Stockertim Dictionary · Artificial intelligence · Technology
AI agent
Also called AI agents, agentic AI, autonomous agent
In short
An AI agent is an AI system that plans and carries out tasks of several steps by itself, choosing which permitted tools to use and when, in pursuit of a goal a person has set.
- Intermediate
- Technology
- Checked
In detail
An assistant replies and leaves the next move to a person. A workflow follows a path fixed in advance. An agent is given a goal and decides its own steps within limits: it plans, calls a tool such as a search, a database or an email draft, observes the result, checks whether it is closer to the goal and chooses what to do next. The loop repeats until the goal is met, a limit is reached or it stops to ask a person.
At its core an agent is a language model wrapped in software that supplies tools, keeps track of progress and enforces rules. The model proposes each action; the surrounding software checks whether the action is allowed, runs it and feeds back the result. How much the agent may do without approval is a design choice, set action by action.
Agents fail differently from assistants. A wrong answer in a chat is text a person can ignore; a wrong step by an agent can be an email sent, a record changed or money moved. Errors can also compound over many steps, and outside content the agent reads can try to steer it. Agent designs therefore lean heavily on permissions, approval points, logs and testing.
Worked example
A coordinator at Lagoon Logistics asks an agent why a shipment is late.
- The agent looks up the order and finds the shipment number.
- It queries the tracking system and sees the container held at the port.
- It searches the port's notices, finds a customs backlog and a new release date.
- It drafts a reply to the customer and stops, because sending requires a staff member's approval.
The answer is assembled in minutes from three systems, and a person still decides what reaches the customer.
In practice
Organisations use agents for research tasks, IT and support tickets, data gathering across systems and drafting, usually starting with read-only tools and approval before any action with outside effect.
Watch out
- Not every chat assistant is an agent; the test is who chooses the next step.
- Giving an agent more tools is not the same as making it more reliable.
Learn it properly
Official sources
- Guidelines for secure AI system development · National Cyber Security Centre
- Top 10 for large language model applications · OWASP Foundation
- International AI Safety Report 2025 · UK Government
Facts last checked on 3 November 2026.
Education, not advice: the Stockertim Dictionary explains what terms mean. It never recommends buying or selling anything.