What is an AI agent?
Also known as: autonomous agent, LLM agent
An AI agent is a software system that uses a language model to pursue a goal with some autonomy, combining instructions, access to data or tools, and the ability to take actions on a user's behalf. Agents range from vendor assistants such as Microsoft Copilot agents to custom builds on any platform. Because an agent reaches data and acts, its permissions, grounding, and ownership are the core governance concerns.
The word “agent” spans a lot of ground, from a scripted assistant that answers questions to a system that plans and executes multi-step tasks against live systems. What they share is the combination that makes them useful and risky at once: a model, access to data or tools, and permission to act.
That combination is why an agent is a governance object, not just a feature. Whatever data it can reach, it can surface or act on. So governing agents, on any platform, comes down to knowing what each one can access, who owns it, and whether that access is still warranted.
