Autonomous agent

An autonomous agent perceives events, makes decisions and executes tasks independently, rather than waiting for someone to type a request. It's defined by three things: triggers (what wakes it up — a schedule, a new email, a changed record), instructions (what it should do and why), and guardrails (what it must never do).

The difference from ordinary automation is judgement. A rule-based automation follows a path someone drew in advance; an autonomous agent is given a goal and works out its own route, including for cases nobody scripted. In practice the most reliable designs split the work — the agent reasons and decides, while a deterministic flow performs the mechanical execution.

Because an autonomous agent acts when nobody is watching, the security decisions matter more than the AI ones: validate that triggers are authentic, scope the agent's permissions tightly, keep humans in the loop for consequential actions, and log everything it does.