** $0 ## The only number that moves an agency Every agency runs the same equation. Blended cost per delivery hour on one side, blended bill rate on the other, utilisation deciding whether the difference survives contact with reality. Software rarely touches that equation. A new tracker moves admin time around; it does not change what an hour of research or a first draft costs to produce. A metered agent hour does touch it, and it touches it in the least glamorous part of the work: the desk research under a strategy deck, the first pass at twelve pages of copy, the competitor scan that a junior spends a day on, the weekly client status write-up that is pure cost and nobody enjoys. That layer is where agency margin quietly leaks, and it is exactly the layer a briefed worker can return finished as a comment with the file attached. ## Where the hours currently go The unglamorous half of a retainer, and what changes when a worker takes the first pass. | Recurring job | How it gets done now | With a briefed worker | | --- | --- | --- | | Desk research under a strategy deck | A junior spends most of a day, billed or absorbed | A research analyst runs a live search loop and returns sourced findings as a file | | First draft of campaign or site copy | A writer starts from nothing, or a senior rewrites a thin draft | A content writer works from a SKILL.md carrying the client's voice rules | | Weekly client status write-up | Unbillable, always late, assembled from three tools | Drafted from the workstream, reviewed and closed by the account lead | | Competitor and category monitoring | Promised in the pitch, dropped by month three | A standing task on a schedule you control, logged every time it runs | | Tool seats for freelancers and contractors | Per seat, per tool, per month, whether or not they log in | No seat count. Add them to the workstream at no cost | ## How an agency shapes a client engagement Workstreams are just containers, so the structure mirrors how you already run accounts. 1. **One workstream per client** — A bucket holds everything for that account: tasks, docs, files and the conversation. Lanes are shared between list and board view, so the account lead and the delivery team organise once and look at it however they prefer. 2. **Lanes that match your delivery stages** — Scoping, in production, in client review, shipped. Everything drags. The Focus lane sits pinned on top with the non-negotiables for today, this week and the next thirty days across every account you run. 3. **Hire the workers the account actually needs** — Tell the Chief of Staff what keeps slipping on that account. The interview takes about a minute and produces a worker with an editable skill file, so a client with a strict tone of voice gets a writer briefed on that tone rather than a generic one. 4. **Assign, then read the work log** — A real cloud machine wakes for the task, runs its tool loop, ticks its own acceptance criteria and posts the deliverable as a comment. The estimated hours land on the work log alongside the output, so the cost of that deliverable is visible at the moment you review it. 5. **A human closes and rates it** — The machine never marks a task done. Your account lead closes it, which keeps the quality gate exactly where your client contract assumes it is: with a named person at the agency. ## The workers an agency hires first Roles that map onto billable production, not onto internal admin. - **Research analyst** — The desk research floor under pitches, audits and strategy work, with live web search and sources attached to the delivery. - **Content writer** — First drafts at volume, briefed per client through a skill file rather than through a fresh brief every time. - **SEO specialist** — Keyword and intent work, on-page recommendations and the content briefs your writers are waiting on. - **Social media manager** — The scheduling and drafting load on accounts where the retainer promised a cadence that is expensive to keep by hand. > **The client contract question, answered honestly** > > Connections are authorised once, org-wide, and credentials are stored server-side. That is convenient and it is also a fact your client MSAs may have an opinion about. If an agreement restricts third-party processing of client material, keep that account out of Polaris until your own review clears it, and start with an internal workstream instead. Do not discover this in a renewal conversation. ## Next, for an agency - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [use-cases/marketing/seo-content-production](https://www.polarishq.co/use-cases/marketing/seo-content-production) - [use-cases/sales/proposal-writing](https://www.polarishq.co/use-cases/sales/proposal-writing) - [for/consultancies](https://www.polarishq.co/for/consultancies) ## Questions people ask **Can we bill agent hours through to the client?** That is a commercial decision inside your contract, not something the product decides for you. What the product gives you is the evidence: an itemised work log with an estimated human-equivalent hour count per job, produced by a formula you can read. Some agencies will pass it through at a markup, some will keep it as margin on a fixed retainer, and the log supports either conversation. **How is the hour estimated, exactly?** From observable effort rather than a wall clock: base pickup time, searches at roughly twelve minutes each, finished prose at about ninety characters per minute, plus fixed overheads for checklist items ticked, comments addressed and files produced. The result is clamped between five minutes and eight hours per session and logged per job. If a line looks wrong, you challenge it from the log. **Does this replace juniors?** It replaces the part of a junior's week that is a research floor and a first pass, which is the part juniors learn least from and clients value least. Judgement, client relationships, creative direction and quality control stay with people, and a human closes every task. If your model depends on billing juniors' desk research at a senior rate, this changes your economics in a way worth thinking about before you adopt it. **We run fifteen client accounts. Does the workspace handle that?** Workstreams are unlimited and free, so fifteen accounts is fifteen buckets with their own lanes, docs and conversation. The Focus lane cuts across all of them, which is the view most account directors are missing today: the non-negotiables for the week regardless of which client they belong to. **What does it cost to add a freelancer for one project?** Nothing. There is no seat count and no tier, so contractors, freelancers and client-side collaborators can sit in a workstream at no cost. This is the point where a per-seat stack usually punishes agencies hardest, because the people who need access for six weeks cost the same as the people who need it all year. **Is Polaris ready for a client-facing agency today?** It is a free public beta with no agencies using it yet, and it would be dishonest to pretend otherwise. The sensible first move is one internal workstream, such as your own new business research or your own content pipeline, where a bad delivery costs you an afternoon rather than a client relationship. ## Related - https://www.polarishq.co/for/consultancies - https://www.polarishq.co/for/design-studios - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/use-cases/marketing/seo-content-production - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/stack-cost-10-person-team --- --- title: "Polaris for solo founders: the colleague you cannot hire yet" description: "A one-person company has no one to delegate to and no one to be accountable to. Polaris gives a solo founder a roster, a work log and a Chief of Staff." url: https://www.polarishq.co/for/solo-founders section: Solutions updated: 2026-08-21 --- # Polaris for solo founders Being the whole org chart is not a time problem. It is an accountability problem. ## The short answer A solo founder has no colleague to delegate to and nobody who notices when something slips. Polaris ships with a Chief of Staff on the roster from the first sign-in that keeps every task owned and dated, builds workstreams when work needs a home, and hires the workers that are missing. AI workers deliver finished work as comments with files, billed only when they deliver. - **Seats required:** 1 - **Chief of Staff:** Included from first sign-in - **Bill in a week with no deliveries:** $0 ## The specific failure of working alone Solo founders are usually told they have a time problem, and time management is offered as the fix. The actual failure is quieter. Nobody is waiting on anything. A task with no owner other than you and no date other than soon does not get done, it gets reclassified as something you are thinking about. Six weeks later it is still an idea and the reason it never moved has become invisible. The second failure is context. Working alone means every piece of knowledge lives in one head with no reason to write it down, so the moment you do bring someone in, whether a contractor, a co-founder or an assistant, you discover that explaining your business takes three days you do not have. ## The four things that quietly fall over when you are one person - **Anything with no external deadline** — The pricing page rewrite, the accounting cleanup, the customer interviews. No client is chasing them, so they lose every week to whatever is loudest. - **The research floor under a decision** — You know you should compare five options properly. You look at two, then decide on instinct, then live with it for a year. - **Anything that needs three uninterrupted hours** — A solo day is fragmented by definition. Work that requires a long block gets attempted at eleven at night and abandoned. - **Writing down how anything works** — There is no audience for a process document when you are the only person who runs the process, right up until the moment there is. ## What a solo founder's Monday looks like in Polaris The point is not more structure. It is that something other than you is holding the structure. 1. **Open Focus, not a project list** — Three horizons: today, this week, the next thirty days. The Focus lane is pinned first in every view, so the non-negotiables are the first thing you see rather than a backlog that makes you feel behind. 2. **Tell the Chief of Staff what is slipping** — In chat, in plain language. It keeps every task owned and dated, builds a workstream when a pile of work clearly needs a home, and tells you which worker you are missing rather than waiting for you to work it out. 3. **Hand off the long block** — The three-hour research task goes to a worker instead of to eleven at night. A cloud machine wakes for it, runs live web search, ticks its acceptance criteria and posts the result as a comment on the task. 4. **Close the loop in the evening** — You read the delivery, close the task and rate it. The machine never marks its own work done, which means the one quality gate in a one-person company stays where it belongs. > **The side effect that matters more than the hours** > > Every worker you hire produces a SKILL.md file describing what it should be great at and how it should work. After a few months of this, the thing you could never write down, how your business actually operates, exists as readable files. That is the asset a first hire or a co-founder walks into, and you built it without ever sitting down to write documentation. ## Where it is genuinely not the answer If your bottleneck is that you have not decided what to build, no roster helps. Delegating research is a way to postpone a decision as easily as it is a way to inform one, and a solo founder is the person least protected from that. It is also a free public beta with no users. As a single operator you have no colleague to catch a bad delivery for you, so keep a human check on anything that goes to a customer, and start with work where the worst case is a wasted afternoon. ## Next, for a solo founder - [ai-workers/executive-assistant](https://www.polarishq.co/ai-workers/executive-assistant) - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [glossary/focus-lane](https://www.polarishq.co/glossary/focus-lane) - [use-cases/executive/meeting-follow-ups](https://www.polarishq.co/use-cases/executive/meeting-follow-ups) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) ## Questions people ask **What is the Chief of Staff, in practical terms?** It is an AI worker that is on your roster in every organisation from the first sign-in, so you never start with an empty team. Its job is the coordination layer: keeping tasks owned and dated, creating workstreams when work needs a container, turning signals from your connected tools into task suggestions, and hiring the workers you are missing when you describe a gap. **I already use Todoist or Notion alone. Why change?** For the list itself, you probably should not. A personal task app is fine at holding a list. The reason to move is that a list cannot do any of the work, cannot notice that an item has sat untouched for three weeks, and cannot be assigned a four-hour research job overnight. If you only want a better list, keep the one you have. **How much would a solo founder actually spend per month?** It depends entirely on how much work you assign, because the software is free and there is no seat charge. Billing is roughly two dollars per human-equivalent hour delivered, so a month where you hand over the equivalent of a day of research and drafting is a small number, and a month where you assign nothing is zero. **Can I use it from my phone?** Yes. There is a Flutter mobile app with the copilot at the centre, including voice input and spoken replies, and it is live as an installable PWA. For a solo founder that mostly matters for the gap between having a thought and it becoming an owned, dated task, which is where most solo work dies. **What happens when I do hire my first employee?** They join the same workspace at no additional cost, because there is no seat count. They also inherit the workstreams, the docs, the task history and the skill files, which is a meaningfully better first week than the usual arrangement where the founder explains everything twice from memory. ## Related - https://www.polarishq.co/for/startups - https://www.polarishq.co/for/technical-founders - https://www.polarishq.co/for/small-business - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "Polaris for remote teams: decisions that survive timezones" description: "Distributed teams lose decisions to Slack scrollback and hours to the timezone gap. Polaris turns signals into approved tasks and keeps working overnight." url: https://www.polarishq.co/for/remote-teams section: Solutions updated: 2026-08-21 --- # Polaris for remote teams The problem was never the meetings. It is what happens to a decision made while half the team is asleep. ## The short answer Remote teams lose decisions in chat scrollback and lose hours in the gap between timezones. Polaris catches what your tools hear in an Inbox, where a Slack message arrives as a prefilled task suggestion with bucket, lane, labels and owner already set, waiting for one click. Assigned AI work runs on cloud machines overnight and lands as a comment with files, timestamped on a work log. - **Slack signals:** Arrive as prefilled task suggestions - **Silent tasks created:** None. Every suggestion needs a click - **Overnight work:** Runs on a cloud machine - **Audit trail:** Every job on the work log ## What actually goes wrong across timezones The standard account of remote work blames meetings and calendars. That is not what breaks. What breaks is that a decision gets made in a thread at four in the afternoon in Lisbon, three people react to it, and it is never written anywhere with an owner and a date. Berlin acts on it. Austin never sees it. Two weeks later someone rebuilds the same thing and the disagreement is unrecoverable because the evidence is nine hundred messages up. The second failure is dead time. In a team spread across eight hours of offset, roughly a third of every day is a handoff window where nothing progresses because the person who could progress it is asleep. Most remote tooling tries to fix this by making the asleep person catch up faster. A cloud machine that keeps working is a different answer to the same hours. ## The same decision, two systems **In a chat thread** - Owner is implied by whoever replied last - The date is soon, which is not a date - Reconstructing it later means scrolling and guessing - Anyone offline when it happened learns about it by accident **In the Polaris Inbox** - The message arrives as a suggestion with bucket, lane, labels and owner prefilled - One click turns it into a real task in the right workstream - Nothing is created silently, so the Inbox never becomes another backlog - The task carries its own history for whoever wakes up next ## A follow-the-sun day, concretely One task, three timezones, nobody waiting on a call. 1. **Morning in Europe: the signal arrives** — Something is agreed in Slack. It shows up in the Inbox as a prefilled suggestion. One click and it is a task in the right workstream, owned and dated, visible to everyone regardless of when they log in. 2. **Afternoon: the long part is assigned to a worker** — The research, the draft, the reconciliation, whatever needs uninterrupted hours goes to an AI worker rather than to a person who is about to close their laptop. 3. **Overnight: a machine works while nobody does** — A cloud machine wakes for the job, runs a live tool loop with real web search, ticks the acceptance criteria as it goes and posts progress. It does not stop because Europe went to bed. 4. **Morning in the Americas: the delivery is already there** — The output is a comment on the task, with files attached and an itemised work log next to it. Nobody had to be awake to see what happened, which is the actual definition of async that works. 5. **A human closes it** — Agents deliver, humans close. Whoever owns the task reviews and rates it, so the decision to accept work stays with a person in a named timezone. ## What distributed teams stop doing - **The synchronous handoff call** — Twenty minutes at the one hour that works for everyone, to explain what a task is. The task carries its own context, its history and its files. - **Copying the same decision into three tools** — Tasks, docs and team chat are one product, so a decision does not need to be minuted in a doc, ticketed in a tracker and announced in a channel. - **Reconstructing who did what** — Every AI job is logged with its estimated hours and its output attached, which is the record that used to exist only as a memory of a standup. > **Do not switch chat tools on a distributed team overnight** > > Polaris includes team chat, but a remote team's chat history is its institutional memory and moving it is a genuine cost. Slack is in the connection catalog, so the honest first move is to connect it, let signals arrive in the Inbox as suggestions, and leave the conversation where it is. Decide about replacing chat later, from experience rather than from a migration plan. ## Next, for a remote team - [integrations/slack](https://www.polarishq.co/integrations/slack) - [cloud-claude-code/claude-code-when-laptop-is-closed](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [use-cases/operations/cross-team-coordination](https://www.polarishq.co/use-cases/operations/cross-team-coordination) - [replace/notion-and-slack](https://www.polarishq.co/replace/notion-and-slack) ## Questions people ask **Does the Inbox create tasks automatically from Slack?** No, and that is deliberate. A signal arrives as a suggestion with the bucket, lane, labels and owner already filled in, and it becomes a task when a person clicks. Automatic task creation from chat produces a second inbox that everyone learns to ignore within a fortnight, which is worse than the problem it was meant to solve. **How does an overnight AI job avoid becoming a surprise in the morning?** It posts progress as it works and ticks its own acceptance-criteria checklist, so the state is visible rather than a black box that resolves at dawn. The deliverable arrives as a comment on the task with any generated files attached, and the machine never marks the task done. A human closes it. **We are across eight timezones. Does anything need to happen synchronously?** Closing a task does, in the sense that a person has to do it, but they can do it whenever they are awake. Hiring a worker is a chat interview that takes about a minute and one person can run it for the whole organisation, because connections are authorised once and shared org-wide rather than per person. **What about the team members who are not technical?** Sign-in is passwordless, an emailed code with no password path in the product, and there is a mobile app with voice input and spoken replies that is live as an installable PWA. Nobody on a remote team needs a terminal or a local setup, which matters more when you cannot walk over and fix someone's environment. **Is there a real audit trail, or just a feed?** Every AI job is logged with its estimated human-equivalent hours, the output it produced and the files it generated. That log is also the billing record, so it is not decorative. For a distributed team it doubles as the answer to what happened while you were offline. ## Related - https://www.polarishq.co/integrations/slack - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/use-cases/operations/cross-team-coordination - https://www.polarishq.co/use-cases/executive/meeting-follow-ups - https://www.polarishq.co/ai-workers/project-coordinator - https://www.polarishq.co/replace/notion-and-slack --- --- title: "Polaris for software teams: the work that is not the code" description: "Engineering teams are good at shipping code and bad at everything around it. Polaris puts AI workers on triage, release notes, postmortems and doc drift." url: https://www.polarishq.co/for/software-teams section: Solutions updated: 2026-08-21 --- # Polaris for software teams Nobody joined your team to write the release notes. Something still has to write them. ## The short answer Software teams handle code well and handle the work around code badly: bug triage, release notes, incident write-ups, documentation drift and sprint admin. Polaris puts AI workers on that layer, assigned exactly like human teammates because humans and agents share one members table. GitHub, Linear and Slack are in the connection catalog, and the workspace is free for unlimited engineers. - **Humans and agents:** One members table - **Connections:** GitHub, Linear, Slack, Notion - **Cost per engineer:** $0 - **Billing unit:** ~$2 per human-hour delivered ## The half of the sprint nobody optimises Engineering organisations have spent fifteen years getting good at the code path. Review, CI, deploys, rollbacks, on-call rotations. All of it is instrumented and most of it is automated. Then there is the other half of the week, which is entirely manual and universally resented: triaging the issues that came in overnight, writing the release notes, turning an incident into a document somebody will read, updating the runbook that stopped matching production in April. This work does not get automated because it is not mechanical. It needs judgement and it needs prose, which is why it lands on whoever is least able to say no. It is also almost exactly what a briefed AI worker with tool access can take a first pass at, and where a first pass removes most of the cost. ## The five jobs to hand over first Ranked by how much engineering time they consume against how little engineering judgement they need. | Job | Where it lands today | What a worker returns | | --- | --- | --- | | Overnight bug triage | The engineer who opens the tracker first, before their own work | Sorted, labelled and duplicated-flagged, posted as a comment for a human to confirm | | Release notes | Written at the end of the day of the release, or not at all | A draft assembled from the shipped work, ready to edit rather than to start | | Incident postmortems | Promised in the retro, written three weeks later from memory | A structured write-up drafted while the timeline is still recoverable | | Documentation drift | Nobody, until a new joiner follows the runbook into a wall | A standing task that reads the docs against the current state and reports the gaps | | Sprint and board admin | The tech lead, in the hour before planning | Tasks kept owned and dated by the Chief of Staff, with suggestions rather than silent edits | ## Why the assignment model matters to engineers specifically Most AI tooling sits beside your process. This sits inside it. - **Humans and agents are the same table** — Members carry a kind of human or agent, and assignment works identically for both. There is no separate AI panel to check, no parallel queue, no second place where work might be. - **Capabilities are a file, not a hidden prompt** — A worker's skills are a SKILL.md you can read, edit and review like any other file in the team. Engineers are the audience most likely to distrust a black box and most able to fix a file. - **A real machine, not a chat window** — A cloud machine wakes per task, runs a live tool loop and delivers the work as a comment with files. It keeps running after the person who assigned it closes their laptop. - **Connect rather than migrate** — GitHub, Linear, Slack and Notion are all in the connection catalog, authorised once and stored server-side. You do not have to leave your tracker to put a worker next to it. ## What Polaris is not, for an engineering team It is not an IDE and it does not sit in your editor. It does not review pull requests inline, it is not a replacement for your CI, and it will not do anything to your repository that you have not connected and asked for. The delivery model is a comment on a task with files attached, which is deliberately outside the code path. It is also a free public beta. No engineering team is running on it yet. Start with the layer where a bad first draft costs an afternoon, which is the release notes and the doc audit, not the incident timeline your compliance team relies on. ## Next, for a software team - [use-cases/engineering](https://www.polarishq.co/use-cases/engineering) - [use-cases/engineering/bug-triage](https://www.polarishq.co/use-cases/engineering/bug-triage) - [use-cases/engineering/incident-postmortems](https://www.polarishq.co/use-cases/engineering/incident-postmortems) - [integrations/github](https://www.polarishq.co/integrations/github) - [integrations/linear](https://www.polarishq.co/integrations/linear) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) ## Questions people ask **Do we have to leave Linear or Jira?