Glossary
AI agent
An AI agent is software given a goal and a set of tools, which takes actions toward that goal and reacts to the results, rather than only producing text.
The four parts
- Reasoning — a model interpreting the goal and choosing a next step.
- Planning — breaking the goal into steps and adapting when one fails.
- Memory — short-term for the conversation, long-term so it does not start from nothing.
- Tools — the connections that let it act outside the chat.
Against neighbouring terms
A chatbot answers and cannot act. An assistant helps a person with a task they are actively doing. An agent is given a job and can work when nobody is watching.
The practical test
Can it do something you would have to undo? If yes, it is an agent, and it needs an approval boundary and an audit trail. If no, most agent concerns do not apply.
Related terms
FAQ
What is an AI agent?
Software given a goal and a set of tools, which takes actions toward that goal and reacts to the results rather than only producing text. Every agent has four parts: reasoning, planning, memory and tool use.
What is the difference between an AI agent and a chatbot?
A chatbot answers from a script or document set and cannot act. An agent is given a job and tools and changes things outside the conversation. The test is whether it can do something you would have to undo.