AI agent

An AI agent is a system that receives a goal and decides on its own the steps to achieve it: it queries data, writes emails, updates a CRM, places orders or calls other agents. The difference from a chatbot is that it acts rather than only answering.

The AI Act has no dedicated chapter for agents: the same rules apply as to any AI system, depending on how it is used. If an agent takes part in recruitment or credit decisions, it is high-risk. What changes is practical management, because risks grow with autonomy.

The basics for governing them: an inventory of agents and what they can touch, clear authorisation limits (what they can do alone and what needs approval), an escalation procedure to a person, and an audit trail of every action.