** No. Linear is in the connection catalog and GitHub is too, so a worker can be given access to the tracker you already run. Whether you eventually move the tracking itself is a separate decision, and the honest first step is putting a worker next to your existing process rather than replacing it. **Is this a coding agent?** Not in the sense of an autonomous pull-request bot. Workers run on cloud machines with a live tool loop and real web search, and they deliver work as comments and files. That covers the research, prose and coordination layer around engineering well. If what you want is inline code review in your editor, that is a different category of tool and you should keep using one. **How do we stop an AI worker from closing its own tickets?** You cannot make it do that, because the product does not allow it. The machine ticks its own acceptance criteria and posts the delivery, but it never marks a task done. A human closes it and rates it, which keeps the completion signal on your board meaningful. **What does it cost for a twenty-engineer team?** The workspace is zero, because there is no seat count and no tier, so twenty engineers, unlimited tasks, workstreams and docs cost nothing. The bill is roughly two dollars per human-equivalent hour that a worker delivers, itemised on the work log, so it scales with work assigned rather than with headcount. **Where is our data and what runs the workers?** Supabase provides Postgres, auth, realtime, row-level security and edge functions, and it is the API that both the frontend and the workers talk to. A worker runtime on Fly.io claims jobs off a queue. The queue contract is runtime-agnostic, so the machine behind it is swappable. **Can a worker be given our credentials safely?** Connections are authorised once, org-wide, verified live and stored server-side. Workers use them; browsers cannot read them back. That is the design, and it is worth reviewing against your own security posture before you connect anything with production access. ## Related - https://www.polarishq.co/for/technical-founders - https://www.polarishq.co/for/product-teams - https://www.polarishq.co/use-cases/engineering - https://www.polarishq.co/use-cases/engineering/bug-triage - https://www.polarishq.co/use-cases/engineering/technical-documentation - https://www.polarishq.co/integrations/github - https://www.polarishq.co/integrations/linear - https://www.polarishq.co/cost/stack-cost-25-person-team --- --- title: "Polaris for consultancies: delegate the research floor" description: "Every consulting engagement starts with the same research floor, paid for in senior time. Polaris moves that layer to metered AI workers who deliver files." url: https://www.polarishq.co/for/consultancies section: Solutions updated: 2026-08-21 --- # Polaris for consultancies The scarce resource in a consultancy is not hours. It is partner attention, spent on work a partner should not be doing. ## The short answer Consultancies sell judgement, but every engagement begins with a research and synthesis floor that consumes senior time before any judgement is possible. Polaris moves that layer to AI workers that run on cloud machines, use live web search, and deliver the output as files attached to a task. Hours are estimated by an open formula and logged, so the cost of the floor becomes a visible line. - **Deliverables:** Arrive as files on the task - **Hours:** Open formula, itemised per job - **Cost per consultant seat:** $0 ## The fixed tax on every engagement A consultancy prices its work on expertise, then spends the first two weeks of every engagement on something that is not expertise. Reading the client's market. Assembling the comparable set. Pulling the regulatory background. Turning nineteen interviews into four themes. Building the same category map that was built for a similar client eighteen months ago and cannot be found now. This floor is unavoidable, which is exactly why it deserves attention. It is the largest block of hours in the engagement that does not depend on who is doing it, and in most firms it is done by the people whose time is either the most expensive or the most developmentally valuable. Neither is a good outcome. ## The four costs hiding inside the floor - **Senior time spent reading, not deciding** — The partner or principal reads to form a view. Half of that reading is orientation any competent researcher could have compressed into a briefing note. - **Work already done, done again** — Firms rebuild the same market maps because the previous version lives in a deck on someone's laptop rather than in a place with a search box. - **Synthesis that arrives too late to change the answer** — Interview themes surface in week four. The hypothesis was set in week one and the evidence is now decoration. - **The proposal that costs a week to lose** — Every unsuccessful pitch consumed real research hours that nobody billed and nobody logged. ## Running a diagnostic engagement in Polaris The structure is ordinary. The difference is who does the floor. 1. **The engagement becomes a workstream** — One bucket per engagement, holding tasks, docs, files and the conversation. Docs are a nested tree with a block editor, versioned files, review and comments, so the interim work products live where the work lives. 2. **Hire a research analyst briefed on this client** — The Chief of Staff runs a short interview and the answers become a readable SKILL.md. That file is the difference between a generic researcher and one that knows this client's sector, vocabulary and evidence standard. 3. **Assign the floor as discrete tasks** — Comparable set, regulatory background, category map, competitor positions. Each is a task with acceptance criteria that the worker ticks as it goes, so a half-finished piece of research is visibly half-finished. 4. **Deliverables arrive as files** — The machine posts its output as a comment on the task, including generated files such as PDFs and documents. The consultant reads a document rather than a chat transcript. 5. **The principal spends their time on the view** — A human closes and rates every task. The judgement, the client relationship and the recommendation stay exactly where the fee assumes they are. ## Where a principal's week goes **Today** - Orientation reading that could have been a briefing note - Rebuilding an analysis the firm has done before - Chasing an interim deliverable from three people - Unbilled research on a proposal that may not land **With the floor delegated** - Reading a sourced briefing note produced overnight - Reusing a skill file that carries the firm's method - Watching acceptance criteria tick on the task itself - A logged, itemised hour count against every proposal > **Check your engagement letters before you connect anything** > > Consulting agreements frequently restrict where client material may be processed and who may see it. Connections in Polaris are authorised once, org-wide, with credentials stored server-side. That is a real convenience and a real question for your risk function. Start with your own new-business research, where the material is yours, and take client data through your normal review before it goes anywhere near a worker. ## Next, for a consultancy - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [ai-workers/competitive-analyst](https://www.polarishq.co/ai-workers/competitive-analyst) - [use-cases/executive/strategic-research](https://www.polarishq.co/use-cases/executive/strategic-research) - [use-cases/sales/proposal-writing](https://www.polarishq.co/use-cases/sales/proposal-writing) - [cloud-claude-code/ai-agents-that-produce-files](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) ## Questions people ask **How is this different from the agencies argument?** An agency's constraint is the blended cost of a delivery hour across a team. A consultancy's constraint is senior attention, which cannot be blended, hired quickly or scaled. The agency case is about margin per hour. This one is about what the most expensive person in the building spends their week reading. **Can a worker be trusted with primary research quality?** Treat it as a first pass with sources attached, not as a finished view. Workers run a live tool loop with real web search and tick their own acceptance criteria, and the delivery arrives as a file you read like any junior's work product. A human closes every task, so nothing reaches a client without a consultant having accepted it. **What stops us rebuilding the same analysis for the next client?** The skill file. A worker's capabilities are a SKILL.md you can read, edit and reuse, so the method your firm applies to a market scan becomes a durable artifact rather than a habit in a partner's head. Docs are versioned and nested, so the outputs stay findable in the workstream rather than in an attachment. **Does the hour estimate hold up under scrutiny from a finance team?** It is derived from observable effort: base pickup time, searches at roughly twelve minutes each, finished prose at about ninety characters per minute, and fixed overheads for checklist items, comments addressed and files produced, clamped between five minutes and eight hours per session. The formula is open and every job is logged, so any line can be challenged from the log itself. **Is a small consultancy the right size for this?** Boutique firms feel the argument hardest, because the research floor lands on partners directly with no junior bench to absorb it. Larger firms have that bench, so the case becomes an economic one about what that bench costs rather than a capacity one. Both work; the boutique version is sharper. ## Related - https://www.polarishq.co/for/agencies - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/use-cases/executive/strategic-research - https://www.polarishq.co/use-cases/sales/proposal-writing - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files - https://www.polarishq.co/cost/per-seat-vs-usage-pricing --- --- title: "Polaris for ecommerce: capacity that follows the calendar" description: "Ecommerce work is seasonal but headcount is not. Polaris keeps the workspace free year-round and bills only for the hours AI workers actually deliver." url: https://www.polarishq.co/for/ecommerce-brands section: Solutions updated: 2026-08-21 --- # Polaris for ecommerce brands You staff for peak and pay for it in February, or you staff for February and suffer in November. ## The short answer Ecommerce workload is lumpy: product launches, seasonal campaigns and support volume that multiplies for six weeks and then collapses. Fixed headcount and per-seat software do not follow that curve. Polaris keeps the workspace free year-round and bills roughly two dollars per human-hour an AI worker delivers, so a quiet month costs nothing and a peak month costs what the work was worth. - **Cost in a quiet month:** $0 if nothing is delivered - **Connections:** Stripe, Instagram, Gmail, Slack - **Workspace:** $0, unlimited seasonal staff - **Billing unit:** ~$2 per human-hour delivered ## The staffing curve nobody can get right Retail workload is not a flat line and it never has been. A drop, a seasonal campaign, a marketplace event and the six weeks before Christmas each produce two or three times the normal volume of copy, imagery coordination, customer messages and reporting, and then it stops. Everyone in the category runs the same two bad options: hire for the peak and carry the cost through the flat months, or hire for the flat months and burn the team out twice a year. Agencies and seasonal contractors exist because of this, and they carry their own problems: a ramp-up cost every time, a minimum retainer that outlives the need, and a per-seat charge on four tools for someone who is here for nine weeks. ## How the cost line behaves across a trading year Same workspace, same roster, different amount of assigned work. | Month type | What the work looks like | What the bill does | | --- | --- | --- | | Flat trading month | Ordinary content cadence, normal support volume, routine reporting | Software is free. Billing only reflects work actually delivered | | Launch or drop | Product copy, campaign assets coordination, launch comms, a spike in questions | Rises with assigned work, itemised per job on the work log | | Peak season | Support volume multiplies, daily social, daily performance reporting | Rises further, still with no seat charge for temporary staff | | The month after peak | Post-mortem reporting, returns handling, planning next year | Falls immediately, because nothing is billed for capacity you are not using | ## The workers an ecommerce brand hires first The four roles that carry most of the seasonal load. - **Content writer** — Product descriptions at volume, campaign copy, email and landing page drafts, all briefed through a skill file carrying the brand voice rather than a fresh brief per drop. - **Support specialist** — The where-is-my-order tier that triples in December. Triage and drafted responses, with a person closing every task before anything is treated as resolved. - **Social media manager** — The daily cadence a small brand promises itself and abandons by week three, with Instagram in the connection catalog. - **Competitive analyst** — Pricing and promotion monitoring across the category during the weeks when a competitor's discount actually changes your day. > **Do not do this in November** > > Changing how your team coordinates during peak trading is a genuinely bad idea, and Polaris is a free public beta with no ecommerce brands running on it yet. The honest advice is to pick one workstream in a flat month, such as product copy for a small collection or the social calendar, run it end to end, and decide about peak season from evidence rather than optimism. ## Next, for an ecommerce brand - [integrations/stripe](https://www.polarishq.co/integrations/stripe) - [integrations/instagram](https://www.polarishq.co/integrations/instagram) - [ai-workers/support-specialist](https://www.polarishq.co/ai-workers/support-specialist) - [ai-workers/social-media-manager](https://www.polarishq.co/ai-workers/social-media-manager) - [use-cases/customer-support/ticket-triage](https://www.polarishq.co/use-cases/customer-support/ticket-triage) - [use-cases/marketing/social-media-scheduling](https://www.polarishq.co/use-cases/marketing/social-media-scheduling) ## Questions people ask **Can a worker answer customer emails directly?** It can triage and draft, and Gmail is in the connection catalog, but the completion signal stays with a person. The machine never marks a task done. For customer-facing messages during peak that is the correct arrangement anyway, because the cost of a wrong automated reply to an anxious customer in December is much higher than the minute it takes to approve one. **Does it connect to Shopify?** Not today. The connection catalog covers Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web research. Stripe is there, so payments data has a path, but if your operational truth lives in a store platform, expect to bring context to a worker manually for now. **We already pay an agency for content. Why change?** You may not need to. The realistic comparison is a retainer with a monthly minimum against a metered line that goes to zero in a flat month. If your agency relationship is mostly creative direction, keep it. If a large part of it is volume production of product and campaign copy, that is the part worth testing on one collection. **How many people can we put in the workspace during peak?** As many as you like. There is no seat count and no tier, so seasonal staff, warehouse coordinators and temporary support agents cost nothing to add and nothing to remove afterwards. This is the specific place per-seat pricing punishes a seasonal business twice, once when you add people and again when you forget to remove them. **What does peak actually cost, roughly?** It depends entirely on how many hours of work you assign, because there is no fixed component at all. Billing is roughly two dollars per human-equivalent hour delivered, estimated from observable effort by an open formula and itemised per job, so the number moves with the work rather than with the calendar or the headcount. ## Related - https://www.polarishq.co/integrations/stripe - https://www.polarishq.co/integrations/instagram - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/ai-workers/social-media-manager - https://www.polarishq.co/use-cases/customer-support/ticket-triage - https://www.polarishq.co/use-cases/marketing/social-media-scheduling - https://www.polarishq.co/cost/stack-cost-10-person-team --- --- title: "Polaris for nonprofits: no seat charge for a volunteer" description: "Per-seat software charges nonprofits for volunteers who log in twice a quarter. Polaris has no seat count, and bills only when an AI worker delivers work." url: https://www.polarishq.co/for/nonprofits section: Solutions updated: 2026-08-21 --- # Polaris for nonprofits Your org chart is six staff and forty people who help. Per-seat pricing was designed for the opposite shape. ## The short answer Nonprofits usually have a small paid staff surrounded by many occasional participants: volunteers, trustees, part-time coordinators and partner organisations. Per-seat software charges for all of them. Polaris has no seat count and no tier, so unlimited people share one free workspace with tasks, docs and team chat, and the only cost is roughly two dollars per human-hour an AI worker delivers. - **Cost per volunteer:** $0 - **People in one workspace:** Unlimited - **Billed when nothing is delivered:** $0 ## The seat tax on the shape you actually are Per-seat pricing assumes an organisation is a set of full-time employees who all use the software daily. A nonprofit is rarely that. It is four or six paid staff, a board that shows up quarterly, a rota of volunteers who need access for one campaign, a partner organisation running a joint programme, and a part-time bookkeeper. Every one of those people costs the same per month as a full-time employee in most tools. So the standard workaround is a shared login, or a spreadsheet, or keeping volunteers permanently outside the system and forwarding them things. All three cost more in coordination than the licence would have, and the last one is why volunteer effort so often gets wasted on work that was already done. ## A six-staff, forty-volunteer organisation **On per-seat tools** - Every occasional helper is a full monthly licence - Volunteers get excluded to control cost, then work in the dark - A second tool for docs doubles the count - Adding AI means another per-person add-on on top **On Polaris** - All forty-six people in one workspace at no cost - Tasks, docs and team chat included, with no tier to outgrow - A volunteer sees the workstream they are helping with, and its history - The only spend is work an AI worker actually delivered ## The work most worth handing to a worker Jobs that are essential, repetitive and chronically underserved in a small organisation. - **Funder and grant research** — The scan of who is funding what, with deadlines and eligibility, which everyone knows should be continuous and is instead done in a panic. A research analyst runs live web search and returns sourced findings as a file. - **Report drafting** — Funder reports, impact summaries and the board pack. First drafts assembled from the workstream, then edited by the person who actually knows the programme. - **Process documentation** — The volunteer handbook, the event runbook, the induction notes. Written once, versioned, and no longer resident in the head of the coordinator who left in June. - **Budget and spend reporting** — The recurring finance summary that a part-time bookkeeper produces at the end of a long week, drafted in advance for them to correct. > **Beneficiary and donor data comes last, not first** > > If you hold personal data about beneficiaries or donors, your obligations do not soften because a tool is free. Polaris is a public beta, connections are authorised org-wide and credentials sit server-side. Start with a workstream that holds no personal data at all, such as funder research or the volunteer handbook, and let your own governance process decide about anything else on its own timetable. ## What to try first, concretely Take one programme, put it in a workstream, and add the volunteers who work on it. That alone tests the thing per-seat pricing has been preventing: whether occasional helpers do better work when they can see the whole picture rather than receiving forwarded fragments. Then assign one funder research task to a worker and read what comes back. It costs roughly two dollars per human-equivalent hour of the work delivered, itemised on a log you can challenge, which is a defensible line in a budget in a way that a monthly licence for forty part-time people is not. ## Next, for a nonprofit - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [ai-workers/research-analyst](https://www.polarishq.co/ai-workers/research-analyst) - [ai-workers/content-writer](https://www.polarishq.co/ai-workers/content-writer) - [use-cases/finance/budget-reporting](https://www.polarishq.co/use-cases/finance/budget-reporting) - [use-cases/operations/process-documentation](https://www.polarishq.co/use-cases/operations/process-documentation) ## Questions people ask **Is there a nonprofit discount?** There is no discount because there is no price on the software. It is free with no seat count and no tier, for any organisation of any size, and that is not a charitable programme that could be withdrawn. Revenue comes only from delivered AI work at roughly two dollars per human-equivalent hour. **How is free sustainable? We have been burned by tools that changed pricing.** It is a fair question and the honest answer is the revenue mechanism, not a promise. Polaris makes money when AI workers deliver work, so the free workspace is the thing that makes the metered part possible rather than a loss leader waiting to be closed. That said, Polaris is an early-stage product in public beta, and no early-stage product can guarantee its own future. Keep your data exportable and judge accordingly. **Can volunteers use it without training?** Sign-in is passwordless, an emailed code with no password anywhere in the product, which removes the single largest source of volunteer support requests. The interface is tasks, docs and chat, and there is a mobile app with voice input live as an installable PWA. Nobody needs a terminal or a local setup. **What about our compliance requirements around processing?** Treat this as a real constraint rather than a formality. Connections are authorised once at organisation level with credentials stored server-side, and the product is in public beta. If you operate under a data protection agreement, a funder condition or a safeguarding policy that governs where records may be processed, run this through that process before any personal data goes in, and start with a workstream that has none. **We are two paid staff and a lot of goodwill. Is this too much tool?** Possibly, for the workspace. Two people can coordinate in a shared document for a long time. The part that is not too much tool is the worker: funder research and report drafting are jobs a two-person organisation genuinely cannot do properly, and that is worth testing on its own before you move any coordination at all. ## Related - https://www.polarishq.co/for/small-business - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/ai-workers/research-analyst - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/use-cases/finance/budget-reporting - https://www.polarishq.co/use-cases/operations/process-documentation --- --- title: "Polaris for Product Teams: the Translation Tax on a PM Week" description: "Most of a product manager's week is format conversion: research into themes, feedback into tickets, shipped work into release notes. Delegate that layer." url: https://www.polarishq.co/for/product-teams section: Solutions updated: 2026-08-21 --- # Polaris for product teams Look at a product manager's calendar and subtract the meetings. What is left is mostly reformatting. ## The short answer Product managers spend much of the week converting work from one format into another: interviews into themes, feedback into tickets, tickets into release notes, a roadmap into three different presentations of itself. Polaris lets that translation layer be assigned to AI workers, who deliver it as comments and files on the task, while the prioritisation decisions stay with the person who closes the task. ## The job nobody writes in the job description Product management is sold as prioritisation and judgement, and a small part of the week genuinely is. The rest is translation. Nineteen user interviews become four themes. Four themes become a prioritisation argument. The argument becomes a roadmap. The roadmap becomes a version for engineering, a version for the leadership review and a version for the customer-facing team. What shipped becomes release notes, then a changelog entry, then a slide. None of that is low-value, which is why it never gets cut. It is also not where a product manager's specific judgement lives, and every hour of it is an hour not spent with a customer or with the engineers building the thing. That is the tax, and it is the largest recoverable block in the role. ## Four translations worth delegating Each one is a task with a clear input, a clear output and an acceptance criterion the worker can tick. - **Interviews into themes** — Raw research is unread research. A worker produces a first synthesis with the quotes attached, which turns the argument from what do we think people said into a document people can disagree with. - **Feedback into candidate tickets** — Support conversations, sales notes and Slack complaints arrive in the Inbox as prefilled suggestions with bucket, lane, labels and owner already set. One click makes a real task; nothing gets created silently. - **Shipped work into release notes** — A draft assembled from the workstream rather than reconstructed on the afternoon of the release, when the details have already gone. - **The market into a standing competitive picture** — Competitor tracking that continues past the first month, running as a scheduled task with live web search and sources attached. ## From a research round to a prioritised backlog The judgement steps are the short ones. That is the point. 