Agentic AI

Agentic AI refers to AI systems that operate with a degree of autonomy: they break a goal into steps, choose actions, observe the results, and adapt. Instead of a single request and response, an agentic system runs a loop of reasoning and tool use until it decides the task is done.

The term captures a shift in the risk model. When an AI plans and acts across many steps, the security question is no longer just what it says, but what it does at each step, in what order, and with what data. Agentic systems need controls on every action, correlation across the whole run, and a record that survives after the fact.