AI agent

An AI system that works in a loop: it decides what to do next, uses a tool to do it, observes the result, and repeats until the task is done or a person steps in.

A regular chatbot generates one response and waits for your next message. An AI agent keeps going on its own. It breaks a goal into steps, picks a tool for each step (searching files, querying a database, sending a message), reads what comes back, adjusts its plan, and moves to the next step. The loop continues until the task is complete, a limit is reached, or a human intervenes. That ongoing loop of reasoning and action is what separates an agent from a one-shot assistant.

Builder example

If you are building an agent, you are building a system that takes real actions in the world without waiting for permission at every step. That changes what you need to test: a chatbot can give a bad answer, while an agent can send a bad email, delete a file, or spend money. Design your loop with clear stop conditions, permission boundaries, and logging from the start.

A customer asks for a refund. The agent can read the order, check policy, and issue the refund.

Give it read tools first, require approval for payment actions, and log every decision input.

Common confusion: Many products labeled "AI agents" are single model calls wrapped in a UI. A true agent has a decision loop, tool access, and the ability to act on what it observes across multiple steps.