1. **Put the round in a workstream** — Interviews, notes and recordings in a bucket with lanes shared between list and board view. Docs are nested with a block editor, versioned files, review and comments, so the raw material and the conclusions live in the same place. 2. **Assign the synthesis** — A worker produces themes with supporting quotes, ticking its acceptance criteria as it goes and posting progress. This is the four-hour job that a product manager usually does at the end of a fragmented week. 3. **Argue with the draft** — You read a document you did not write, which is a better starting position for a prioritisation discussion than a blank page and your own recall of interview eleven. 4. **Decide, and close the task** — The machine never marks a task done. You close and rate it, so the prioritisation call and the record of who made it stay with a named person. ## Where a product manager's week goes **Before** - Synthesis attempted in the gaps between meetings - Feedback triaged from four tools by copy and paste - Release notes written from memory on release day - Competitive tracking abandoned in month two **With the translation layer assigned** - A themed synthesis waiting, with quotes attached - Suggestions in one Inbox, each one click from being a task - A release note draft ready to correct - A standing competitive task that keeps running > **The line worth holding** > > Delegate the translation, never the decision. Agents deliver and humans close, which is enforced in the product rather than left to discipline: a worker ticks its own acceptance criteria and posts its output, but it cannot mark a task done. If a product team ever finds itself accepting a prioritisation because an agent proposed it, the problem is not the agent. ## Next, for a product team - [use-cases/product](https://www.polarishq.co/use-cases/product) - [use-cases/product/user-research-synthesis](https://www.polarishq.co/use-cases/product/user-research-synthesis) - [use-cases/product/release-notes](https://www.polarishq.co/use-cases/product/release-notes) - [use-cases/product/competitive-tracking](https://www.polarishq.co/use-cases/product/competitive-tracking) - [ai-workers/competitive-analyst](https://www.polarishq.co/ai-workers/competitive-analyst) ## Questions people ask **Does this replace Linear or Jira for the engineering team?** It can, because Polaris has tasks, lanes and board and list views, but it does not have to. Linear, Jira-adjacent workflows and GitHub are questions of migration cost, and Linear and GitHub are both in the connection catalog. Plenty of product teams will run the research and synthesis side in Polaris while engineering stays where it is. **How good is research synthesis from an AI worker, honestly?** Good enough to argue with, not good enough to accept. Treat it as a first pass with the quotes attached, produced by a worker briefed through a skill file you wrote. The value is that a themed draft exists on Monday instead of on the following Thursday, not that the themes are correct. **What stops the Inbox becoming another backlog nobody reads?** Suggestions are never created as tasks silently. Each one arrives with the bucket, lane, labels and owner prefilled and requires one click to become real, so the Inbox is a decision queue rather than a second list that accumulates. If you ignore it, it does not quietly fill your board with tickets. **Can a worker be given access to our customer feedback tools?** Partly. The connection catalog covers Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp, Instagram and open web research. If your feedback lives in a dedicated product-feedback tool, it is not in the catalog today and you would bring that material in manually. **Where does the Focus lane fit for a PM?** Focus is the home screen and it is time-based rather than project-based: today, this week, the next thirty days, pinned first in every view. For a product manager running four workstreams at once, that horizon view is closer to how the job is actually experienced than a per-project board is. ## Related - https://www.polarishq.co/for/software-teams - https://www.polarishq.co/for/design-studios - https://www.polarishq.co/use-cases/product - https://www.polarishq.co/use-cases/product/user-research-synthesis - https://www.polarishq.co/use-cases/product/feature-prioritization - https://www.polarishq.co/use-cases/product/release-notes - https://www.polarishq.co/ai-workers/competitive-analyst - https://www.polarishq.co/ai-workers/project-coordinator --- --- title: "Polaris for small business: someone to delegate to, finally" description: "A six-person business has plenty to delegate and nobody to delegate to. Polaris adds AI workers with no software cost, no terminal and no password to remember." url: https://www.polarishq.co/for/small-business section: Solutions updated: 2026-08-21 --- # Polaris for small business The jobs that never get done in a small business are not hard. There is simply nobody free to do them. ## The short answer Small businesses accumulate work that matters but never reaches the top of anyone's day: chasing invoices, keeping procedures written down, answering routine customer questions, producing the monthly summary. Polaris adds AI workers who take that work, hired through a chat interview in about sixty seconds. The software costs nothing, sign-in is an emailed code, and nobody opens a terminal. - **Software cost:** $0 - **Passwords to manage:** None. Emailed code - **Terminal required:** No - **Billing unit:** ~$2 per human-hour delivered ## Delegation without a bench In a company of six, everybody has a job and everybody is doing it. There is no spare capacity, no junior to hand things to, and no budget that comfortably absorbs a part-time hire for work that amounts to perhaps a day a week. So a specific category of task simply never happens: important, not urgent, and nobody's actual role. The tools sold to small businesses do not address this. A tracker gives you a better list of the things you are not doing. A doc tool gives you somewhere to write the procedure you have not written. Neither of them does any of it, and both charge per person per month for the privilege of holding your intentions. ## The jobs that never reach the top of the day Every small business has some version of this list, and it is remarkably stable across industries. - **Invoices that are late and nobody has chased** — The follow-up sequence everyone agrees should be systematic and which currently depends on somebody being annoyed enough to write an email. - **How we do things, written down** — Opening procedures, handover notes, the way the booking system is actually used. It lives in two people's heads and it leaves with them. - **The same customer question, answered from scratch** — Forty times a month, slightly differently each time, by whoever happens to see it first. - **The monthly picture** — A summary of what came in, what went out and what changed, produced late enough that it is history rather than a decision. ## Hiring your first worker without a technical person Every answer in the interview is a click. There is nothing to install. 1. **Sign in with an emailed code** — Enter your email, get a code, you are in. There is no password path anywhere in the product, which removes the most common reason a small team abandons a tool in month two. 2. **Tell the Chief of Staff what keeps slipping** — In ordinary language, in chat. It is on your roster from the first sign-in, so you never start with an empty screen and a create-project button. 3. **Answer the hiring interview** — A name, the role, what they should be great at, which tools they need. Option pills and multi-select, so every answer is a click rather than a blank field. 4. **Connect only what that worker needs** — One click per connection from a fixed catalog, authorised once for the whole business. Credentials are verified live and stored server-side; browsers cannot read them back. 5. **Give it one real job this week** — The worker card lands in chat ready for its first task. Assign something ordinary, read what comes back as a comment with any files attached, and close it yourself. > **If your business is physical, be clear about what this touches** > > Polaris does nothing about a van, a kitchen, a treatment room or a shop floor. What it can take is the paperwork orbiting them: the supplier emails, the procedure documents, the invoice chasing, the monthly numbers. If the office side of your week is already under control and the constraint is people on the ground, this is not your problem solved and you should not pretend otherwise. ## What it costs, plainly The workspace is free for everyone in the business, permanently, with no seat count and no tier. Tasks, docs, team chat, file versioning and the Chief of Staff are all included. AI work is billed at roughly two dollars per human-equivalent hour delivered, estimated by an open formula from observable effort and itemised job by job on a work log. If a line looks wrong you can challenge it from the log. A month with nothing assigned costs nothing at all. ## Next, for a small business - [ai-workers/bookkeeper](https://www.polarishq.co/ai-workers/bookkeeper) - [ai-workers/executive-assistant](https://www.polarishq.co/ai-workers/executive-assistant) - [use-cases/finance/invoice-tracking](https://www.polarishq.co/use-cases/finance/invoice-tracking) - [use-cases/operations/sop-maintenance](https://www.polarishq.co/use-cases/operations/sop-maintenance) - [cost/stack-cost-5-person-team](https://www.polarishq.co/cost/stack-cost-5-person-team) ## Questions people ask **We are not a tech company. Is this built for us?** The parts that require technical knowledge are optional. Hiring a worker is a chat interview with clickable answers, connections are one click each, and sign-in is an emailed code. The technical detail exists underneath and you can read it if you want to, but nothing in the daily use of the product requires it. **Is it really free, or is there a seat charge hiding somewhere?** The software is free with no seat count and no tier. Unlimited people, tasks, workstreams and docs, plus the Chief of Staff, are included for every organisation. The only bill is roughly two dollars per human-equivalent hour of work an AI worker delivers, and nothing delivered means nothing billed. **Who is responsible if a worker gets something wrong?** You are, in the same way you are responsible for a new employee's first draft. That is why the machine never marks a task done: it delivers the work as a comment, ticks its acceptance criteria and stops, and a person in your business closes and rates it. Never send anything to a customer or a tax authority that nobody has read. **How long before it is useful?** Hiring the first worker takes about a minute. Getting a useful delivery out of it takes as long as one task, which is the only honest way to evaluate this. Assign something real from this week rather than a test, because a test task produces a test-quality answer and tells you nothing. **What happens to our information if we stop using it?** Polaris is an early-stage product in public beta, so this is a reasonable thing to ask before you commit. Keep the documents that matter exportable, treat the first workstream as an experiment rather than a migration, and expand once you have seen a few deliveries you would have paid a person for. ## Related - https://www.polarishq.co/for/nonprofits - https://www.polarishq.co/for/solo-founders - https://www.polarishq.co/ai-workers/bookkeeper - https://www.polarishq.co/ai-workers/executive-assistant - https://www.polarishq.co/ai-workers/support-specialist - https://www.polarishq.co/use-cases/finance/invoice-tracking - https://www.polarishq.co/use-cases/operations/sop-maintenance - https://www.polarishq.co/glossary/passwordless-authentication --- --- title: "Polaris for design studios: the unbillable half of the week" description: "Studios bill for design and lose money on everything around it: handoff specs, client updates, asset naming, brand audits. Polaris puts workers on that half." url: https://www.polarishq.co/for/design-studios section: Solutions updated: 2026-08-21 --- # Polaris for design studios Nobody has ever put a line item for asset renaming on an invoice, and every studio pays for it anyway. ## The short answer Design studios bill for design work and absorb everything surrounding it: handoff specifications, weekly client updates, asset naming and organisation, brand consistency audits and scoping documents. Polaris lets that surrounding layer be assigned to AI workers, with Figma in the connection catalog, deliverables arriving as files on the task, and a designer closing every task before it counts as done. ## Where a studio actually loses money A studio quotes for a piece of design work, and the design work usually lands roughly where it was estimated. What blows the estimate is the layer around it. The specification a developer needs before they can build the thing. The weekly update to a client who is anxious because they cannot see progress. Renaming three hundred exports so the handoff is usable. The audit of whether the brand still looks like itself across nine surfaces after a year of small decisions. None of this is design and all of it is required. It is also the work that gets deprioritised when a deadline tightens, which is exactly when its absence causes the most damage: a rushed handoff produces a build that does not match, and the studio absorbs the correction round. ## Five jobs that eat a studio week - **Handoff specification** — Spacing, states, tokens, behaviour and edge cases, written up so an engineer does not have to guess or ask. Slow, careful, and almost never quoted for. - **The weekly client update** — Pure cost, always late, assembled from a board and two conversations. It is also the single biggest driver of whether a client feels well served. - **Design system upkeep** — The component that drifted, the token that got overridden, the pattern used two ways in the same product. Everyone agrees it should be maintained continuously and nobody has the hours. - **Brand consistency audits** — Checking a brand against itself across site, product, social and print. Genuinely valuable to a client and hard to sell as a line item. - **Scoping and estimating** — Turning a vague brief into a scoped proposal, which consumes senior designer time before any money exists. ## What changes when the layer is assigned The design stays with designers. The documentation around it does not have to. | Job | Today | Assigned to a worker | | --- | --- | --- | | Handoff spec | Written by the designer at the end of the project, when they are already on the next one | Drafted from the workstream and the Figma connection, then corrected by the designer | | Weekly client update | Unbillable, written on a Friday afternoon | Drafted from the workstream, reviewed and sent by the account lead | | Design system audit | Deferred indefinitely | A standing task that reports drift, with the findings as a file on the task | | Proposal and scope document | Senior designer time spent before the work is won | A structured draft to argue with, so the senior time goes into the estimate rather than the formatting | > **Polaris does not do design, and will not pretend to** > > There is no canvas, no image generation and no visual output in the product. Figma is a connection, which means a worker can be given access to it, not that anything here replaces it. If you are looking for a tool that produces design work, this is the wrong page. What is on offer is the writing, tracking and documentation that surrounds design and currently comes out of your margin. ## How a studio would set it up **The structure** - One workstream per client or per project - Lanes matching your stages, shared between list and board view - Docs nested per project, versioned, with review and comments - Freelance collaborators added at no cost, because there are no seats **The roster** - A technical writer for handoff specs and documentation - A content writer for client updates and case study drafts - A research analyst for competitive and category work in pitches - The Chief of Staff keeping every task owned and dated across projects ## Next, for a design studio - [integrations/figma](https://www.polarishq.co/integrations/figma) - [use-cases/design](https://www.polarishq.co/use-cases/design) - [use-cases/design/asset-handoff](https://www.polarishq.co/use-cases/design/asset-handoff) - [use-cases/design/brand-consistency-audits](https://www.polarishq.co/use-cases/design/brand-consistency-audits) - [for/agencies](https://www.polarishq.co/for/agencies) ## Questions people ask **Can a worker read our Figma files?** Figma is in the connection catalog, so a worker can be given access to it with one click, authorised once for the organisation and stored server-side. What it does with that access depends on the task you assign and the skill file you wrote for it. It is not a design tool and it does not produce design work. **Is this better than Notion plus a tracker for a studio?** For holding project information, Notion and a tracker are perfectly good and you should not switch on aesthetics. The difference is that neither of them writes the handoff spec or the client update. If the layer costing you margin is documentation rather than organisation, that is the reason to look; if your problem is genuinely organisation, be honest that a free workspace is a smaller win. **How would we price this into client work?** Hours are itemised on the work log with an estimated human-equivalent hour count per job, produced by an open formula. Whether that becomes a passed-through cost, an absorbed margin improvement or a reason to quote a documentation phase properly is your commercial decision. The evidence supports all three conversations. **We are four people. Does the workspace cost anything at that size?** No. There is no seat count and no tier, so four designers plus whatever freelancers a project needs cost nothing, and removing them later costs nothing either. The only bill is roughly two dollars per human-equivalent hour that an AI worker delivers. **Is a design studio a good first user of a beta product?** Honestly, only in one place. Client-facing deliverables in a public beta carry a reputational risk that a studio cannot really afford. Start with something internal, such as your own design system audit or your own case studies, where a weak first draft costs an afternoon and nobody outside the studio ever sees it. ## Related - https://www.polarishq.co/for/agencies - https://www.polarishq.co/for/product-teams - https://www.polarishq.co/integrations/figma - https://www.polarishq.co/use-cases/design - https://www.polarishq.co/use-cases/design/design-system-maintenance - https://www.polarishq.co/use-cases/design/asset-handoff - https://www.polarishq.co/ai-workers/technical-writer - https://www.polarishq.co/cost/stack-cost-5-person-team --- --- title: "Polaris for technical founders: your agents are invisible" description: "You already run coding agents locally. Nobody else can see them, assign to them, or use what they produced, and they stop when your laptop closes." url: https://www.polarishq.co/for/technical-founders section: Solutions updated: 2026-08-21 --- # Polaris for technical founders The agent that wrote half your week's work is single-player, local, and gone at midnight. ## The short answer Technical founders already run coding and research agents on their own machine. Those agents are single-player: teammates cannot see them, assign work to them, or find what they produced, and they stop when the laptop closes. Polaris puts AI workers in a shared roster where humans and agents are the same members table, runs each job on a cloud machine, and logs every delivery. - **Runtime:** A cloud machine per job - **Worker identity:** An editable SKILL.md - **Humans and agents:** One members table - **Laptop closed:** Work continues ## The problem is not capability, it is topology If you are a technical founder in 2026, you are probably already getting real output from agents. That is not the gap. The gap is that all of it happens in a terminal on one machine, under one person's account, with the results ending up as a paste into Slack or a commit with no context attached. Your co-founder cannot assign anything to the thing doing a third of the work. Your first non-technical hire cannot use it at all. And it stops. Not because it failed, but because you closed the lid to get on a train. A local agent's working life is bounded by your working day, which is a strange constraint for something that has no reason to sleep. ## Four things a local agent cannot do None of these are model problems. They are all consequences of where the process runs and who owns it. - **Be assigned work by someone else** — Your designer cannot hand it a task. In Polaris, members carry a kind of human or agent and assignment works identically for both, so a task goes to a worker exactly the way it goes to a person. - **Keep going after you disconnect** — A worker runtime on Fly.io claims jobs off a queue and a real machine wakes per task. It runs its tool loop, posts progress and delivers whether or not you are at your desk. - **Leave a record anyone can audit** — Local sessions leave scrollback. Every Polaris job is logged with its estimated human-equivalent hours and its output attached, and that log is also the billing record, so it has to be accurate. - **Carry a shared, reviewable identity** — Your local configuration is yours. A Polaris worker's capabilities are a SKILL.md file the team can read and edit, so what the agent is good at becomes a team artifact rather than a personal setup. ## Local agent versus a hired worker **On your machine** - One person's terminal, one person's account - Stops with the lid, resumes when you remember - Output arrives as a paste with no task attached - Credentials live in your local environment - Nothing to hand to a non-technical colleague **On the roster** - A member of the org, assignable by anyone on the team - A cloud machine wakes per job and keeps working - Delivery is a comment on the task, with files - Connections authorised once, org-wide, stored server-side - Hired through a chat interview in about sixty seconds ## From a local habit to a shared roster You are not throwing away how you work. You are giving it an address the rest of the company can reach. 1. **Name what you already delegate locally** — The research passes, the drafting, the recurring digests, the analysis you re-run by hand. That list is the job description of your first hired worker. 2. **Run the hiring interview** — The Chief of Staff asks for a name, a role, what it should be great at and which tools it needs. Every answer is a click, and the capabilities become a SKILL.md you can open and edit rather than a hidden prompt. 3. **Authorise the connections once** — GitHub, Slack, Linear, Notion, Supabase and the rest of the catalog, connected at organisation level and verified live. Credentials are stored server-side; workers use them and browsers cannot read them back. 4. **Assign a real task and read the work log** — The machine compiles the worker's identity, its skill files and the task context, runs a live tool loop with real web search, ticks its own acceptance criteria and posts the deliverable as a comment with any generated files. 5. **Close it yourself** — Agents deliver; humans close. The completion signal stays with a person, which is what makes the board trustworthy once the team is bigger than you. > **Why the members table is the whole argument** > > Humans and agents are rows in the same table with a kind of human or agent, and assignment works identically for both. That is not an implementation detail, it is the positioning expressed as a schema. Every product that bolts an AI panel onto a tracker has made the opposite choice, and the consequence is that AI work permanently lives beside the real work instead of inside it. ## Next, for a technical founder - [cloud-claude-code](https://www.polarishq.co/cloud-claude-code) - [cloud-claude-code/claude-code-for-teams](https://www.polarishq.co/cloud-claude-code/claude-code-for-teams) - [cloud-claude-code/claude-code-when-laptop-is-closed](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) - [cloud-claude-code/skill-files-for-ai-workers](https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) ## Questions people ask **Does Polaris replace running agents locally?** For your own solo work, probably not, and it does not try to. A local setup is fast, private and yours. The case here is for the work that other people need to see, assign or reuse, and for the jobs that should keep running after you shut the laptop. Most technical founders will end up doing both. **What actually runs the job?** A worker runtime on Fly.io claims jobs off an agent_jobs queue and a machine wakes for the task. It compiles the worker's instructions, its skill files and the task context, then runs a live tool loop with real web search. The queue contract is runtime-agnostic, so the machine behind it is swappable. **Where does the data live?