For agencies
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.
Next, for an agency
The cost model in detail, the production roles, and the pages your delivery team will care about.
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.
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.
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.
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.
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.
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.
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
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.
Polaris for design studios
Nobody has ever put a line item for asset renaming on an invoice, and every studio pays for it anyway.
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.
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.
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.
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.
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.
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.