Simple definition
An AI agent combines a model, instructions, and optional tools to interpret goals and carry out actions.
Purpose and use
An AI agent uses a model and optional tools to pursue an objective over multiple steps.
How does it work?
It can choose a next action from context and observed results within limits enforced by the executing application.
Example and verification
For an operational task, verify completion criteria, authorized tools and behavior with missing or contradictory data.
Points to watch
An agent can repeat errors or act in the wrong context. Permissions and stop conditions require control independent of explanation quality.
In short
Useful autonomy depends on a verifiable objective, bounded scope and observable outcomes.