** Supabase provides Postgres, auth, realtime, row-level security and edge functions, and it is the API for both the frontend and the workers. The frontend is a Next.js static export on Cloudflare Pages. Auth is passwordless everywhere: email, then an emailed code, with no password path in the product. **Can I edit what a worker knows, or is it a black box?** You can edit it. Capabilities are a real SKILL.md file that you read and change like any other file, which is the point of doing it that way. If a worker keeps making the same wrong assumption, that is a file edit and a re-run, not a support ticket. **How is this priced for someone who will assign a lot of work?** The software is free with no seats. Billing is roughly two dollars per human-equivalent hour delivered, estimated from observable effort by an open formula and clamped between five minutes and eight hours per session, itemised job by job on the work log. Heavy use costs more, which is the honest trade for a model with no fixed component. **It is a beta. What is genuinely missing?** There are no users yet, so there is no operational track record to point at, and there are no customer stories because there are no customers. The mobile app is App Store prepped but TestFlight is still blocked on an App Store Connect API key. Judge it on the demos and on your own first delivery, not on adoption. ## Related - https://www.polarishq.co/for/software-teams - https://www.polarishq.co/for/solo-founders - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates - https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/integrations/github --- --- title: "Glossary of agentic work: 24 terms, defined plainly" description: "Twenty-four definitions for the language of AI teammates and agentic work — origins, practical usage, and the neighbouring term each one gets confused with." url: https://www.polarishq.co/glossary section: Glossary updated: 2026-08-21 --- # The vocabulary of agentic work Definitions written to be quoted: where the term came from, how practitioners use it, and what it is not. ## The short answer Agentic work has its own vocabulary: AI workers, agent runtimes, skill files, acceptance criteria, human-in-the-loop review, human-equivalent hours. This glossary defines twenty-four of those terms in plain language, covering where each came from, how practitioners use it, and which neighbouring term it gets confused with. Sixteen are defined as the wider industry uses them. Eight are Polaris usage, labelled as such. - **Terms defined:** 24 - **Industry definitions:** 16 - **Polaris usage:** 8 - **Last checked:** 21 August 2026 ## Why this glossary exists The words around AI agents got adopted faster than they got defined. Vendors named the same thing four ways, and four different things one way. A buyer comparing two products often cannot tell whether both mean the same thing by autonomous, or by orchestration, or by usage-based. So each page here does four jobs in the same order: a definition that stands on its own, where the term came from, how it is actually used in practice, and what it is commonly mistaken for. Where a definition is contested, the page says so rather than picking the convenient one. ## Two kinds of entry, marked differently - **Industry terms** — Model Context Protocol, headless agent, human-in-the-loop, generative engine optimization, cloud development environment, autonomous agent, agent orchestration, per-seat pricing, usage-based pricing and tool sprawl are defined as the field defines them, checked against primary sources and dated. - **Polaris usage** — Workstream, focus lane, delivery comment, work log, human-equivalent hours, AI worker, skill file and agent runtime are words the industry has not settled. Each page gives the general sense first, then says plainly what Polaris means by it. ## Every term - [AI worker](https://www.polarishq.co/glossary/ai-worker) — The difference between an agent you prompt and an agent you assign work to. - [Agentic project management](https://www.polarishq.co/glossary/agentic-project-management) — The line is whether the agent holds the work item or only summarises it. - [AI teammate](https://www.polarishq.co/glossary/ai-teammate) — An assistant belongs to a person. A teammate belongs to the team. - [Cloud development environment](https://www.polarishq.co/glossary/cloud-development-environment) — The compute, filesystem and toolchain live somewhere else, and are usually thrown away afterwards. - [Headless agent](https://www.polarishq.co/glossary/headless-agent) — Nobody is typing at it, so its output is actions and artefacts rather than replies. - [Skill file](https://www.polarishq.co/glossary/skill-file) — The capability an agent has, written down where a person can read and edit it. - [Model Context Protocol (MCP)](https://www.polarishq.co/glossary/model-context-protocol) — One protocol between models and the systems they need, instead of one integration per pair. - [Agent orchestration](https://www.polarishq.co/glossary/agent-orchestration) — Deciding which agent does what, in what order, and what happens when one fails. - [Human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) — The system cannot complete the loop without a person, by design. - [Acceptance criteria](https://www.polarishq.co/glossary/acceptance-criteria) — Written before the work, checkable after it, and binary either way. - [Work log](https://www.polarishq.co/glossary/work-log) — The record that makes an unwatched run reviewable afterwards. - [Human-equivalent hours](https://www.polarishq.co/glossary/human-equivalent-hours) — A billing unit denominated in the work replaced, not the compute consumed. - [Usage-based pricing](https://www.polarishq.co/glossary/usage-based-pricing) — The bill follows consumption, which cuts both ways for buyer and vendor. - [Per-seat pricing](https://www.polarishq.co/glossary/per-seat-pricing) — The bill tracks headcount, which is only a proxy for value received. - [Tool sprawl](https://www.polarishq.co/glossary/tool-sprawl) — The expensive part is not the licences. It is that nothing is authoritative any more. - [All-in-one workspace](https://www.polarishq.co/glossary/all-in-one-workspace) — One data model behind documents, tasks and conversation, instead of three products and a pile of integrations. - [Task queue](https://www.polarishq.co/glossary/task-queue) — The thing that lets a request survive the process that made it. - [Agent runtime](https://www.polarishq.co/glossary/agent-runtime) — Not the model, not the framework: the thing that actually runs the job. - [Autonomous agent](https://www.polarishq.co/glossary/autonomous-agent) — Autonomy is a range, and the interesting question is where the boundary sits. - [Delivery comment](https://www.polarishq.co/glossary/delivery-comment) — Work arrives where the task already lives, attributed and reviewable, and the task stays open. - [Workstream](https://www.polarishq.co/glossary/workstream) — A strand of work that keeps going, rather than a project that ends. - [Focus lane](https://www.polarishq.co/glossary/focus-lane) — A commitment for a horizon, not a filter over everything you have. - [Passwordless authentication](https://www.polarishq.co/glossary/passwordless-authentication) — Removing the shared secret the user has to remember, and attackers have to steal. - [Generative engine optimization](https://www.polarishq.co/glossary/generative-engine-optimization) — Optimising to be quoted inside an answer rather than ranked beneath one. ## Where the vocabulary gets used - [cloud-claude-code](https://www.polarishq.co/cloud-claude-code) - [ai-workers](https://www.polarishq.co/ai-workers) - [cost](https://www.polarishq.co/cost) - [integrations](https://www.polarishq.co/integrations) - [use-cases](https://www.polarishq.co/use-cases) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) ## Questions people ask **How often is this glossary checked?** Every entry was verified on 21 August 2026 against primary sources: protocol specifications, published research papers, and vendor documentation. Definitions in a field moving this fast decay quickly, so each page carries its own last-checked date rather than a single site-wide one. **Are these definitions written to favour Polaris?** No. Industry terms are defined as the industry uses them, including where the accurate definition is inconvenient. Terms that describe Polaris specifically are labelled as Polaris usage so a reader can tell the difference between a shared standard and one product's vocabulary. **Can these definitions be quoted?** Yes. Each page opens with a self-contained definition of 45 to 75 words that can be lifted into another document without losing meaning. Attribution to Polaris is appreciated but not required. **What is the difference between an AI agent and an AI worker?** An AI agent is the general technical category: a system that pursues a goal by choosing its own steps and tool calls. An AI worker is an agent configured as a named team member with a role, a toolset and a queue of assigned tasks. Every AI worker is an agent; most agents are not workers. ## Related - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/model-context-protocol - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/usage-based-pricing --- --- title: "What is an AI worker? Definition and examples" description: "An AI worker is an agent set up as a named team member with a role, tools and assigned tasks. Definition, origin, and how it differs from a chatbot or copilot." url: https://www.polarishq.co/glossary/ai-worker section: Glossary updated: 2026-08-21 --- # AI worker The difference between an agent you prompt and an agent you assign work to. ## The short answer An AI worker is an AI agent configured as a named member of a team: it has a role, written instructions, a defined set of tools, and a queue of tasks assigned to it by people. Unlike a chatbot, which answers when someone opens it, an AI worker holds standing responsibilities and returns finished deliverables that a human reviews and accepts. - **Category:** Emerging industry term - **Nearest older term:** Software agent - **Key property:** Assignable, not just promptable ## Where the term came from The phrase spread through 2024 and 2025 as vendors tried to name the shift from AI you talk to towards AI you hand work to. Older vocabulary did not fit. Assistant implied a person invoking it. Copilot implied a person sitting alongside. Agent was accurate but described a technical property rather than a place in an organisation chart. Worker is a claim about organisational position, not capability. It says the thing shows up in the same list as the humans, takes assignments the same way, and produces output that goes through the same review. ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Chatbot | A conversational interface to a model | Stateless towards work. It answers a question and forgets the responsibility. | | Copilot | In-context suggestions inside a tool a human is using | The human is doing the task. The copilot narrows the next keystroke. | | Autonomous agent | A system that chooses its own steps toward a goal | Describes decision latitude, not team membership. An AI worker is usually an autonomous agent given a job title. | | Workflow automation | A fixed sequence triggered by an event | The path is authored in advance. An AI worker decides its own path within the brief. | ## How Polaris uses the term Polaris means something specific and checkable by AI worker. - **Same table as the humans** — Humans and AI workers are both rows in one members table with kind set to human or agent. Assignment, ownership and mentions work identically for both. - **Hired in about sixty seconds** — A short interview in chat produces a name, a role, a capability set and the tool connections the worker needs. The capabilities are written to an editable SKILL.md file rather than a hidden prompt. - **Delivers on a real machine** — Assigning a task to an AI worker wakes a cloud machine that runs a live tool loop, ticks its acceptance-criteria checklist, and posts the finished work as a comment with any generated files attached. - **Never closes its own task** — The machine delivers. A human closes and rates the work. That boundary is the reason the output stays reviewable. ## Related terms - [glossary/ai-teammate](https://www.polarishq.co/glossary/ai-teammate) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [glossary/skill-file](https://www.polarishq.co/glossary/skill-file) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [ai-workers](https://www.polarishq.co/ai-workers) ## Questions people ask **Is an AI worker the same as an AI agent?** Not quite. AI agent is the technical category: software that pursues a goal by choosing its own steps and tool calls. AI worker describes an agent that has been given a role, a toolset and a place in a team's assignment flow. Every AI worker is an agent; most agents in production are not workers. **Do AI workers replace headcount?** In practice, teams use them for work that was not getting done rather than work someone was already doing: research backlogs, documentation debt, recurring reports. The honest framing is capacity, not replacement, because an AI worker still needs a human to write the brief and accept the result. **What stops an AI worker from doing the wrong thing?** Three limits are standard: the tools it has been granted, the acceptance criteria on the task, and a human review step before anything is marked complete. Tool access is the strongest of the three, because an agent cannot act on a system it has no credentials for. **How is an AI worker billed?** Billing models vary by vendor: per seat, per credit, per token, or per unit of delivered work. Polaris bills roughly two dollars per human-equivalent hour delivered, estimated by an open formula and itemised on the worker's work log, with no charge when nothing is delivered. ## Related - https://www.polarishq.co/glossary/ai-teammate - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/ai-workers - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent --- --- title: "Agentic project management: definition and practice" description: "Agentic project management is project work where AI agents hold assigned items and act toward goals within guardrails. Definition, origin, and what it is not." url: https://www.polarishq.co/glossary/agentic-project-management section: Glossary updated: 2026-08-21 --- # Agentic project management The line is whether the agent holds the work item or only summarises it. ## The short answer Agentic project management is the practice of running projects in which AI agents hold assigned work items and act toward goals within human-defined guardrails, rather than only summarising or autofilling for a person. Agents plan their own steps, gather data and produce deliverables; people set the goal, the acceptance criteria and the approval gate. Traditional project management keeps every action with a human. - **Term in use since:** 2024 - **Distinguishing test:** Can an agent be an assignee? - **Human role:** Goals, criteria, approval ## Where the term came from The phrase follows agentic AI, which distinguishes systems that pursue goals over multiple steps from systems that respond to a single prompt. Project management vendors picked it up through 2025 and 2026, and by now Atlassian, Planview and Wrike all publish material under the heading. The vendor definitions converge on one idea: an agent that detects a risk, reconciles a dependency or drafts a status report on its own initiative and routes it to a person, instead of waiting to be asked. What they disagree about is how much latitude counts as agentic, which is why the useful test is structural rather than adjectival. ## The test that actually separates the two Ignore the marketing adjectives and ask three questions of the product. - **Can an agent be the assignee?** — In AI-assisted project management, every task still belongs to a person and the AI decorates it. In agentic project management, the agent owns the item and appears in the same assignee list as the humans. - **Does work continue without a session?** — Assistance runs while a human sits in the tool. Agentic work runs off a queue on infrastructure that does not care whether anyone has a browser tab open. - **Is there an audit trail of what it did?** — A goal-pursuing agent takes steps nobody watched. Without a per-job record of searches, tool calls, files and time, agentic project management is unreviewable and therefore unusable for anything that matters. ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | AI-assisted project management | Summaries, autofill, smart suggestions inside a tracker | The human performs every action. The AI reduces typing. | | Workflow automation | Rules that fire on events: when status changes, notify channel | The branches are authored in advance by a person. No goal-seeking. | | Agentic AI | The general capability class of goal-pursuing AI systems | The parent category. Agentic project management is one application of it. | | Autonomous project management | Marketing phrasing implying no human involvement | Rarely accurate. Every credible implementation keeps a human approval gate. | ## What it looks like in Polaris Polaris is built on the structural version of the definition. Humans and AI workers are rows in the same members table, so an AI worker can be the assignee on any task. Assigned work lands on a job queue and a cloud machine claims it, runs a tool loop with live web search, ticks the acceptance criteria as it completes them, and posts the result as a comment with attached files. The one thing the machine never does is close the task. A person closes it and rates the work, which keeps the review gate where the definition requires it. ## Related terms - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [use-cases](https://www.polarishq.co/use-cases) ## Questions people ask **Is agentic project management a real methodology or a marketing label?** Both, currently. There is no standards body and no certification behind the phrase, so it functions as a label vendors apply to varying levels of capability. The structural test, whether an agent can be an assignee and work off a queue, separates the products that mean it from the products that do not. **Does agentic project management replace Scrum or Kanban?** No. It changes who performs the work items, not how work is planned, sequenced or reviewed. Teams running Kanban keep their board; some cards are now owned by an agent and arrive back as a delivered comment for review. **What is the main risk?** Unreviewed output entering the record as if a person had produced it. The mitigation is structural rather than cultural: acceptance criteria written before the work starts, a per-job log of what the agent actually did, and a human close step that the agent cannot perform. ## Related - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/agent-orchestration - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/use-cases --- --- title: "AI teammate: definition and how it differs from an assistant" description: "An AI teammate is an AI that appears in a team's own tools as a named member with assignable work. Definition, origin, and the difference from an AI assistant." url: https://www.polarishq.co/glossary/ai-teammate section: Glossary updated: 2026-08-21 --- # AI teammate An assistant belongs to a person. A teammate belongs to the team. ## The short answer An AI teammate is an AI system that appears inside a team's own tools as a named member: visible to everyone, mentionable, assignable, and producing work that lands in shared channels rather than one person's private session. The framing contrasts with an AI assistant, which is invoked by an individual, and whose context, history and output stay with that individual. - **Unit of ownership:** The team, not the person - **Visibility test:** Can a colleague see the work? - **Related terms:** AI worker, copilot, assistant ## Why the distinction is not just wording Most AI at work in 2026 is single-player. A person opens a chat window, gets an answer, and pastes some of it somewhere. The context that made the answer good, the prompt, the files, the corrections, dies in that session. A colleague cannot see it, build on it, or assign the same job to it next week. Teammate framing describes the alternative: one shared roster, one shared history, one place where the output lands with an author name attached. The practical consequence is that a team develops institutional knowledge about how to brief the thing, in the same way it develops knowledge about how to brief a new hire. ## Commonly confused with | Term | Belongs to | Where the output lands | | --- | --- | --- | | AI assistant | One person, in a session | That person's clipboard | | Copilot | One person, inside one tool | The document being edited | | AI teammate | The team, on a shared roster | A shared task, doc or channel, attributed | | AI worker | The team, with a defined role and queue | A delivered task, with a work log behind it | ## What a system needs before the label is honest - **A name and a profile everyone can see** — Not a model selector. A member other people can look up, whose role and capabilities are readable by anyone on the team. - **Assignment that works like human assignment** — If assigning to the AI uses a different mechanism from assigning to a colleague, it is a feature bolted onto a tracker rather than a member of it. - **Shared, attributable output** — Work that arrives in a place colleagues already look, with the author recorded, so it can be reviewed, corrected and cited later. - **Persistent identity across tasks** — The same instructions and the same capability files apply on Tuesday as on Monday. A teammate that starts from zero every session is a session, not a teammate. ## Related terms - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [glossary/workstream](https://www.polarishq.co/glossary/workstream) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [glossary/skill-file](https://www.polarishq.co/glossary/skill-file) - [cloud-claude-code/share-claude-code-with-teammates](https://www.polarishq.co/cloud-claude-code/share-claude-code-with-teammates) - [ai-workers](https://www.polarishq.co/ai-workers) ## Questions people ask **Is AI teammate just a rebrand of AI assistant?** The words are used loosely, but the underlying difference is real and checkable. Ask whether a colleague can see the work, assign to the same entity, and read what it did. If the answer is no on all three, the system is an assistant regardless of what it is called. **Do AI teammates need their own accounts?** They need an identity in the workspace, which some products implement as a full account and others as a member record with no login. What matters is that the identity is addressable and that its actions are attributed to it in the audit trail. **How does Polaris implement this?** Humans and AI workers are rows in the same members table, distinguished only by a kind column set to human or agent. Assignment, ownership, comments and activity work identically for both, which is the architectural version of the teammate claim rather than the marketing version. ## Related - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/cloud-claude-code/claude-code-for-teams - https://www.polarishq.co/ai-workers --- --- title: "Cloud development environment (CDE): definition" description: "A cloud development environment is a remote, pre-configured dev environment reached over the network. Definition, origin, and how it differs from a VDI." url: https://www.polarishq.co/glossary/cloud-development-environment section: Glossary updated: 2026-08-21 --- # Cloud development environment The compute, filesystem and toolchain live somewhere else, and are usually thrown away afterwards. ## The short answer A cloud development environment (CDE) is a software development environment (compute, filesystem, dependencies and toolchain) hosted on remote infrastructure and accessed over the network from a browser or a local editor. Environments are typically defined as configuration in the repository and created on demand, so every developer gets an identical setup and can discard it when the branch is done. - **Abbreviation:** CDE - **Examples:** Codespaces, Gitpod, Coder - **Defining property:** Reproducible from config - **Typical lifetime:** Minutes to days ## Where the idea came from Remote development is older than the term. Developers have compiled on shared servers since timesharing, and browser IDEs such as Cloud9 shipped around 2010. What made CDE a named category was the combination of three things arriving together: container images that pin a toolchain, environment definitions checked into the repository, and per-branch provisioning fast enough that developers stopped maintaining a laptop setup. GitHub Codespaces, Gitpod, Coder and Daytona are the products most often named in the category. The problem they exist to kill is the one every team recognises: an environment that works on one machine and not another, and the day of onboarding that goes into fixing it. ## What distinguishes a CDE from a remote server - **Defined as code** — The environment comes from a file in the repository, whether a devcontainer definition, a Dockerfile or a Nix expression, so it is reproducible rather than hand-built. - **Ephemeral by default** — Environments are created for a branch or a task and destroyed afterwards. State that matters lives in version control, not on the box. - **Attached to an editor, not a terminal alone** — A browser IDE or a local editor connected over the network, with language servers, debuggers and port forwarding working as they would locally. - **Provisioned on demand** — Start time measured in seconds to a couple of minutes, because a developer waiting five minutes will go back to their laptop. ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Virtual desktop (VDI) | A remote Windows or Linux desktop session | A whole desktop for general work, persistent, not defined by a repo. | | Remote SSH box | A long-lived server you configure by hand | Not reproducible and not disposable. The classic thing a CDE replaces. | | Agent runtime | Infrastructure that executes an AI agent's job | No human editor attached. The consumer of the environment is a program, not a person. | | CI runner | Ephemeral compute that executes a pipeline | Runs a defined script to completion. Nobody is developing inside it. | ## The overlap with AI agents The CDE category and the agent-runtime category are converging, because an AI agent needs roughly what a developer needs: a filesystem, a toolchain, network access and credentials, created on demand and thrown away after. The difference that still matters is the interface. A CDE assumes a person is typing in it; an agent runtime assumes nobody is watching and therefore has to record what happened. Polaris runs a worker runtime on Fly.io that wakes a machine per task. It is not a CDE and is not sold as one. There is no editor to attach to, and the output is a delivered task comment rather than a branch you inspect by hand. ## Related terms - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [cloud-claude-code/run-claude-code-in-the-cloud](https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud) - [cloud-claude-code/cloud-agents-vs-local-agents](https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) ## Questions people ask **Is a cloud development environment the same as running Claude Code in the cloud?** No. A CDE gives a human developer a remote machine to work in. Running an agent in the cloud gives a program a machine to work in, with no editor session and no person present. The infrastructure looks similar; the interface, the audit requirements and the failure modes are different. **Do CDEs cost more than local development?** They move the cost from hardware to metered compute, which usually shows up as a per-hour or per-seat line item. Teams that adopt them tend to justify it on onboarding time and environment-drift incidents rather than on raw cost, since a laptop is a sunk cost and a CDE is a recurring one. **What is the main limitation?** Latency and offline work. Editing over a network connection is noticeably worse on a bad connection, and nothing works on a plane. Teams that adopt CDEs generally keep local development possible as a fallback rather than deleting it. ## Related - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/cloud-claude-code - https://www.polarishq.co/cloud-claude-code/cloud-agents-vs-local-agents - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed --- --- title: "Headless agent: definition and how it is triggered" description: "A headless agent is an AI agent with no user-facing interface, started by an event, schedule, queue or API call. Definition, origin, and what it is not." url: https://www.polarishq.co/glossary/headless-agent section: Glossary updated: 2026-08-21 --- # Headless agent Nobody is typing at it, so its output is actions and artefacts rather than replies. ## The short answer A headless agent is an AI agent that runs without a user-facing interface. Instead of a person typing into a chat window, it is started by an API call, a webhook, a schedule, a queue message or a database change, and its output is actions and artefacts (a written file, an updated record, a sent message) rather than a conversational reply. The term borrows from headless CMS and headless commerce. - **Term lineage:** Headless CMS, headless commerce - **Trigger:** Event, schedule, queue, API - **Output:** Actions and files, not replies ## What headless actually removes In web architecture, headless means the system has no presentation layer of its own: a headless CMS stores and serves content that some other front end renders. Applied to agents, the head that is missing is the chat interface. The reasoning, tool use and model calls are unchanged; what disappears is the assumption that a human is present to read a response and type the next turn. That absence changes the engineering. A headless agent has to decide when it is finished without anyone confirming, has to fail in a way an operator can diagnose later, and has to leave a record, because there is no transcript anybody watched. ## How headless agents are usually triggered Four patterns cover most production deployments. - **Queue message** — A row or message representing a unit of work is claimed by a worker process. The most common pattern where ordering, retries and concurrency limits matter. - **Webhook or event** — An external system fires on a state change, such as a ticket created, a payment failed or a pull request opened, and the agent handles it. - **Schedule** — A cron expression runs the agent periodically: a nightly report, a weekly audit, an hourly inbox sweep. - **Direct API call** — Another program invokes the agent synchronously and consumes its structured result, treating it as a function with judgement. ## Commonly confused with | Term | What it describes | The difference | | --- | --- | --- | | Autonomous agent | How much latitude the agent has over its own steps | A property of decision-making, not of interface. An agent can be headless and tightly scripted, or interactive and highly autonomous. | | Background job | Deferred work executed outside a request cycle | Runs a fixed procedure. A headless agent chooses its steps at run time. | | Batch processing | Large volumes of records processed on a schedule | Uniform work, no per-item reasoning. | | Daemon | A long-running system process | Describes process lifetime. A headless agent is usually invoked per unit of work and exits. | ## The observability problem The defining operational fact about headless agents is that nobody saw what happened. Interactive agents get corrected mid-run by the person reading along; headless ones do not. Every serious deployment therefore pairs the agent with a durable record of its steps: tool calls, searches, files produced, and the criteria it believed it satisfied. Polaris runs its AI workers headless, claiming jobs off a queue with a machine on Fly.io, and pairs each run with a work log and progress comments posted onto the task, so the run can be read after the fact by whoever reviews the delivery. ## Related terms - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) - [cloud-claude-code/ai-agent-audit-trail](https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail) ## Questions people ask **Is a headless agent the same as an autonomous agent?** No, and conflating them is the most common error with this term. Headless describes the interface: no chat window, triggered programmatically. Autonomous describes latitude: the agent chooses its own steps. A headless agent following a rigid script is neither autonomous nor unusual. **Can a headless agent ask a human a question?** Yes, through an asynchronous channel rather than a live turn. Common patterns are posting a comment and pausing, opening an approval request, or escalating to a queue a person monitors. The agent cannot block on an answer the way an interactive one can, so the design has to handle nobody replying. **What are headless agents good for?** Work that is triggered by systems rather than people, or that takes longer than someone will sit and watch: overnight research, recurring reports, ticket triage, document generation. Anything requiring rapid back-and-forth clarification is a poor fit. ## Related - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/cloud-development-environment - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed --- --- title: "Skill file (SKILL.md): definition and structure" description: "A skill file teaches an agent one capability: when to use it, the procedure, the constraints. Definition, structure, and how it differs from a system prompt." url: https://www.polarishq.co/glossary/skill-file section: Glossary updated: 2026-08-21 --- # Skill file The capability an agent has, written down where a person can read and edit it. ## The short answer A skill file is a plain-text document, conventionally named SKILL.md, that teaches an AI agent one capability: when the skill applies, the procedure to follow, the constraints to respect, and any reference material or scripts it needs. Skill files are loaded into the agent's context when relevant, which keeps capabilities modular, version-controlled and readable by the people who rely on them. - **Conventional filename:** SKILL.md - **Format:** Markdown, often with frontmatter - **Loaded:** When the task matches ## Where the convention came from The pattern was popularised by Anthropic's Agent Skills, which package a capability as a folder containing a SKILL.md file with YAML frontmatter giving a name and a description of when to use it, plus whatever reference documents and scripts the procedure needs. An agent reads the descriptions, and pulls the full file into context only for tasks where the skill applies. The design solves a context problem. Everything an agent might ever need to know cannot fit in one prompt, and stuffing it in degrades performance on every other task. Splitting capabilities into files that load conditionally keeps the always-on instructions short. ## What a good skill file contains - **A trigger description** — When this skill applies and, just as important, when it does not. This is the line the agent reads to decide whether to load the rest. - **The procedure** — Steps in order, written as instructions to a competent colleague rather than as a description of the domain. - **Constraints and prohibitions** — What must never happen. Negative constraints do more work than positive ones, because they are the part a model will otherwise reason its way around. - **Reference material** — Tables, examples, checklists, or scripts the procedure calls. Kept in the same folder so the skill travels as a unit. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | System prompt | Always-on instructions defining identity and behaviour | Loaded on every call. A skill file loads only when the task matches. | | Tool definition | A machine-readable schema describing a function the agent can call | Tells the agent what it can do. A skill file tells it how and when to do it. | | RAG document | Content retrieved from a knowledge base to answer a question | Supplies facts. A skill file supplies procedure. | | Fine-tuning | Changing model weights on training examples | Bakes behaviour into the model. A skill file is editable in a text editor at any time. | ## How Polaris uses skill files When you hire an AI worker in Polaris, the interview at the end of hiring writes the worker's capabilities to a real SKILL.md file rather than a hidden system prompt. You can open it, read exactly what your worker was told, and edit it. When a task is assigned, the runtime compiles the worker's identity from its instructions, its skill files and the task context before starting the tool loop. The reason for exposing the file is auditability. A worker that behaves oddly has a text file behind it that a person can read and correct, which is not true of an agent whose behaviour lives in a prompt you never see. ## Related terms - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/model-context-protocol](https://www.polarishq.co/glossary/model-context-protocol) - [cloud-claude-code/skill-files-for-ai-workers](https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [ai-workers](https://www.polarishq.co/ai-workers) ## Questions people ask **How is a skill file different from a prompt?** A prompt is what you send for one request. A skill file is a durable document that any task matching its trigger will load, version-controlled and editable independently of any single request. The practical difference is that a prompt improves one answer and a skill file improves every future task of that kind. **Can a non-technical person write a skill file?** Yes. It is Markdown, and the hardest part is procedural clarity rather than syntax. The most common failure is writing a description of a topic instead of instructions for performing a task, which produces an agent that knows about the work rather than one that does it. **How many skill files should one agent have?** As many as it has distinct capabilities, since they load conditionally and unused ones cost nothing. The design pressure is towards one file per capability with a sharp trigger description, rather than one large file covering an entire role. ## Related - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/model-context-protocol - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/cloud-claude-code/skill-files-for-ai-workers - https://www.polarishq.co/ai-workers --- --- title: "Model Context Protocol (MCP): definition and how it works" description: "The Model Context Protocol is an open standard from Anthropic for connecting AI applications to tools and data. Definition, architecture, and what MCP is not." url: https://www.polarishq.co/glossary/model-context-protocol section: Glossary updated: 2026-08-21 --- # Model Context Protocol (MCP) One protocol between models and the systems they need, instead of one integration per pair. ## The short answer The Model Context Protocol (MCP) is an open standard, released by Anthropic in November 2024, for connecting AI applications to external tools and data. An MCP server exposes tools, resources and prompts over a defined JSON-RPC interface; an MCP client inside an AI application discovers and calls them. One protocol replaces bespoke per-integration code, so any compliant client can use any compliant server. - **Released:** November 2024 - **Published by:** Anthropic - **Licence:** Open standard, open source - **Transport:** JSON-RPC ## The problem it was designed to solve Before a shared protocol, connecting M AI applications to N data sources meant writing M times N integrations, each with its own authentication, schema and failure behaviour. Every new model host reimplemented the same connectors, and every tool vendor reimplemented the same adapters for each host. MCP turns that into an M plus N problem: implement the protocol once on either side and the combinations come for free. Anthropic published the specification and reference implementations in November 2024 and open-sourced the surrounding SDKs. Adoption spread beyond Anthropic's own products through 2025, and by 2026 MCP is the most widely implemented interface for exposing tools and context to language models. ## The three things an MCP server can expose The specification defines distinct primitives, and mixing them up is the most common implementation mistake. - **Tools** — Functions the model can invoke, with a described input schema. Model-controlled: the model decides when to call them, subject to whatever approval the host enforces. - **Resources** — Data the client can read and attach to context, such as files, records and query results, identified by URI. Application-controlled rather than model-invoked. - **Prompts** — Reusable templates a user can select, typically surfaced in the host application as slash commands or menu entries. User-controlled. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Function calling | A model capability: emitting a structured call against a supplied schema | MCP standardises how tools are discovered, transported and authorised. Function calling is what happens once a tool is in front of the model. | | Plugin API | A vendor-specific extension interface for one product | Tied to one host. MCP is host-neutral by design. | | Agent framework | A library for building agent loops, memory and orchestration | MCP does not run agents. It is a connection layer that agents use. | | API gateway | Infrastructure routing and securing HTTP APIs | Serves general clients. MCP describes capabilities in terms a model can reason about. | ## A concrete example A team wants an assistant to answer questions from their Postgres database. Without MCP, someone writes a bespoke connector for whichever assistant they use, and writes it again when they switch. With MCP, they run a Postgres MCP server that exposes a query tool and the schema as resources. Any MCP-capable client can then be pointed at it, and the same server also serves the coding agent and the internal chat app. Polaris does not require MCP to give an AI worker tool access. Connections come from a curated catalog, are authorised once for the whole organisation, and are stored server-side so workers can use credentials that browsers can never read back. ## Related terms - [glossary/skill-file](https://www.polarishq.co/glossary/skill-file) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [integrations](https://www.polarishq.co/integrations) - [cloud-claude-code/give-an-ai-agent-tool-access](https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access) - [glossary/agent-orchestration](https://www.polarishq.co/glossary/agent-orchestration) ## Questions people ask **Who controls the Model Context Protocol?** Anthropic published the specification in November 2024 and maintains it as an open standard with open-source SDKs and reference servers. It is not proprietary to Anthropic's products: other model providers, IDEs and agent hosts implement the same protocol. **Is MCP a security risk?** It is a connection layer, so it inherits the risk of whatever it connects. The specific concerns are prompt injection through resource content, over-broad tool permissions, and servers from untrusted sources. The mitigations are conventional: least-privilege credentials, human approval for consequential tool calls, and running only servers you have reason to trust. **Do I need MCP to give an AI agent access to my tools?** No. MCP is one way to expose tools to a model, and direct API integrations remain common, particularly where a product curates its own connection catalog. MCP's advantage is portability: the same server works with any compliant client rather than one vendor's. **What does JSON-RPC have to do with it?** JSON-RPC is the message format MCP uses for requests and responses between client and server. It matters to implementers rather than users, but it is why an MCP server can run locally over standard input and output or remotely over HTTP without changing its logic. ## Related - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/agent-orchestration - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/integrations - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access --- --- title: "Agent orchestration: definition and common patterns" description: "Agent orchestration is the coordination of AI agents and their steps: routing, sequencing, state and retries. Definition, the four patterns, and what it is not." url: https://www.polarishq.co/glossary/agent-orchestration section: Glossary updated: 2026-08-21 --- # Agent orchestration Deciding which agent does what, in what order, and what happens when one fails. ## The short answer Agent orchestration is the coordination of AI agents and their steps: routing a task to the right agent or model, sequencing multi-step work, passing state between steps, enforcing limits, handling retries, and managing handoffs between agents and to humans. Orchestration is a control layer above the individual agent loop, and it is where most reliability problems in multi-agent systems are solved or created. - **Layer:** Above the agent loop - **Main patterns:** Single loop, manager, pipeline, graph - **Failure domain:** State, retries, handoffs ## Why it became a separate concern A single agent in a single loop needs no orchestration: it calls tools until it is done. Coordination becomes a distinct layer as soon as one of three things is true: several agents with different specialities are involved, work spans multiple runs or machines, or the run is long enough that failures have to be recovered rather than restarted. The vocabulary was borrowed wholesale from distributed systems and workflow engines, which is appropriate, because the problems are the same ones: idempotency, at-least-once delivery, partial failure, and knowing what state the world was in when something broke. ## The four patterns you will meet - **Single agent, tool loop** — One agent calls tools repeatedly until the goal is met. The simplest thing that works, and correct far more often than the alternatives suggest. - **Manager and sub-agents** — A coordinating agent decomposes the goal and delegates to specialists, then assembles their results. Good for breadth; the coordinator becomes the reliability bottleneck. - **Sequential pipeline** — Fixed stages, each with a defined input and output, such as research, then draft, then review. Predictable and easy to debug, at the cost of flexibility. - **Graph or state machine** — Nodes with explicit transitions and conditions. The most controllable option and the most work to author; the pattern behind most production agent frameworks. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Workflow automation | Event-triggered rules with authored branches | Every path exists before run time. Orchestration coordinates agents that choose paths at run time. | | Multi-agent conversation | Several agents exchanging messages in a shared thread | One technique within orchestration, not a synonym for it. | | Prompt chaining | Feeding one model output into the next prompt | A single-process technique with no scheduling, state store or failure handling. | | Container orchestration | Scheduling containers across machines, as Kubernetes does | Unrelated to agents. The shared word causes real confusion in search results. | ## Where Polaris sits Polaris uses the simplest arrangement that supports the product: an agent_jobs queue in Postgres that a runtime service claims work from, one job per assigned task, running a single agent in a tool loop with retries on failure. There is no manager agent and no graph, because a task assigned to a named worker with written acceptance criteria is already decomposed by the person who wrote it. The queue contract is deliberately runtime-agnostic, so the machine behind it can be swapped without changing the product. ## Related terms - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/model-context-protocol](https://www.polarishq.co/glossary/model-context-protocol) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) ## Questions people ask **Do I need a multi-agent system?** Usually not at first. Multiple agents add coordination overhead, more failure modes and harder debugging, and they pay off mainly when specialities genuinely differ or when work must run in parallel. A single well-briefed agent with the right tools handles most tasks that teams reach for orchestration frameworks to solve. **What breaks most often in orchestrated agent systems?** State handoff between steps. An agent that produced good work in step one hands an ambiguous summary to step two, which acts on a subtly wrong premise. Explicit, structured outputs between stages fix more failures than better prompts do. **Is orchestration the same as an agent framework?** A framework is a library that helps you implement orchestration; orchestration is the design problem the library addresses. You can orchestrate agents with a database table and a worker process, which is what many production systems actually do. ## Related - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/model-context-protocol - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks --- --- title: "Human-in-the-loop (HITL): definition and variants" description: "Human-in-the-loop means a person's judgement is a required step in an automated system. Definition, origin, and the difference from human-on-the-loop." url: https://www.polarishq.co/glossary/human-in-the-loop section: Glossary updated: 2026-08-21 --- # Human-in-the-loop The system cannot complete the loop without a person, by design. ## The short answer Human-in-the-loop (HITL) describes a system design in which a person's judgement is a required step, not an optional one: the process pauses for human input, approval or correction before it can continue or take effect. The term originated in control and simulation engineering, spread to machine learning through data labelling and active learning, and now describes approval gates in AI agent systems. - **Abbreviation:** HITL - **Origin field:** Control and simulation engineering - **Sibling terms:** Human-on-the-loop, human-in-command ## Where the term came from The phrase comes from control engineering and military simulation, where a human-in-the-loop simulation is one that cannot run without a live operator making decisions inside it. Machine learning borrowed it for training pipelines where people label data, correct model outputs, or choose which examples the model should see next. AI agents inherited the term with a narrower meaning: a checkpoint where the agent must stop and get a person's decision before proceeding or before an action takes effect. Regulatory frameworks, including EU AI Act provisions on human oversight of high-risk systems, have since given it legal weight in some contexts. ## Commonly confused with These three are genuinely different levels of oversight, and the distinction matters in policy documents. | Term | The human's position | If the human does nothing | | --- | --- | --- | | Human-in-the-loop | A required step inside the process | Nothing happens. The process is blocked. | | Human-on-the-loop | Monitoring, able to intervene or veto | The system proceeds on its own. | | Human-in-command | Setting policy and scope, not individual decisions | The system operates within the boundaries already set. | | Fully autonomous | Outside the operating loop entirely | The system acts and reports afterwards, if at all. | ## Where the checkpoint usually goes in agent systems - **Before a consequential action** — Sending an external email, moving money, deleting records, publishing. The cost of the action is irreversible, so the approval sits in front of it. - **At delivery** — The agent produces work and a person accepts, rejects or asks for changes. The most common gate, and the cheapest to operate. - **On low confidence** — The agent escalates only when it is uncertain or when the case falls outside its brief. Efficient, and dependent on confidence estimates being honest. - **On a sample** — A percentage of completed work is reviewed to detect drift. Used where per-item review would be more expensive than the work itself. ## The failure mode nobody plans for A review gate that always approves is not oversight, it is a click. Rubber-stamping is well documented in automation research under automation bias: people asked to check machine output repeatedly, and rarely finding a fault, stop checking properly. A gate placed on every trivial output produces exactly that outcome and then fails on the one case that mattered. The practical answer is to place fewer gates on higher-stakes decisions and give the reviewer something to check against: acceptance criteria written before the work started, and a record of what the agent actually did. ## Related terms - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [glossary/autonomous-agent](https://www.polarishq.co/glossary/autonomous-agent) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [glossary/agentic-project-management](https://www.polarishq.co/glossary/agentic-project-management) ## Questions people ask **What is the difference between human-in-the-loop and human-on-the-loop?