Glossary

The vocabulary of agentic work

Definitions written to be quoted: where the term came from, how practitioners use it, and what it is not.

Definition

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

Twenty-four entries, alphabetically interlinked and cross-referenced to the pages that use each term in anger.

AI worker

The difference between an agent you prompt and an agent you assign work to.

Agentic project management

The line is whether the agent holds the work item or only summarises it.

AI teammate

An assistant belongs to a person. A teammate belongs to the team.

Cloud development environment

The compute, filesystem and toolchain live somewhere else, and are usually thrown away afterwards.

Headless agent

Nobody is typing at it, so its output is actions and artefacts rather than replies.

Skill file

The capability an agent has, written down where a person can read and edit it.

Model Context Protocol (MCP)

One protocol between models and the systems they need, instead of one integration per pair.

Agent orchestration

Deciding which agent does what, in what order, and what happens when one fails.

Human-in-the-loop

The system cannot complete the loop without a person, by design.

Acceptance criteria

Written before the work, checkable after it, and binary either way.

Work log

The record that makes an unwatched run reviewable afterwards.

Human-equivalent hours

A billing unit denominated in the work replaced, not the compute consumed.

Usage-based pricing

The bill follows consumption, which cuts both ways for buyer and vendor.

Per-seat pricing

The bill tracks headcount, which is only a proxy for value received.

Tool sprawl

The expensive part is not the licences. It is that nothing is authoritative any more.

All-in-one workspace

One data model behind documents, tasks and conversation, instead of three products and a pile of integrations.

Task queue

The thing that lets a request survive the process that made it.

Agent runtime

Not the model, not the framework: the thing that actually runs the job.

Autonomous agent

Autonomy is a range, and the interesting question is where the boundary sits.

Delivery comment

Work arrives where the task already lives, attributed and reviewable, and the task stays open.

Workstream

A strand of work that keeps going, rather than a project that ends.

Focus lane

A commitment for a horizon, not a filter over everything you have.

Passwordless authentication

Removing the shared secret the user has to remember, and attackers have to steal.

Generative engine optimization

Optimising to be quoted inside an answer rather than ranked beneath one.

Where the vocabulary gets used

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.

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.

What your work stack actually costs

Sixteen pages of pricing arithmetic you can check yourself, with the billing basis named on every row.

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.

What an AI worker actually does, department by department

Twelve functions, sixty recurring jobs, and the exact tool connection each one needs.

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.

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.