What is an AI agent?
An agent is a system that can plan steps towards a goal, use tools and check the results.
What is it?
An AI agent is a system that plans its own steps towards a goal, uses tools such as a browser, calendar or files, and checks the results. It does not just answer a question; it tries to see a task through. In doing so it can also make mistakes with real consequences.
An everyday example
An agent can draft an email; sending it should stay under your control.
Why does it matter?
Agents move AI from conversation to action: bookings, purchases, programming. That increases both the benefit and the risk.
A common misconception
That an agent knows what it must not do. More autonomy needs more oversight, narrower permissions and human approval for important steps.
In more detail
How is it different from a chatbot?
A chatbot can only answer. An agent can, for example, read a document, use a calendar or prepare a draft action.
Does an agent work entirely on its own?
How independent it is depends on its rules and permissions. For important actions, human approval makes sense.
What are tools?
Software functions or services the agent uses to search, calculate, read files or change data.
What is the biggest risk?
A mistake or a malicious instruction can turn into an action, not just a wrong sentence.
Remember: More autonomy calls for more oversight and narrower permissions.
Reliable sources
Last reviewed 1 November 2026. An explainer, not professional, legal or medical advice. For rules and figures that change, check the official version.