AI Automation in Dutch Businesses: Where Do You Start?

AI automation for Dutch businesses: choose one process, involve your team, and account for privacy, oversight and the EU AI Act.

Written by Christian and Michel.

AI automation only works when it fits a process your business already understands. For Dutch organisations, that does not start with a tool. It starts with a clear choice: which recurring work needs more structure, and where should human review remain in place?

That is a practical starting point, but it also matters for how you handle data, staff and accountability. Dutch government research describes how SMEs encounter different motivations and barriers when applying AI to improve business processes. A useful first step should therefore fit your organisation’s daily way of working.

1. Choose one process your team recognises

Start with a process that repeats regularly and that the people in your team can explain well. It could be preparing enquiries, gathering information for a case file, checking data between systems or producing a first summary.

A sound first process has three characteristics:

• It has a clear trigger and outcome.
• It uses information your team already works with.
• A team member can check whether the outcome is useful.

This keeps AI from becoming a separate experiment. It becomes a focused improvement to work that already exists. For a broader first inventory, you can also use our AI roadmap for your business.

2. Map people, information and systems together

Next, look at the entire workflow rather than one action in isolation. Who starts the process? What information arrives? Which systems are involved? Where does someone need to choose, check or make contact?

A short conversation with the people doing the work often offers more insight than a list of disconnected AI ideas. They know where duplicate work occurs, which exceptions recur and where information is not clear enough.

Also define what an application should not do. AI may organise information, prepare a first proposal or signal a next step. A team member remains responsible for exceptions and decisions that affect customers, colleagues or the organisation.

3. Make oversight part of the design

Oversight is not a final check after implementation. Agree in advance when someone reviews an outcome, when a process stops and who can investigate an exception.

That helps your team work with the application and makes it easier to learn from it. A practical starting question is: which outcome may be prepared automatically, and for which outcome must a team member always decide whether a next step follows?

This approach fits the risk-based design of the European AI Act. The European Commission makes clear that rules and obligations depend on the use and its risk, not simply on the fact that AI is involved.

4. Address privacy from the start

Where an AI application processes personal data, the GDPR remains relevant. The Dutch Data Protection Authority advises organisations to first consider whether generative AI can be used without processing personal data. Where processing is necessary, its tool helps organisations consider GDPR duties and technical and organisational measures from the beginning.

Make these points clear during the first exploration:

• Which data the application needs.
• Whether that includes personal data.
• Who can access the input and outcomes.
• How your team can report an error, unwanted outcome or question.

For more on these principles, read our guide to GDPR and AI automation.

5. Include the EU AI Act in your way of working

The AI Act has applied in its main part since 2 August 2026. Some elements applied earlier, including prohibited AI practices and AI literacy obligations from 2 February 2025.

For many organisations, this does not mean every AI project follows the same route. It does mean you should be able to explain what the application does, who works with it, which control points exist and which arrangements are relevant to your situation.

Take these questions into account:

1. Does the application support a team member, or does it make a decision that directly affects someone?

2. Can a team member understand the outcome and intervene where needed?

3. Do the people working with it know what the application can and cannot do?

4. Is it clear who is responsible for its use in daily practice?

For the role of your team, see our page on AI literacy and Article 4.

6. Build further only once the first step is sound

A first application answers questions you cannot fully predict in advance. Does the input fit the real work? Does the team understand when an outcome needs adjustment? Are the agreements on data and oversight clear enough?

Only when that foundation is sound is it useful to choose a next step. You may extend the same process. Another workflow may be more suitable. Or you may find that the data or way of working needs more structure first.

At NordX, this starts with understanding how a business works. We then decide together where AI automation or custom software can fit its daily practice. Explore AI automation for businesses or contact NordX for an initial conversation.

Sources

• Dutch Government, Research on AI use in SMEs: Ambition or hesitation?
• European Commission, AI Act application dates and risk-based approach
• Dutch Data Protection Authority, guidance for generative AI and the GDPR

Meer weten? Bekijk onze andere artikelen op het NordX blog of neem contact op via nordx.ai.