Glossary — questions and answers

Plain definitions for the vocabulary of agentic work, written to be quoted.

Each question links to the page that answers it. The answers are not repeated here — they belong to the page that earns them.

80 questions across 25 pages · browse the Glossary pages →

• How often is this glossary checked?

  The vocabulary of agentic work

• Are these definitions written to favour Polaris?

  The vocabulary of agentic work

• Can these definitions be quoted?

  The vocabulary of agentic work

• What is the difference between an AI agent and an AI worker?

  The vocabulary of agentic work

• Is an AI worker the same as an AI agent?

  AI worker

• Do AI workers replace headcount?

  AI worker

• What stops an AI worker from doing the wrong thing?

  AI worker

• How is an AI worker billed?

  AI worker

• Is agentic project management a real methodology or a marketing label?

  Agentic project management

• Does agentic project management replace Scrum or Kanban?

  Agentic project management

• What is the main risk?

  Agentic project management

• Is AI teammate just a rebrand of AI assistant?

  AI teammate

• Do AI teammates need their own accounts?

  AI teammate

• How does Polaris implement this?

  AI teammate

• Is a cloud development environment the same as running Claude Code in the cloud?

  Cloud development environment

• Do CDEs cost more than local development?

  Cloud development environment

• What is the main limitation?

  Cloud development environment

• Is a headless agent the same as an autonomous agent?

  Headless agent

• Can a headless agent ask a human a question?

  Headless agent

• What are headless agents good for?

  Headless agent

• How is a skill file different from a prompt?

  Skill file

• Can a non-technical person write a skill file?

  Skill file

• How many skill files should one agent have?

  Skill file

• Who controls the Model Context Protocol?

  Model Context Protocol (MCP)

• Is MCP a security risk?

  Model Context Protocol (MCP)

• Do I need MCP to give an AI agent access to my tools?

  Model Context Protocol (MCP)

• What does JSON-RPC have to do with it?

  Model Context Protocol (MCP)

• Do I need a multi-agent system?

  Agent orchestration

• What breaks most often in orchestrated agent systems?

  Agent orchestration

• Is orchestration the same as an agent framework?

  Agent orchestration

• What is the difference between human-in-the-loop and human-on-the-loop?

  Human-in-the-loop

• Does human-in-the-loop slow everything down?

  Human-in-the-loop

• How does Polaris implement it?

  Human-in-the-loop

• How many acceptance criteria should one task have?

  Acceptance criteria

• Who writes the acceptance criteria?

  Acceptance criteria

• What makes an acceptance criterion bad?

  Acceptance criteria

• Who writes the work log, the AI or the system?

  Work log

• How long should work logs be kept?

  Work log

• Can a work log be challenged?

  Work log

• Is a human-equivalent hour the same as an hour of machine time?

  Human-equivalent hours

• How can an estimate be fair if nobody did the work by hand?

  Human-equivalent hours

• What does an hour cost?

  Human-equivalent hours

• Is usage-based pricing cheaper than a subscription?

  Usage-based pricing

• How do buyers protect themselves from a surprise bill?

  Usage-based pricing

• Why do AI products favour it?

  Usage-based pricing

• What makes a usage meter trustworthy?

  Usage-based pricing

• Why do most SaaS companies still price per seat?

  Per-seat pricing

• What is the true cost of a per-seat stack?

  Per-seat pricing

• Are AI seats the same as software seats?

  Per-seat pricing

• How do I know if my team has tool sprawl?

  Tool sprawl

• Is consolidating always the right answer?

  Tool sprawl

• Does AI make tool sprawl better or worse?

  Tool sprawl

• Are all-in-one workspaces worse at each individual job?

  All-in-one workspace

• When is an all-in-one workspace the wrong choice?

  All-in-one workspace

• What is the migration cost?

  All-in-one workspace

• Can a database table be a task queue?

  Task queue

• What happens if a worker dies mid-job?

  Task queue

• Do task queues make agents slower?

  Task queue

• Is the agent runtime the same as the model?

  Agent runtime

• Why does an agent need its own machine?

  Agent runtime

• Can the runtime be swapped?

  Agent runtime

• Is an autonomous agent the same as a headless agent?

  Autonomous agent

• How autonomous should an agent be?

  Autonomous agent

• Who is responsible for what an autonomous agent does?

  Autonomous agent

• Why not just mark the task done when the agent finishes?

  Delivery comment

• What if the delivered work is wrong?

  Delivery comment

• Can a delivery comment include files?

  Delivery comment

• What is the difference between a workstream and a project?

  Workstream

• How many workstreams should a team have?

  Workstream

• Do AI workers belong to a workstream?

  Workstream

• How is a focus lane different from marking tasks high priority?

  Focus lane

• How many items should be in a focus lane?

  Focus lane

• Can AI workers have tasks in a focus lane?

  Focus lane

• Is passwordless authentication more secure than a password plus MFA?

  Passwordless authentication

• What happens if someone loses access to their email?

  Passwordless authentication

• Why do products drop passwords entirely rather than offering both?

  Passwordless authentication

• Is GEO different from SEO?

  Generative engine optimization

• Does llms.txt help?

  Generative engine optimization

• How do you measure GEO?

  Generative engine optimization

• What matters most for being cited?

  Generative engine optimization