** In-the-loop means the process cannot continue without the person: if nobody acts, nothing happens. On-the-loop means the person supervises and can intervene, but the system proceeds by default. The distinction decides what happens when a reviewer is asleep, on holiday, or simply not looking. **Does human-in-the-loop slow everything down?** It adds latency at the gate, which is the point. The design question is not whether to have gates but where to put them: on irreversible actions and final deliveries rather than on every intermediate step, so reviewers keep enough attention for the decisions that carry risk. **How does Polaris implement it?** AI workers deliver work as a comment on the task with any generated files attached, and tick their acceptance-criteria checklist as they go. The machine cannot mark a task done. A human closes and rates it, so the completion decision is always a person's. ## Related - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail --- --- title: "Acceptance criteria: definition and how to write them" description: "Acceptance criteria are the conditions a deliverable must meet to be accepted, written before work begins. Definition, agile origin, and use with AI agents." url: https://www.polarishq.co/glossary/acceptance-criteria section: Glossary updated: 2026-08-21 --- # Acceptance criteria Written before the work, checkable after it, and binary either way. ## The short answer Acceptance criteria are the conditions a deliverable must satisfy to be accepted by the person who requested it, written before work begins and evaluated as pass or fail rather than as a matter of opinion. The practice comes from agile requirements work, where criteria attached to a user story define its scope and tell the team unambiguously when the story is finished. - **Origin:** Agile user-story practice - **Written:** Before work starts - **Evaluated as:** Pass or fail ## Where the practice came from Acceptance criteria entered mainstream software practice through extreme programming and the user-story format, where Mike Cohn's conditions of satisfaction described the checks a story had to pass before a customer would accept it. Behaviour-driven development later formalised one common shape, Given-When-Then, and made criteria executable as tests. The reason the practice survived every methodology fashion is that it does one specific job: it moves the argument about scope to before the work, when changing the answer is cheap. ## The two formats in common use - **Checklist** — A list of conditions, each independently verifiable. Best for deliverables that are documents, analyses or artefacts rather than behaviours. - **Given-When-Then** — Given a starting state, when an action occurs, then an observable outcome follows. Best for system behaviour, and directly translatable into automated tests. ## Commonly confused with | Term | Scope | The difference | | --- | --- | --- | | Definition of Done | Every item the team produces | A team-wide standard such as tests written and documentation updated. Acceptance criteria are specific to one item. | | Requirements | The whole system or feature | Describe what to build. Acceptance criteria describe how to check it was built. | | Test cases | Verification procedure | Written after, derived from the criteria, and usually more numerous and more detailed. | | Success metrics | Business outcome after release | Measured later, in production. Criteria are checkable at handover. | ## Why they matter more with AI agents than with people A human colleague who receives a vague brief asks a question. An agent, given the same brief, produces something confidently wrong-shaped and hands it back, having burned the time. Criteria written in advance are the cheapest available fix, because they convert an ambiguous instruction into a set of checks the agent can steer against while working. In Polaris, a task's checklist is handed to the AI worker as its acceptance criteria. The worker ticks each item the moment it is genuinely complete, so the reviewer opens a delivery with a visible record of which conditions the agent believes it met, and can disagree with any of them before closing the task. ## Related terms - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [glossary/agentic-project-management](https://www.polarishq.co/glossary/agentic-project-management) ## Questions people ask **How many acceptance criteria should one task have?** Enough to remove ambiguity and few enough to read at a glance, which in practice is usually three to seven. A task needing fifteen criteria is normally two or three tasks that have not been separated yet. **Who writes the acceptance criteria?** The person requesting the work, because they are the person who will accept or reject it. The performer can propose additions and should challenge criteria that are untestable, but the requester owns the definition of accepted. **What makes an acceptance criterion bad?** Any criterion that cannot be checked without a debate. Fast, clear, user-friendly and well-researched all fail the test. Replace each with an observable condition: a named source count, a stated response time, a specific section that must exist. ## Related - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work --- --- title: "Work log: definition and what belongs in one" description: "A work log is a per-job record of what was done: steps, sources, artefacts and time. Definition, what belongs in one, and why it makes a bill challengeable." url: https://www.polarishq.co/glossary/work-log section: Glossary updated: 2026-08-21 --- # Work log The record that makes an unwatched run reviewable afterwards. ## The short answer A work log is a per-job record of what was done to complete a unit of work: the steps taken, the sources consulted, the artefacts produced, and the time attributed to each. In agent systems a work log is the primary audit surface, because nobody watched the run, and it is the evidence a reviewer or a payer checks a claim against. - **Granularity:** One entry per job - **Primary purpose:** Audit and billing - **Written by:** The runtime, not the model ## General usage and Polaris usage In general usage, a work log is any chronological record of work performed: the timesheet, the engineering logbook, the case notes. Professions that bill by the hour have kept them for a century, for exactly one reason: a client who cannot see what was done has no way to check what was charged. Polaris uses the term in that older billing sense rather than as a generic activity feed. Each AI worker has a work log listing every job it completed, what the run consisted of, and the human-equivalent hours attributed to it, which is what the bill is computed from. ## What belongs in an agent work log - **The observable effort** — How many searches were run, how much finished prose was written, how many checklist items were ticked, how many comments posted, how many files produced. Countable things, not the model's own account of its diligence. - **The artefacts** — Every file the run produced, attached and retrievable, so a reader can check the claim against the output. - **The attributed time** — The estimate the billing formula produced from that effort, shown per job rather than rolled up into a monthly total. - **Failures and retries** — Runs that errored, and how many attempts were made. A log showing only successes is a marketing artefact. ## Commonly confused with | Term | What it records | The difference | | --- | --- | --- | | Activity feed | Discrete events: assigned, status changed, run started | Event-shaped and system-wide. A work log is job-shaped and effort-shaped. | | Chat transcript | Every message in a conversation | Complete but unstructured, and far too long to audit routinely. | | Audit log | Security-relevant actions and who performed them | Answers who touched what. A work log answers what was produced and what it cost. | | Timesheet | Hours a person claims to have spent | Self-reported. An agent work log is computed from observable effort. | ## Why it decides whether usage billing is trustworthy Any usage-priced product faces one question from a buyer: how do I know you are not inflating the meter? The only durable answer is a record detailed enough that a customer can recompute the number, and a formula published clearly enough that they can check the arithmetic. That is why Polaris shows the human-hours estimate job by job on the worker's work log and publishes the formula behind it. Any line on the bill can be challenged from the log entry that produced it. ## Related terms - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [glossary/usage-based-pricing](https://www.polarishq.co/glossary/usage-based-pricing) - [glossary/delivery-comment](https://www.polarishq.co/glossary/delivery-comment) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [cloud-claude-code/ai-agent-audit-trail](https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) ## Questions people ask **Who writes the work log, the AI or the system?** The runtime writes it, from counters it maintains while the job runs. Letting the model report its own effort would make the log an assertion rather than a measurement, and the whole point of the artefact is that it is not the agent's opinion of its work. **How long should work logs be kept?** At least as long as the billing period they support, and in practice longer, because disputes and quality reviews arrive after the fact. Teams using agent output in regulated work usually align retention with whatever their record-keeping obligations already require. **Can a work log be challenged?** In Polaris, yes: each job's hour estimate is shown on the worker's work log next to the effort it was computed from, and the formula is published. A line you disagree with can be disputed from the entry itself rather than from an aggregate invoice. ## Related - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/cloud-claude-code/ai-agent-audit-trail - https://www.polarishq.co/cost/what-an-ai-worker-costs --- --- title: "Human-equivalent hours: definition and formula" description: "Human-equivalent hours measure how long delivered work would have taken a person, not how long the machine ran. Definition, the Polaris formula and its limits." url: https://www.polarishq.co/glossary/human-equivalent-hours section: Glossary updated: 2026-08-21 --- # Human-equivalent hours A billing unit denominated in the work replaced, not the compute consumed. ## The short answer Human-equivalent hours are a unit of measurement that expresses delivered work in terms of how long it would have taken a competent person, rather than how long a machine ran or how many tokens it consumed. The unit exists because machine runtime is meaningless to a buyer comparing an agent against hiring, and because faster hardware should not make the same delivered work cheaper to describe. - **Measures:** Work delivered, not compute used - **Polaris rate:** ~$2 per human-hour - **Clamp:** 5 minutes to 8 hours per job - **Estimated from:** Observable effort ## Why the unit exists Every other billing unit for AI work is denominated in something the buyer does not care about. Tokens measure text volume. GPU-seconds measure hardware. Seats measure headcount. None of them answer the question a manager is actually asking, which is whether this is cheaper than the alternative way of getting the work done. Human-equivalent hours answer that question directly, and they have an uncomfortable property that makes them honest: if the model gets faster, the bill does not fall, because the work replaced has not changed. Nobody would design this unit to flatter a vendor. ## The Polaris formula, in full Polaris estimates each job from observable effort. The formula is published so it can be checked. - **Base** — 15 minutes for a delivery job, 10 minutes for a review job. The fixed cost of picking up a piece of work and understanding it. - **Research** — 12 minutes per web search performed during the run. - **Writing** — Finished prose counted at 90 characters per minute, which is a normal composing-and-editing rate rather than a typing speed. - **Checklist and conversation** — 8 minutes per acceptance-criteria item ticked, 10 minutes per comment posted. - **Artefacts** — 10 minutes per file produced, plus 5 minutes for each generated PDF. - **Clamp** — The total is bounded to a minimum of 5 minutes and a maximum of 8 hours per job, so no single run can produce an absurd line on the bill. ## Commonly confused with | Unit | What it measures | Who it favours | | --- | --- | --- | | Human-equivalent hours | Work delivered, in human time | Buyer: faster models do not raise the bill | | Compute time | Wall-clock machine runtime | Neither, reliably: it tracks hardware, not value | | Tokens | Text processed in and out | Vendor: verbose work costs more regardless of usefulness | | Full-time equivalent (FTE) | Headcount capacity over a period | Different purpose entirely: workforce planning, not billing a job | ## The honest limitations This is an estimate, not a measurement of a real person doing the same task. The counters are objective, but the conversion rates are judgements: twelve minutes per search and ninety characters per minute are defensible averages, not laws. Work heavy on thinking and light on output is undercounted by a formula that counts output. The mitigation is transparency rather than precision. The formula is published, every job's estimate is itemised on the worker's work log next to the effort it came from, and any line can be challenged from that entry. ## Related terms - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [glossary/usage-based-pricing](https://www.polarishq.co/glossary/usage-based-pricing) - [glossary/per-seat-pricing](https://www.polarishq.co/glossary/per-seat-pricing) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) ## Questions people ask **Is a human-equivalent hour the same as an hour of machine time?** No, and they are usually far apart. A job that a machine finishes in minutes may represent an hour or more of human-equivalent work, because the unit describes the work delivered rather than the time the hardware was busy. **How can an estimate be fair if nobody did the work by hand?** It cannot be exact, and Polaris does not claim it is. What makes it checkable is that the inputs are counted rather than asserted, the conversion rates are published, and each job's estimate appears next to the effort that produced it so a customer can recompute and dispute it. **What does an hour cost?** Polaris bills roughly two dollars per human-equivalent hour delivered, with no charge for the software itself and no charge when nothing is delivered. Unlimited humans, tasks, workstreams and docs are included at no cost. ## Related - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/delivery-comment - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/per-seat-vs-usage-pricing --- --- title: "Usage-based pricing: definition, models and trade-offs" description: "Usage-based pricing charges by a consumption metric rather than a fixed subscription. Definition, origin in cloud infrastructure, the variants and trade-offs." url: https://www.polarishq.co/glossary/usage-based-pricing section: Glossary updated: 2026-08-21 --- # Usage-based pricing The bill follows consumption, which cuts both ways for buyer and vendor. ## The short answer Usage-based pricing is a model in which the amount a customer pays is determined by a consumption metric such as API calls, gigabytes stored, messages sent, tasks completed or hours delivered, rather than by a fixed subscription fee. Charges are normally metered continuously and billed in arrears, so a customer who uses nothing in a period pays nothing or only a small platform fee. - **Also called:** Consumption-based, pay-per-use - **Origin:** Cloud infrastructure, mid-2000s - **Billed:** In arrears, on a meter ## Where it came from and why it spread Metered billing is old, since utilities have charged this way for a century, but its arrival in software came with cloud infrastructure in the mid-2000s, when Amazon Web Services made per-hour compute and per-gigabyte storage normal. Developer tools followed, then communications and payments, then everything with a countable unit. OpenView's State of Usage-Based Pricing survey found adoption among SaaS companies rising sharply through the early 2020s, reporting that around 45% of surveyed companies had adopted some form of usage-based model by 2021. Later surveys disagree on the exact figure, partly because hybrid models, a platform fee plus a meter, blur the definition. ## The variants, which are not interchangeable - **Pure pay-per-use** — No commitment, no floor. Pay for what the meter records. Rare in enterprise software because the vendor's revenue becomes hard to forecast. - **Hybrid** — A platform or seat fee plus a meter for consumption above an included allowance. The most common shape in practice. - **Prepaid credits** — Buy a balance, draw it down. Popular with AI products because it caps the customer's exposure and improves the vendor's cash position. Unused credits often expire, which is where trust is lost. - **Outcome-based** — Charging per completed unit of work rather than per resource consumed: per resolved ticket, per delivered hour. Aligns price with value and requires a definition of done both sides accept. ## The honest trade-offs **What buyers gain and lose** - No cost for users or capacity that go unused - Cost scales with value received rather than headcount - Budgeting is harder: the bill varies month to month - Fear of runaway usage discourages exploration unless caps exist **What vendors gain and lose** - Low barrier to starting, since small usage costs little - Revenue expands automatically as customers grow - Forecasting is harder and churn is quieter, since usage drops before anyone cancels - The meter must be legible, or every invoice becomes a support ticket ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Pay-as-you-go seats | Monthly rather than annual per-user billing | Still per seat. Only the commitment length changed, not the unit. | | Freemium | A free tier with paid upgrades | A packaging decision about who pays, not a decision about what the meter counts. | | Tiered pricing | Fixed price bands with feature or volume limits | The price is a step function chosen in advance rather than a continuous meter. | | Overage charges | Fees for exceeding a plan's allowance | A meter bolted onto a subscription, usually priced punitively rather than proportionally. | ## Related terms - [glossary/per-seat-pricing](https://www.polarishq.co/glossary/per-seat-pricing) - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/what-an-ai-worker-costs](https://www.polarishq.co/cost/what-an-ai-worker-costs) - [cost](https://www.polarishq.co/cost) ## Questions people ask **Is usage-based pricing cheaper than a subscription?** It depends entirely on usage relative to the subscription's break-even point. Light and uneven users almost always pay less; heavy consistent users can pay more. The reliable difference is not the total but the shape: cost tracks activity instead of headcount. **How do buyers protect themselves from a surprise bill?** Spending caps, alerts at defined thresholds, and a meter they can inspect in near real time. A vendor that cannot show consumption as it accrues is asking for trust that the invoice will not justify at the end of the month. **Why do AI products favour it?** Because their marginal costs are real. Every inference costs the vendor money, so a flat subscription either overcharges light users or loses money on heavy ones. Metering passes the underlying cost structure through instead of averaging it across everyone. **What makes a usage meter trustworthy?** Three properties: the unit is something the buyer recognises as valuable, the calculation is published rather than proprietary, and each charge is itemised against the activity that produced it so it can be disputed individually. ## Related - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/what-an-ai-worker-costs - https://www.polarishq.co/cost/cost-of-ai-subscriptions --- --- title: "Per-seat pricing: definition and where it breaks" description: "Per-seat pricing charges a fixed fee per named user per period regardless of use. Definition, origin, why it dominates B2B SaaS, and where it breaks." url: https://www.polarishq.co/glossary/per-seat-pricing section: Glossary updated: 2026-08-21 --- # Per-seat pricing The bill tracks headcount, which is only a proxy for value received. ## The short answer Per-seat pricing is a model in which a customer pays a fixed fee for each named user per billing period, regardless of how much any of them uses the product. The model descends from per-user software licensing, and it dominates business software because it is simple to quote, simple to forecast, and grows automatically as a customer's headcount grows. - **Also called:** Per-user pricing - **Unit:** A named user per period - **Vendor benefit:** Predictable, forecastable revenue ## Why it became the default Per-user licensing predates SaaS: on-premise software was sold in user counts long before anything was rented monthly. Subscription software inherited the unit because it solved a real problem. Pricing needed a number that a buyer could count, a salesperson could quote, and a finance team could forecast, and headcount is the only such number every company already knows. The model is also the reason collaboration software is priced the way it is. Value in a workspace tool rises with the number of people in it, so the seat is a rough proxy for value delivered, and a legitimate one for most of the product's history. ## Where the proxy breaks - **Occasional users** — The finance manager who opens the tracker twice a month costs the same as the engineer living in it. Teams respond by not buying seats for people who should have them, which quietly damages the product's usefulness. - **Multiplication across tools** — Four per-seat tools at $10 to $20 each is $40 to $80 per person per month before anyone has done any work. The unit is the same in each, so the count multiplies rather than adds. - **Growth taxes collaboration** — Adding a contractor, a client or a temporary reviewer has a price, so organisations exclude people from the system of record to avoid the line item. - **AI does not fit the unit** — AI features cost the vendor money per use, not per user, so they arrive as per-seat add-ons that overcharge light users and undercharge heavy ones. The unit stopped matching the cost structure. ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Per active user | Charges only for users who used the product in the period | A meter on activity. Rarer, because it makes vendor revenue less predictable. | | Concurrent-user licensing | A pool of simultaneous sessions shared across many people | Common in on-premise and specialist tools; the seat is not tied to a person. | | Tiered pricing | Price bands by feature set or company size | Often combined with seats. Tiers decide what you get; seats decide how many pay for it. | | Platform fee | A flat charge for the account regardless of users | Headcount-independent. Sometimes paired with a usage meter in hybrid models. | ## The alternative Polaris uses Polaris charges nothing per seat and nothing for the software: unlimited humans, tasks, workstreams and docs, with the Chief of Staff included in every organisation. Revenue comes from delivered work at roughly two dollars per human-equivalent hour, itemised on a work log you can challenge. The consequence worth stating plainly is that adding a person to Polaris costs nothing, so there is no financial reason to keep anyone outside the system of record. ## Related terms - [glossary/usage-based-pricing](https://www.polarishq.co/glossary/usage-based-pricing) - [glossary/tool-sprawl](https://www.polarishq.co/glossary/tool-sprawl) - [glossary/human-equivalent-hours](https://www.polarishq.co/glossary/human-equivalent-hours) - [cost/per-seat-vs-usage-pricing](https://www.polarishq.co/cost/per-seat-vs-usage-pricing) - [cost/stack-cost-10-person-team](https://www.polarishq.co/cost/stack-cost-10-person-team) - [cost](https://www.polarishq.co/cost) ## Questions people ask **Why do most SaaS companies still price per seat?** Because it is legible and forecastable for both sides. A buyer can compute next year's bill from next year's hiring plan, and a vendor can model revenue from customer headcount. Those are real advantages that no consumption meter matches. **What is the true cost of a per-seat stack?** Multiply the number of people by the sum of the per-seat prices of every tool they need, then add the AI add-ons that are now sold on top of each. A ten-person team on four mid-tier tools is typically paying several hundred dollars a month before a single task is completed. **Are AI seats the same as software seats?