AI agent

What is an AI agent?

An AI agent is AI that carries out a task in several steps. It is given a goal, chooses the steps itself and uses other software to do so, such as your email, calendar or CRM.

A chatbot or AI assistant mainly gives answers. An agent also does something with that answer.

How does an AI agent work?

At its core is a language model. That model is given tools, for example a search function or an integration with your accounting software. The model thinks of a step, carries it out, looks at the result and chooses the next step. This repeats until the task is done, or until the agent gets stuck.

That is how an agent differs from ordinary workflow automation. There, the steps are fixed in advance: if this happens, do that. With an agent, the model chooses the route. That makes it flexible, but also less predictable.

Example in an SME

Where are the limits?

• A person checks. An agent can make mistakes, and each step builds on the previous one. So a mistake at the start carries through. Have a person approve steps with consequences, such as sending, paying or deleting.
• Limit the permissions. Only give the agent access to what it needs for the task.
• Start simple. Often a fixed automation or an assistant is enough. An agent suits work where the steps differ each time.
• Rules. If an agent works with personal data, the GDPR applies. If it makes decisions about people, human intervention is important. In recruitment and selection, for example, the EU AI Act sets stricter requirements.

When an agent makes sense in your business, you can read in An AI agent for your business. In a business analysis I look with you at which work is suitable, and which is not. Read more about AI in your business.

Frequently asked questions

Sources: Anthropic: Building effective agents and the Dutch Data Protection Authority (Autoriteit Persoonsgegevens): Guidance on generative AI and the GDPR (in Dutch). Consulted on 7 October 2026.