Also known as: agent
In plain English
A chatbot answers questions; an AI agent gets things done. Ask it to book a meeting and it can check calendars, find a time and send the invite itself.
In practice
Agents suit multi-step work like research, form filling or triaging tickets across systems. Start with narrow tasks, limited permissions and clear approval points, because an agent's mistakes are actions, not just wrong words.
Under the hood
An agent typically runs a loop: the model reasons about the goal, selects a tool call, observes the result and decides the next step, until it finishes or hits a limit. Key design elements are tool definitions, memory, planning, error handling and stopping conditions.
Example
"The agent opened the ticket, checked the logs and suggested a fix."