An AI agent combines a language model with tools it can use, such as searching documents, reading email, calling an API, updating a spreadsheet, or operating an application. Given a goal, it decides which steps to take, observes the results, and continues until the task is done or it needs help.
Because agents act, they need more governance than chatbots. Good agent systems define exactly which tools are available, apply least-privilege permissions, set cost and time budgets, log every action, and require human approval for consequential steps such as sending messages or making payments.