What is an AI agent, and how is it different from a chatbot?
The short answer
A chatbot answers your question and stops. An agent is given a goal and works towards it: it decides its own next step, uses tools such as your email or your business system, looks at the result, and carries on until the job is done or it needs you. The real difference is who decides the next step.
A chatbot answers; an agent acts
You type a question into a chatbot. It writes an answer, and it stops. Whatever happens next, you do it yourself: you copy the answer, open the system, make the change.
An agent is given a goal instead of a question. For example: “Prepare a reply to every client who wrote this week and is still waiting.” The agent decides what to do first. It reads the mailbox, opens the client’s record, drafts the reply, checks what is left, and moves to the next client.
Anthropic, one of the companies that build these models, describes agents in simple terms: a language model that uses tools, based on what it finds, in a loop.
The difference that matters: who decides the next step
Between a chatbot and an agent there is a middle ground, and many companies need exactly that. The steps are written in advance, in a fixed order, and the model does one or more of those steps. Anthropic calls this a workflow: the path is set in code, and the model does not choose it. In an agent, the model chooses the path.
This is a management choice before it is a technical one. A workflow is predictable: the same request takes the same route every time. An agent is flexible: it can handle a case nobody planned for. For the same reason, it can also take a route nobody planned for.
What an agent needs before it can do useful work
- Access to your information. An agent that cannot see your contracts, prices and client history will guess about them.
- Tools. Each tool is one thing the agent is allowed to do, such as “search the mailbox” or “read an invoice”. The agent can do nothing beyond the tools it is given.
- Limits and a stopping point. What it must never do on its own, and when it must stop and ask a person.
Where to start
Anthropic’s advice is to use the simplest solution that works, and to add complexity only when it is needed. If the steps are always the same, you need a workflow, not an agent. Use an agent where the cases vary too much to write the steps in advance.