Home / AI glossary / What is an AI agent?
AI glossary

What is an AI agent?

An AI agent pursues a goal on its own, planning steps, using tools and acting in your systems.

In short

An AI agent is a software system that uses an AI model to plan and carry out multi-step tasks on its own — choosing actions, using tools and calling other systems to reach a goal. Unlike a chatbot, which answers questions, an agent can take action: book the appointment, update the record, run the workflow. See AI agent vs chatbot →

An AI agent is software that pursues a goal on its own. Instead of answering a single question, it plans the steps, calls the tools it needs, checks the result and keeps going until the job is done or it gets stuck.

The distinction that matters commercially: a chatbot informs and a human acts; an agent acts. That makes agents more useful and more risky, which is why serious deployments put approval steps around anything irreversible.

What does an AI agent actually do?

A working agent has four parts: a goal you set, a model that reasons about how to reach it, tools it can call (your CRM, an order system, a search index), and a loop that lets it check its own work and retry. Remove the tools and you have a chatbot. Remove the loop and you have a single prompt.

Where AI agents actually fail

Agents fail quietly rather than loudly. The common modes are: looping on a step they cannot complete, calling a tool with plausible but wrong arguments, and confidently reporting success on a task they only partly finished. None of these throw an error, which is why logging and a human approval step matter more than model choice. Treat an unattended agent with write access to a production system as an outage waiting to happen.

Agent, chatbot or automation script?

Use a script when the steps are fixed and known, it is less expensive and it never improvises. Use a chatbot when people need answers from your content. Use an agent when the steps vary per case and a human would otherwise decide them each time. Most SME processes that look like agent work are actually script work with a language model at one step.

What does building one cost?

At Crux Digits an agent is scoped like any other build: a €2,500 audit to confirm the process is worth automating, a €20,000 Production-ready MVP that runs the agent on your real data for 4–6 weeks, and production from €50,000 once it has proven itself. The audit exists partly to tell you when a €4,000 script would do the same job.

Frequently asked questions

What is the difference between an AI agent and a chatbot?

A chatbot answers; an agent acts. A chatbot retrieves information from your content and hands it to a person, who then does something with it. An agent uses tools to complete the task itself, looking up an order, updating a record, sending a reply.

  • Chatbots connect to knowledge bases and documents; agents connect to CRM, ticketing, order systems and internal APIs.
  • Agents need guardrails and approval steps because they can act, not just inform.
  • Most teams get more value from a good chatbot first, then add agent capabilities.

Are AI agents reliable enough for production?

For small jobs with someone watching, yes. For running end to end on its own, usually not yet. The more steps you chain together, the more small mistakes pile up. So keep the job narrow, and put a person in front of anything you cannot undo.

  • Rule of thumb: the more steps an agent chains, the more its per-step accuracy compounds against you.
  • Read-only agents are far safer to deploy than agents with write access.
  • Log every tool call, agents fail silently, so you cannot debug what you did not record.

What does an AI agent cost to build in the Netherlands?

€2,500 for the audit, €20,000 for a Production-ready MVP, production from €50,000. Those are Crux Digits fixed prices, excluding VAT. Across the Dutch market, AI consultant rates generally run €125–€250 per hour, and most agencies publish no price at all.

  • Typical timelines: audit 1–2 weeks, Production-ready MVP 4–6 weeks, production from 8–12 weeks.
  • A simple workflow script is often less expensive and more reliable, the audit will say so if that is the case.
  • Dutch companies can often offset part of the R&D wage cost through WBSO.
Go deeper with Crux Digits

Want this applied in your business? See how we take it to production:

← All AI terms

From concept to working tool?

We build this AI in production, at fixed prices, with one named expert. Start with a free consultation.

Book a free consultation →