** Structurally yes, economically no. A software seat costs the vendor almost nothing to serve, so an unused seat is pure margin. An AI seat carries real inference cost per use, which is why per-seat AI pricing is unstable and why so many AI products have moved to credits or meters. ## Related - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/glossary/human-equivalent-hours - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/cost/per-seat-vs-usage-pricing - https://www.polarishq.co/cost/stack-cost-10-person-team - https://www.polarishq.co/cost/cost-of-ai-subscriptions --- --- title: "Tool sprawl: definition, causes and real costs" description: "Tool sprawl is the accumulation of overlapping software until no tool is the system of record. Definition, causes, costs, and how it differs from shadow IT." url: https://www.polarishq.co/glossary/tool-sprawl section: Glossary updated: 2026-08-21 --- # Tool sprawl The expensive part is not the licences. It is that nothing is authoritative any more. ## The short answer Tool sprawl is the accumulation of overlapping software across an organisation to the point where no single tool is the system of record for a given kind of work. Symptoms include the same information maintained in several places, licences nobody uses, decisions recorded where the people affected will not find them, and staff who cannot say which tool is authoritative. - **Apps per company, 2025:** 101 average (Okta) - **Main cost:** Fragmented context, not licences - **Distinct from:** Shadow IT ## How much software a company actually runs Okta's Businesses at Work 2025 report, drawn from its own customer base, put the average number of applications per organisation at 101, the first time the figure crossed one hundred, after sitting below ninety from 2019 through 2023. That is the identity-provider view, so it counts anything anyone signed into, which is exactly the population that matters here. The number is a symptom rather than the disease. Ninety of those applications can coexist happily. Sprawl is what happens in the small subset that overlap: three places to write a document, two places to track a task, four places a decision could plausibly have been recorded. ## What it actually costs Licence spend is the easiest cost to measure and rarely the largest. - **Context fragmentation** — The requirement is in one tool, the discussion in another, the task in a third, the file in a fourth. Reconstructing what was decided becomes archaeology, and the reconstruction is done by the most expensive people in the company. - **Duplicate maintenance** — The same roadmap kept in two systems drifts within a fortnight. Whichever copy someone reads is now a coin flip. - **Per-seat multiplication** — Each overlapping tool charges per person, so cost scales with headcount times tool count rather than with work done. - **Onboarding tax** — A new hire must learn not just the tools but the unwritten convention about which one is authoritative for what. That convention is rarely written down, because it is not a decision anyone made. ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Shadow IT | Software procured without IT approval | A cause of sprawl, not the same thing. Fully approved tools sprawl too. | | SaaS sprawl | Unmanaged growth in subscription count and spend | Framed around procurement and cost. Tool sprawl is framed around overlapping function. | | Technical debt | Accumulated shortcuts in a codebase | Lives in the product. Tool sprawl lives in how the company works. | | Integration debt | Growing web of connections between systems | Often the attempted cure for sprawl, and a cost centre of its own. | ## The two ways out, honestly stated Consolidation means moving work into fewer products, which lowers the bill and restores a single system of record, at the cost of losing whatever the specialist tools were genuinely better at. Integration means keeping the tools and connecting them, which preserves capability but adds a maintenance surface and never quite makes any one place authoritative. Polaris takes the consolidation position for tasks, docs and team chat, and the integration position for everything else: Slack, Notion, Linear, GitHub, Gmail, Google Calendar, Google Drive, Figma, HubSpot, Stripe, Supabase, WhatsApp and Instagram are in the connection catalog, so a team can move the system of record without abandoning the tools around it. ## Related terms - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) - [glossary/per-seat-pricing](https://www.polarishq.co/glossary/per-seat-pricing) - [glossary/workstream](https://www.polarishq.co/glossary/workstream) - [replace](https://www.polarishq.co/replace) - [cost/stack-cost-25-person-team](https://www.polarishq.co/cost/stack-cost-25-person-team) - [integrations](https://www.polarishq.co/integrations) ## Questions people ask **How do I know if my team has tool sprawl?** Ask five colleagues where a given decision from last quarter is recorded and see whether the answers agree. Disagreement is the diagnostic. Counting subscriptions measures spend; that question measures whether anything is authoritative. **Is consolidating always the right answer?** No. Specialist tools exist because general ones handle some jobs badly, and forcing a design team out of Figma or an engineering team off a workflow engine they depend on trades one problem for a worse one. The tools worth consolidating are the overlapping ones, not the good ones. **Does AI make tool sprawl better or worse?** Worse before better, on current evidence. Most teams have added AI subscriptions on top of their existing stack rather than replacing anything, which increases both the bill and the number of places where work happens without leaving a shared record. ## Related - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/glossary/usage-based-pricing - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/replace - https://www.polarishq.co/cost/stack-cost-25-person-team - https://www.polarishq.co/cost/cost-of-ai-subscriptions --- --- title: "All-in-one workspace: definition and trade-offs" description: "An all-in-one workspace combines docs, tasks and chat in one product with one data model. Definition, origin, and the suite-versus-best-of-breed trade-off." url: https://www.polarishq.co/glossary/all-in-one-workspace section: Glossary updated: 2026-08-21 --- # All-in-one workspace One data model behind documents, tasks and conversation, instead of three products and a pile of integrations. ## The short answer An all-in-one workspace is a single product that covers the work categories a team would otherwise buy separately (documents, tasks and project tracking, and often team communication) under one data model, one permission model and one search index. The category's defining claim is that context stays connected because the work never leaves the system, not that each individual feature is best in class. - **Category popularised:** Late 2010s - **Core claim:** One data model, connected context - **Classic objection:** Best-of-breed beats the suite ## Where the category came from Bundled work software is not new, since Microsoft Office and Lotus Notes made the argument decades ago, but the phrase all-in-one workspace attached itself to a wave of products in the late 2010s, Notion most prominently, that merged the document and the database into one editable surface. The pitch was that a page could be a document, a table, a task list or all three, so information did not have to be filed into a category before it could be written down. The category has since widened to include products that add task tracking, chat, and now AI agents to the same surface. What all of them share is the underlying architectural claim: one schema, so a task can live inside a document and a document can be attached to a project without an integration in between. ## Suite versus best-of-breed, without the marketing **What the suite genuinely wins** - One place to search, one permission model to reason about - Context stays attached: the discussion, the decision and the task are the same object graph - One bill and one onboarding path instead of four - No integration to maintain between the parts that matter most **What best-of-breed genuinely wins** - Depth in specialist workflows: Jira's workflow engine, Figma's canvas, Linear's cycle discipline - Faster iteration on one problem instead of five - Lower switching cost when one component disappoints - Teams already know the tools, and retraining has a real price ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Productivity suite | Bundled applications sharing an account and file store | Separate applications with separate data models. Bundling is commercial, not architectural. | | Integrated stack | Separate tools connected by integrations | Data is synchronised between systems, so each copy can drift and each connection can break. | | Digital workplace | Umbrella term for a company's whole set of work tools | Describes a portfolio, not a product. | | Intranet | Internal publishing and communication portal | One-to-many distribution rather than collaborative work. | ## What Polaris adds to the category Polaris covers the usual three (tasks, docs and team chat) with the Focus lane as the home screen rather than a project list. The addition is that AI workers are members of the workspace on the same terms as people, so the roster that gets assigned work contains both. The pricing consequence matters as much as the feature one. The software is free with no seats, so the usual suite argument about consolidating four subscriptions into one becomes consolidating four subscriptions into zero, with billing only for work an AI worker delivers. ## Related terms - [glossary/tool-sprawl](https://www.polarishq.co/glossary/tool-sprawl) - [glossary/workstream](https://www.polarishq.co/glossary/workstream) - [glossary/focus-lane](https://www.polarishq.co/glossary/focus-lane) - [glossary/per-seat-pricing](https://www.polarishq.co/glossary/per-seat-pricing) - [alternatives](https://www.polarishq.co/alternatives) - [replace](https://www.polarishq.co/replace) ## Questions people ask **Are all-in-one workspaces worse at each individual job?** Usually, on the specialist axes. A suite rarely beats a focused tool at that tool's hardest problem. The trade is whether connected context and one bill are worth more to a given team than depth in workflows most of them do not use. **When is an all-in-one workspace the wrong choice?** When a specialist workflow is load-bearing for the business: complex approval routing, regulated audit requirements, or a design or engineering practice built around one tool's specific model. Those teams are better served by keeping the specialist and consolidating everything around it. **What is the migration cost?** Mostly conventions rather than data. Exports and imports handle documents and tasks reasonably well; what does not transfer is the accumulated habit of where things go, which is why staged moves by workstream tend to work better than a single cutover weekend. ## Related - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/glossary/ai-teammate - https://www.polarishq.co/alternatives - https://www.polarishq.co/replace --- --- title: "Task queue: definition and why agents need one" description: "A task queue is a durable list of work that producers append and workers claim, decoupling a request from its execution. Definition and use in agent systems." url: https://www.polarishq.co/glossary/task-queue section: Glossary updated: 2026-08-21 --- # Task queue The thing that lets a request survive the process that made it. ## The short answer A task queue is a durable list of units of work that producers append to and worker processes claim from, decoupling the moment work is requested from the moment it is executed. Queues give a system retries after failure, control over how many jobs run at once, tolerance for bursts of demand, and a record of work that outlives the process that requested it. - **Field:** Distributed systems - **Common implementations:** SQS, Celery, Sidekiq, Postgres - **Key guarantee:** Work survives a crash ## What a queue actually buys you The naive alternative is doing the work inside the request that asked for it. That works until the work takes longer than a request should, until the process restarts mid-way, or until a thousand requests arrive at once. A queue converts the work into a durable record, so the request returns immediately and the execution becomes someone else's problem, specifically a worker's. Everything else queues are praised for follows from durability. Retries are possible because the job still exists after a failure. Concurrency limits are possible because a fixed number of workers claim from one list. Backpressure is visible because the queue length is a number you can watch. ## The properties that matter when choosing one - **Delivery guarantee** — At-least-once is the norm, which means jobs can run twice and handlers must be idempotent. Exactly-once is mostly a marketing claim about a system that does deduplication for you. - **Claiming** — How a worker takes exclusive ownership of a job. Usually a conditional update or a visibility timeout, so two workers cannot claim the same item. - **Retry and backoff** — How many attempts, how long between them, and where a job goes when it has exhausted them, normally a dead-letter queue somebody actually monitors. - **Durability** — Whether an enqueued job survives a restart of the broker. In-memory queues are fast and lose work; database-backed queues are slower and do not. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Message bus / pub-sub | Broadcast of events to many subscribers | Fan-out to any number of listeners. A task queue delivers each job to exactly one worker. | | Cron | Time-triggered execution of a fixed command | Triggered by the clock, not by demand, and with no per-item state. | | Event stream | An ordered, replayable log of events | Consumers track their own position and can replay history. A queue item is claimed and removed. | | Thread pool | In-process concurrency across worker threads | Lives and dies with the process. Nothing survives a restart. | ## Why agent systems depend on them Agent work has exactly the shape a queue is built for: it takes minutes rather than milliseconds, it fails in ways worth retrying, it must not run twice on the same task, and it has to keep going after the person who requested it closes their laptop. Running an agent inside a web request is the single most common architectural mistake in early agent products. Polaris uses a Postgres table called agent_jobs as its queue. A runtime service polls it, claims a job with a conditional update so two machines cannot take the same one, retries a failed job up to two further attempts, and writes results back to the same database the frontend reads. The queue contract is deliberately runtime-agnostic, so the machine behind it can be replaced without touching the product. ## Related terms - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/agent-orchestration](https://www.polarishq.co/glossary/agent-orchestration) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [cloud-claude-code/long-running-agent-tasks](https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks) - [cloud-claude-code/claude-code-when-laptop-is-closed](https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed) ## Questions people ask **Can a database table be a task queue?** Yes, and for most workloads it is the right choice. A Postgres table with a status column and a conditional claim gives durability, retries and exact ordering with no extra infrastructure. Dedicated brokers earn their operational cost at throughputs most products never reach. **What happens if a worker dies mid-job?** That is what claiming with a timeout is for. The job stays in the queue with a claimed timestamp, and if it is not completed within the timeout it becomes claimable again. This is exactly why handlers must tolerate running twice. **Do task queues make agents slower?** They add a small scheduling delay, typically seconds, and remove the far larger risk of losing a multi-minute job to a restart. For work measured in minutes, the trade is not close. ## Related - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/agent-orchestration - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/autonomous-agent - https://www.polarishq.co/cloud-claude-code/long-running-agent-tasks - https://www.polarishq.co/cloud-claude-code/claude-code-when-laptop-is-closed --- --- title: "Agent runtime: definition and what it is responsible for" description: "An agent runtime is the infrastructure that executes an AI agent: loading context, running the tool loop and enforcing limits. Definition and boundaries." url: https://www.polarishq.co/glossary/agent-runtime section: Glossary updated: 2026-08-21 --- # Agent runtime Not the model, not the framework: the thing that actually runs the job. ## The short answer An agent runtime is the infrastructure that executes an AI agent's work: it loads the agent's identity and task context, runs the loop of model calls and tool calls, enforces time and iteration limits, handles failures, and records what happened. The runtime is distinct from the model that reasons, the framework used to write the agent, and the orchestrator that decides which agent runs. - **Responsibility:** Execution, limits, recording - **Not the same as:** Model, framework, orchestrator - **Polaris runtime:** polaris-runtime on Fly.io ## The four layers people collapse into one Conversations about agent architecture get confused because four separate things are all called the agent. The model provides reasoning and lives behind an API. The framework is the library you wrote the agent with. The orchestrator decides which agent handles what and in what order. The runtime is the process that actually executes a single job on real hardware with a real filesystem and real credentials. The distinction is practical rather than pedantic. Almost every operational failure, whether a job that hung, a run that cost too much or a credential that leaked into a log, belongs to the runtime layer, and cannot be fixed by changing model or framework. ## What a runtime is responsible for - **Compiling identity and context** — Assembling the agent's instructions, its skill files and the specific task's context into the input for the run, before the first model call. - **Running the loop** — Alternating model calls and tool calls until the agent finishes or a limit is reached, with a bounded number of rounds so a confused agent cannot loop forever. - **Enforcing limits** — Wall-clock timeouts, iteration caps, and the boundary of which tools and credentials this run may use. - **Recording the run** — Counting the observable effort, posting progress, attaching produced files, and writing the result back to durable storage. Nobody watched the run, so the record is the only account of it. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Model / LLM | The reasoning engine behind an API | Stateless per call. It cannot claim a job, write a file, or enforce a timeout. | | Agent framework | A library for defining agents, tools and loops | Development-time. The runtime is what is running in production at three in the morning. | | Orchestrator | The layer choosing which agent handles what | Decides assignment and sequence. The runtime executes one job to completion. | | Cloud development environment | A remote environment for a human developer | Assumes a person is typing in it. A runtime assumes nobody is watching. | ## How Polaris implements it Polaris runs a worker service on Fly.io called polaris-runtime. It polls the agent_jobs queue, claims one job, compiles the worker's identity from its instructions and skill files plus the task context, then runs a tool loop with live web search and a bounded number of rounds under a wall-clock timeout. During the run it ticks acceptance-criteria items, posts progress comments, attaches generated files, and computes the human-equivalent hours for the work log. Credentials for connected tools are stored server-side and used by the runtime. A browser never sees them, which is the security reason for putting execution on a machine rather than in the client. ## Related terms - [glossary/task-queue](https://www.polarishq.co/glossary/task-queue) - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/agent-orchestration](https://www.polarishq.co/glossary/agent-orchestration) - [glossary/skill-file](https://www.polarishq.co/glossary/skill-file) - [glossary/cloud-development-environment](https://www.polarishq.co/glossary/cloud-development-environment) - [cloud-claude-code/run-claude-code-in-the-cloud](https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud) ## Questions people ask **Is the agent runtime the same as the model?** No. The model is a stateless reasoning service reached over an API; the runtime is a process on a machine that calls it repeatedly, holds the filesystem and credentials, enforces limits and records results. Swapping models changes the quality of the reasoning and nothing about the operational behaviour. **Why does an agent need its own machine?** Because it needs a filesystem to write to, credentials that must not be exposed to a browser, network access for tools, and a process that keeps running when the requester's laptop closes. None of those survive in a client-side session. **Can the runtime be swapped?** It depends on whether the product talks to a queue contract or to the runtime directly. Polaris deliberately defines the job contract in the database, so the service that claims jobs can be replaced without changing anything in the application. ## Related - https://www.polarishq.co/glossary/task-queue - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/agent-orchestration - https://www.polarishq.co/glossary/skill-file - https://www.polarishq.co/glossary/cloud-development-environment - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/cloud-claude-code/run-claude-code-in-the-cloud --- --- title: "Autonomous agent: definition and degrees of autonomy" description: "An autonomous agent pursues a goal by choosing its own actions rather than following a script. Definition, origin, degrees of autonomy, and what it is not." url: https://www.polarishq.co/glossary/autonomous-agent section: Glossary updated: 2026-08-21 --- # Autonomous agent Autonomy is a range, and the interesting question is where the boundary sits. ## The short answer An autonomous agent is a system that pursues a goal by deciding its own sequence of actions, rather than executing a procedure authored in advance. Autonomy is a matter of degree, bounded in practice by the tools the agent can use, the limits placed on its run, and the approval points where a human decision is required before it can proceed. - **Research origin:** Multi-agent systems, 1990s - **Property described:** Decision latitude - **Bounded by:** Tools, limits, approval gates ## The term is older than the current wave Autonomous agents were a research field long before language models. Wooldridge and Jennings' 1995 survey of intelligent agents in the Knowledge Engineering Review set out the properties still cited today: autonomy, reactivity, proactiveness and social ability. Robotics, simulation and distributed AI used the vocabulary for decades. What changed after 2022 is not the definition but the substrate. Deciding what to do next used to require hand-built planners and domain models; a language model with tool access does it directly from a natural-language goal. The old definition survived the change intact, which is why the research usage and the product usage still line up. ## Degrees of autonomy in practice Most product arguments about whether something is truly autonomous are arguments about which row it occupies. | Level | What the system decides | What a human decides | | --- | --- | --- | | Scripted | Nothing. It follows the authored path | Every branch, in advance | | Tool-selecting | Which tool to call next, within a fixed procedure | The procedure and its order | | Goal-pursuing | Its own steps, tools and stopping point | The goal, the criteria, and acceptance of the result | | Self-directed | Which goals to pursue at all | Policy and scope, if anything | ## Commonly confused with | Term | What it describes | The difference | | --- | --- | --- | | Headless agent | The interface: no chat window, triggered programmatically | An orthogonal property. Agents can be headless and scripted, or interactive and highly autonomous. | | Automation | Executing a defined procedure without human effort | The path is authored. Autonomy is about choosing the path. | | AGI | General intelligence across arbitrary domains | A capability claim about breadth. Autonomy is a design property about latitude within a task. | | Self-improving system | A system that modifies its own capabilities | A different axis entirely, and not implied by autonomy. | ## Where the boundary is usually drawn Production systems almost never sit at the self-directed end, and the reason is accountability rather than capability. An agent that picks its own goals has no natural point at which a person is answerable for the outcome. So the common design is goal-pursuing autonomy inside a task, with the goal and the acceptance criteria set by a human and the result accepted by one. Polaris sits exactly there. An AI worker chooses its own approach to a task, runs its own searches and decides when the criteria are met, and it cannot mark the task done. A person closes it and rates the work. ## Related terms - [glossary/headless-agent](https://www.polarishq.co/glossary/headless-agent) - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [glossary/agent-orchestration](https://www.polarishq.co/glossary/agent-orchestration) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [cloud-claude-code/assign-work-to-an-ai-agent](https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent) ## Questions people ask **Is an autonomous agent the same as a headless agent?