What an AI agent is, and how it differs from a chatbot

An AI agent is a program in which a language model decides for itself which steps to take and which tools to use to complete a task.

Anthropic, the maker of Claude, distinguishes two things that are often sold under the same name ¹. In workflows, the path is written in advance in the code: the model does its part, but a person decided the order of the steps. In agents, the model directs its own work and its use of tools ¹.

Chatbot and agent: the difference in an example

A customer writes: "Where is my order?"

• The chatbot answers by explaining how to track the shipment, or reads the order status if someone has connected it to your management system.
• The agent looks up the order, sees that it is held up in the warehouse, opens a ticket with the courier, writes to the customer with a date and leaves you a note.

The first one talks. The second one acts, and for that reason it can get an action wrong, not just a sentence.

Why it matters to you

When a supplier offers you "an agent", ask whether the path is decided by the model or written in advance. For many business tasks a workflow is more predictable, cheaper and easier to control. Anthropic itself says so, advising you to start with the simplest solution and add autonomy only when it is needed ¹.

If an agent is the right choice, the next question is which permissions you give it. OWASP names excessive functionality, permissions and autonomy as the main risk of agents ². This is why guardrails exist.