Definition
An AI system that does not just answer questions but takes actions in the world — browsing websites, writing and running code, sending emails, calling external services — often in a chain of steps, with little or no human approval at each step. Unlike a simple chatbot, an agent pursues a goal by deciding on its own what steps to take.
Why it matters
The more autonomously an AI agent acts, the greater the potential damage if it is manipulated, compromised, or simply wrong. Agents operating with access to business systems can cause real-world harm — deleting data, making purchases, or leaking secrets — at machine speed.