** No. Autonomy describes how much latitude the agent has over its own actions; headless describes whether it has a user interface. The two vary independently, and treating them as synonyms is the most frequent mistake made with both terms. **How autonomous should an agent be?** As autonomous as the reversibility of its actions allows. Drafting a document, running research and producing a file are cheap to undo and can be left to the agent. Sending external messages, moving money and deleting records are not, and belong behind an approval step regardless of how capable the agent is. **Who is responsible for what an autonomous agent does?** The organisation that deployed it, in every legal framework currently in force. That is the practical reason approval gates and audit records exist: responsibility does not transfer to the software, so the humans holding it need a way to see and control what was done. ## Related - https://www.polarishq.co/glossary/headless-agent - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/agent-orchestration - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/cloud-claude-code/assign-work-to-an-ai-agent --- --- title: "Delivery comment: definition and why agents deliver this way" description: "A delivery comment is the comment an AI worker posts on a task carrying the finished work and files. Polaris usage, why the pattern exists, and what it is not." url: https://www.polarishq.co/glossary/delivery-comment section: Glossary updated: 2026-08-21 --- # Delivery comment Work arrives where the task already lives, attributed and reviewable, and the task stays open. ## The short answer A delivery comment is the comment an AI worker posts on a task that carries the finished work, including any generated files, and hands it to a person for review. The pattern keeps the deliverable attached to the request that produced it, attributed to the worker that did it, and open for correction, because posting a comment does not close the task. - **Term type:** Polaris usage - **Carries:** Finished work and files - **Does not:** Close the task ## Polaris usage, stated plainly Delivery comment is Polaris's name for a specific mechanic rather than an industry-standard term. When an AI worker finishes a task, the runtime posts its output as a comment on that task, with generated files attached, and leaves the task open in whatever state it was in. The alternative designs all lose something. Delivering into a separate outputs area breaks the link between the request and the result. Delivering by email leaves the workspace. Delivering by marking the task done removes the review step and makes the agent the judge of its own work. ## What the pattern buys - **Context stays attached** — The brief, the acceptance criteria, the progress notes and the deliverable are all on one object, so a reviewer arriving a week later has everything without reconstructing anything. - **Attribution is automatic** — The comment has an author, and the author is the worker. Six months later it is still clear which work was produced by whom. - **Correction is a reply** — Asking for changes is a comment on the same thread rather than a new task, and the worker's next run has the whole conversation as context. - **The close stays human** — A delivered comment is a proposal, not a completion. Someone reads it, decides, and closes. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Progress comment | A short note posted mid-run about what the worker is doing | Informational, posted during the job. A delivery comment carries the finished work. | | Activity event | A system record: assigned, status changed, run started | Generated by the system about state, not written content with an author. | | Status change | Moving the task to done or in review | A state transition. In Polaris the AI worker cannot perform the transition to done. | | Attachment | A file added to a task | The payload rather than the delivery. A delivery comment usually contains attachments and the explanation of them. | ## The rule underneath it Agents deliver; humans close. That single sentence is the reason the delivery comment exists in this shape, and it is why the runtime has no code path to mark a task done. It also sets the reviewer's job clearly. Read the comment, check it against the acceptance criteria the worker ticked, look at the work log if the estimate looks wrong, then close and rate, or reply and ask for another pass. ## Related terms - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [glossary/human-in-the-loop](https://www.polarishq.co/glossary/human-in-the-loop) - [glossary/work-log](https://www.polarishq.co/glossary/work-log) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [cloud-claude-code/review-and-approve-agent-work](https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work) - [cloud-claude-code/ai-agents-that-produce-files](https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files) ## Questions people ask **Why not just mark the task done when the agent finishes?** Because that makes the agent the judge of its own work, and removes the only reliable quality gate in the system. Delivering as a comment keeps the task open until a person has read the output and decided it is acceptable. **What if the delivered work is wrong?** Reply on the same task with what needs to change. The worker's next run has the original brief, the criteria, its own previous output and your correction, which is normally a far better input than starting a fresh task would be. **Can a delivery comment include files?** Yes. Generated files, including documents and PDFs, are attached to the comment, so the artefact and the explanation of it arrive together on the task that requested them. ## Related - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/glossary/human-in-the-loop - https://www.polarishq.co/glossary/work-log - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/ai-teammate - https://www.polarishq.co/cloud-claude-code/review-and-approve-agent-work - https://www.polarishq.co/cloud-claude-code/ai-agents-that-produce-files --- --- title: "Workstream: definition in project and product management" description: "A workstream is a continuous strand of work with its own owner and outputs. Definition, origin in programme management, and why Polaris calls them buckets." url: https://www.polarishq.co/glossary/workstream section: Glossary updated: 2026-08-21 --- # Workstream A strand of work that keeps going, rather than a project that ends. ## The short answer A workstream is a continuous strand of related work within a larger effort, with its own owner, its own outputs and its own rhythm. The term comes from programme management, where a programme is divided into parallel streams that each run to their own schedule. In everyday use it covers anything ongoing that a team wants to see grouped: a project, an area of responsibility, or a client. - **Origin:** Programme management - **Distinguishing feature:** Ongoing, not time-boxed - **Polaris name in UI:** Workstreams, also called buckets ## Why teams reach for the word Project implies a start, an end and a deliverable. Most work in a company has none of those: keeping the help centre current, running recruiting, handling one client's account. Calling these projects forces an artificial end date onto work that will still be happening next year, and teams notice the mismatch even when they cannot name it. Workstream describes the ongoing thing directly. In programme management the meaning is narrower, covering parallel strands within a defined programme, each with a lead, but the looser usage has spread because it fills a gap the project vocabulary left open. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Project | A time-boxed effort with a defined end and deliverable | Ends. A workstream continues until the responsibility itself ends. | | Programme | A group of related projects managed together | The parent. Workstreams are typically the strands inside a programme. | | Epic | A large body of work broken into stories | A backlog-sizing construct that eventually completes. | | Team | A group of people | Describes who. A workstream describes what, and one team can run several. | ## How Polaris uses the term Polaris calls them workstreams in the interface and buckets in conversation. Both mean the same container. - **Any container of work** — A project, an area or a client. There is no separate concept for each; one container type handles all three. - **Lanes shared between views** — The lanes inside a workstream are the same whether you look at it as a list or as a board, so a team organises once and chooses its view afterwards. - **Not the home screen** — The Focus lane is the home screen, not the workstream list. Workstreams are where work lives; Focus is what you have committed to next. - **Created when work needs a home** — The Chief of Staff will propose a workstream when tasks are accumulating without one, rather than leaving them loose in an inbox. ## Related terms - [glossary/focus-lane](https://www.polarishq.co/glossary/focus-lane) - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) - [glossary/tool-sprawl](https://www.polarishq.co/glossary/tool-sprawl) - [glossary/acceptance-criteria](https://www.polarishq.co/glossary/acceptance-criteria) - [use-cases](https://www.polarishq.co/use-cases) - [alternatives](https://www.polarishq.co/alternatives) ## Questions people ask **What is the difference between a workstream and a project?** A project is time-boxed and ends when its deliverable is complete. A workstream is ongoing and ends only when the responsibility itself goes away. A launch is a project; keeping the documentation current is a workstream. **How many workstreams should a team have?** Few enough that someone can name them all without looking. When the list stops being memorable it has usually become a filing system rather than a map of what the team is responsible for, and the fix is merging rather than nesting. **Do AI workers belong to a workstream?** In Polaris, no. Workers belong to the workspace roster and can be assigned tasks in any workstream, in the same way a colleague is not confined to one project. ## Related - https://www.polarishq.co/glossary/focus-lane - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/glossary/ai-teammate - https://www.polarishq.co/glossary/acceptance-criteria - https://www.polarishq.co/use-cases - https://www.polarishq.co/alternatives --- --- title: "Focus lane: definition and how it differs from a to-do list" description: "A focus lane is a short, pinned list of non-negotiable work across one time horizon. Polaris usage, the reasoning behind it, and why it is not a priority field." url: https://www.polarishq.co/glossary/focus-lane section: Glossary updated: 2026-08-21 --- # Focus lane A commitment for a horizon, not a filter over everything you have. ## The short answer A focus lane is a short, deliberately limited list of work a person has committed to within a chosen time horizon, pinned above everything else they could be doing. It differs from a filtered view or a priority field because membership is a decision rather than a sort order: something enters the lane only when a person puts it there, and the lane is meant to stay short. - **Term type:** Polaris usage - **Horizons:** Today, this week, next 30 days - **Position:** Pinned first in every view ## The problem it exists to solve Priority fields fail in a predictable way. Everything drifts upward, high stops meaning anything, and the sorted list is the full backlog in a different order. A person opening it still has to decide what to do, which is the decision the field was supposed to help with. A focus lane makes the decision the act of adding. Putting something in the lane is a commitment, and the lane is short enough that adding a fourth item forces a look at the other three. The constraint is the feature. ## How Polaris implements it - **Three horizons** — Today, this week, and the next thirty days. A person picks the horizon their work actually runs at, and the Focus screen shows that one. - **Pinned first, everywhere** — The focus lane appears at the top of every view rather than as a separate page you have to remember to open. - **The home screen** — Polaris opens on Focus rather than on a workstream list, because the first question of a working day is what am I committed to, not which projects exist. - **A flag on the task** — Focus membership is a property of the task, so an item stays itself when it enters or leaves the lane and keeps its workstream, its lane and its history. ## Commonly confused with | Term | What it is | The difference | | --- | --- | --- | | Priority field | A rank or level stored on every task | Sorts everything. A focus lane holds a small chosen subset. | | Sprint backlog | Work a team commits to for an iteration | Team-level, fixed-length and ceremonial. A focus lane is personal and continuous. | | Saved filter | A query over tasks matching criteria | Membership is computed. Focus membership is decided by a person. | | Kanban WIP limit | A cap on items in progress in a column | Constrains flow through a board. A focus lane constrains attention across everything. | ## Related terms - [glossary/workstream](https://www.polarishq.co/glossary/workstream) - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [alternatives/todoist](https://www.polarishq.co/alternatives/todoist) - [alternatives/linear](https://www.polarishq.co/alternatives/linear) - [use-cases](https://www.polarishq.co/use-cases) ## Questions people ask **How is a focus lane different from marking tasks high priority?** Priority is a field on every task and drifts upward until it carries no information. A focus lane is a short list you add to deliberately, so its length is a signal in itself and a full lane is a prompt to finish something before starting more. **How many items should be in a focus lane?** Few enough to hold in your head, which for most people is three to five for a day and slightly more for a week. If the lane needs scrolling it has stopped being a commitment and become a second backlog. **Can AI workers have tasks in a focus lane?** Yes. Work assigned to an AI worker can sit in the focus lane like any other task, which is often how a person tracks a delivery they are waiting on and intend to review that day. ## Related - https://www.polarishq.co/glossary/workstream - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/alternatives/todoist - https://www.polarishq.co/alternatives/linear - https://www.polarishq.co/use-cases --- --- title: "Passwordless authentication: definition and methods" description: "Passwordless authentication verifies identity without a user-chosen password. Definition, the methods including passkeys and email codes, and why it is not MFA." url: https://www.polarishq.co/glossary/passwordless-authentication section: Glossary updated: 2026-08-21 --- # Passwordless authentication Removing the shared secret the user has to remember, and attackers have to steal. ## The short answer Passwordless authentication verifies a user's identity without a password they choose and remember. Instead it relies on possession of a registered device or key, control of an email account or phone number, or a biometric check on hardware the user already holds. Removing the password removes the credential most commonly reused, phished, guessed and leaked in breaches. - **Standard behind passkeys:** FIDO2 / WebAuthn - **Common methods:** Passkeys, magic links, email codes - **Not the same as:** MFA or SSO ## Why passwords are the weak link Passwords fail because of how people use them, not because the cryptography is broken. They are reused across sites, so one breach compromises many accounts. They are typed into convincing fake pages. They are stored by services that later leak them. Every mitigation invented for these problems, from complexity rules to rotation policies to security questions, pushed users towards worse habits. Passwordless methods change the shape of the attack rather than hardening the secret. A passkey cannot be phished onto a lookalike domain because the browser binds it to the origin. An emailed code cannot be reused elsewhere because it exists for one login and expires. ## The main methods, with their honest weaknesses - **Passkeys (FIDO2 / WebAuthn)** — A key pair stored on the device or in a platform keychain, unlocked by biometrics or a device PIN. The strongest widely available option, and phishing-resistant by design. The recovery story when a device is lost is still where most deployments struggle. - **Magic links** — A one-time sign-in link emailed to the user. Simple to build; inherits the security of the mailbox, and links are frequently mangled or pre-fetched by corporate email scanners. - **Emailed or texted codes** — A short one-time code entered into the app. Works everywhere with no device registration; SMS delivery is vulnerable to SIM-swap attacks, which is why email is often preferred. - **Authenticator app codes** — Time-based codes generated on a device. Common as a second factor rather than a sole one, and still phishable in real time by a proxy. ## Commonly confused with | Term | What it means | The difference | | --- | --- | --- | | Multi-factor authentication | Requiring two or more distinct factors | MFA adds factors alongside a password. Passwordless replaces the password rather than supplementing it. | | Single sign-on (SSO) | One identity provider authenticating many applications | About where authentication happens, not how. SSO can itself be password-based or passwordless. | | Two-factor authentication | A specific case of MFA with exactly two factors | Usually password plus code. Still a password system. | | Social login | Signing in through Google, GitHub or similar | A form of federated identity. Whether it is passwordless depends on how that provider authenticates you. | ## How Polaris does it Polaris uses emailed one-time codes and has no password path at all. A user enters an email address, receives a code, and signs in. There is nothing to reset, nothing to reuse elsewhere and no password database to breach. The trade-off is stated rather than hidden: access to the mailbox is access to the account, so the security of a Polaris workspace rests on the security of its members' email. Teams already running strong protection on their email domain inherit it directly. ## Related terms - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) - [glossary/agent-runtime](https://www.polarishq.co/glossary/agent-runtime) - [glossary/tool-sprawl](https://www.polarishq.co/glossary/tool-sprawl) - [integrations](https://www.polarishq.co/integrations) - [cloud-claude-code/give-an-ai-agent-tool-access](https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access) - [glossary](https://www.polarishq.co/glossary) ## Questions people ask **Is passwordless authentication more secure than a password plus MFA?** Passkeys are, because they are bound to the origin and cannot be handed to a lookalike site. Email codes and magic links are roughly comparable to password-plus-code systems: they remove reuse and leaked-database risk while making the mailbox the single point of compromise. **What happens if someone loses access to their email?** In an email-code system, that is the recovery problem, and it is why account recovery deserves as much design attention as sign-in. Most products fall back to an administrator re-inviting the user, which places the trust in the workspace owner rather than in a security question. **Why do products drop passwords entirely rather than offering both?** Keeping a password path preserves every weakness it was meant to remove, since attackers simply use the weaker route. Removing it also removes the password reset flow, the storage of hashes, and an entire class of support requests. ## Related - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/glossary/agent-runtime - https://www.polarishq.co/glossary/tool-sprawl - https://www.polarishq.co/glossary/per-seat-pricing - https://www.polarishq.co/integrations - https://www.polarishq.co/cloud-claude-code/give-an-ai-agent-tool-access --- --- title: "Generative engine optimization (GEO): definition and origin" description: "Generative engine optimization is structuring content so AI answer engines cite it. Definition, the 2023 paper that named it, and how it differs from SEO." url: https://www.polarishq.co/glossary/generative-engine-optimization section: Glossary updated: 2026-08-21 --- # Generative engine optimization Optimising to be quoted inside an answer rather than ranked beneath one. ## The short answer Generative engine optimization (GEO) is the practice of structuring content so that generative search systems such as ChatGPT, Claude, Perplexity and Google AI Overviews retrieve it, quote it and attribute it in their answers. The unit of optimisation is the passage rather than the page, because these systems select and cite spans of text rather than ranking whole documents. - **Term coined:** 2023, Aggarwal et al. - **Published at:** KDD 2024 - **Unit optimised:** The passage, not the page ## Where the term came from GEO was named in a research paper: "GEO: Generative Engine Optimization" by Aggarwal and colleagues, posted to arXiv as 2311.09735 in November 2023 and presented at KDD 2024. The authors built a benchmark of around ten thousand queries and tested which content changes increased a source's visibility in generated answers. The reported results favoured adding citations to named sources, adding statistics, and writing with an authoritative rather than hedged tone, with the largest single effect around a 40% relative lift in visibility. Keyword stuffing measured as actively harmful, unlike in classic search where it is merely useless. Those numbers were measured on 2023 and 2024 systems and should be read as directional rather than as guarantees, since every underlying model has changed since. ## What the practice actually involves - **Self-contained answer blocks** — A definition or direct answer near the top of the page that survives being lifted out with no surrounding context: named subject, no unresolved pronouns, complete in itself. - **Structure a retriever can segment** — Question-shaped headings, a strict heading hierarchy, short paragraphs, and tables for anything comparative, since tables are lifted close to verbatim. - **Attributable facts** — Named, dated, linked sources for non-obvious claims. Engines cite the origin of a number, which is why publishing original data is the one advantage a competitor cannot copy by writing better prose. - **Entity clarity** — One canonical name used consistently, plus structured data that lets a system resolve who published the page and what the page is about. ## Commonly confused with | Term | The goal | What success looks like | | --- | --- | --- | | SEO | Rank a page in a list of links | Position on a results page, measured in clicks | | GEO | Be quoted and cited inside a generated answer | Appearing as a named source in an answer, often with no click | | AEO / answer engine optimization | Largely the same goal as GEO | A near-synonym, used more in marketing than in research | | Content marketing | Attract and persuade an audience | Engagement and conversion, independent of retrieval mechanics | ## The uncomfortable part GEO succeeds by being cited, and a citation frequently arrives with no visit. A page can be the source behind thousands of answers and show flat traffic, which makes conventional analytics a poor instrument for judging whether any of it worked. Measurement currently means asking the engines directly, on a schedule, and recording what they say and what they cite. The methods that work are also, awkwardly, just good writing: state the claim clearly, source it, structure it so a reader can find it. The techniques that game the mechanism, including bulk-generated pages and inauthentic mention farming, are the ones both search and answer engines are explicit about penalising. ## Related terms - [glossary/all-in-one-workspace](https://www.polarishq.co/glossary/all-in-one-workspace) - [glossary/ai-worker](https://www.polarishq.co/glossary/ai-worker) - [ai-workers/seo-specialist](https://www.polarishq.co/ai-workers/seo-specialist) - [use-cases/marketing/seo-content-production](https://www.polarishq.co/use-cases/marketing/seo-content-production) - [use-cases/marketing](https://www.polarishq.co/use-cases/marketing) - [glossary](https://www.polarishq.co/glossary) ## Questions people ask **Is GEO different from SEO?** The goals differ. SEO aims to rank a page in a list of links; GEO aims to have a passage quoted and attributed inside a generated answer. The techniques overlap substantially, because both reward clear structure and genuine authority, but the unit of optimisation shifts from the page to the passage. **Does llms.txt help?** There is no public evidence that any major AI system consumes llms.txt as a ranking or retrieval input, and none of the large providers has committed to supporting it. Publishing one is cheap and harmless; treating it as a visibility strategy is not supported by anything measurable today. **How do you measure GEO?** By querying the target engines on a defined schedule with the questions your buyers ask, and recording verbatim whether you were mentioned and what was cited. Referral analytics undercount badly, because most citations are read without a click. **What matters most for being cited?** Publishing something nobody else has. Engines cite the origin of a statistic, so original data, benchmarks or teardowns make you the source that downstream articles point back at. Structural work makes a page quotable; original facts make it worth quoting. ## Related - https://www.polarishq.co/glossary/ai-worker - https://www.polarishq.co/glossary/all-in-one-workspace - https://www.polarishq.co/glossary/agentic-project-management - https://www.polarishq.co/ai-workers/seo-specialist - https://www.polarishq.co/ai-workers/content-writer - https://www.polarishq.co/use-cases/marketing/seo-content-production - https://www.polarishq.co/use-cases/marketing