What is an AI agent?

An AI agent is a program built around an AI model, usually a large language model, that is given a goal rather than a single instruction. It breaks the goal into steps, decides what to do next, uses connected tools such as email, a calendar, or a spreadsheet, looks at the result, and continues until the task is done or it needs help. A chatbot answers; an agent also acts.

For example, an agent in a sales team might be asked to prepare for the next day's customer calls. It could read the calendar, look up each company in the CRM (the customer relationship management system where contacts and deals are stored), summarize recent emails with each contact, and draft a one-page brief for every meeting. A person still reviews the briefs, but the gathering and first draft happen without anyone clicking through each system.

Agents differ from fixed workflows. A workflow follows the same steps every time, while an agent chooses its steps as it goes. That flexibility helps with messy tasks, but it also makes agents less predictable. They can misread a goal, pick the wrong tool, or carry an early mistake through every later step, so an agent is only as reliable as the model's judgment at each point.

Because agents can take real actions, such as sending messages or changing records, it is common to limit what they are allowed to touch and to require human approval for anything that is hard to undo. Starting with read-only access and small, low-risk tasks makes it easier to see how an agent behaves before trusting it with more.

An example

An agent reads new support emails, looks up each customer's order status, and drafts replies for a person to check before sending.