From 8 to 81: the government takes a giant leap in AI

What the government is doing with AI

Municipalities are leading the way with 34 applications, accounting for more than 40% of the total. That is not surprising. There are 342 municipalities in the Netherlands, all providing comparable services. A chatbot that works in Utrecht can, in principle, also work in Eindhoven.

But a striking pattern emerges here. Of all applications, 78% are aimed at employees, chatbots that help summarise documents, answer internal questions, and process knowledge. Only 17% are directed at citizens directly [1].

That choice is deliberate. By deploying AI internally first, the government builds experience without immediate risk to citizens. Organisations are increasingly offering staff their own chatbot to curb uncontrolled use of ChatGPT. A sensible approach, but it also means that citizens will notice little of the AI investments for the time being.

Where things are working well is collaboration. GEM, a partnership of 25 municipalities, is developing a shared chatbot. Open source, so public money means publicly available. And a growing number of organisations are placing the social challenge at the centre rather than the technology.

Where the tension lies is that the risk classification of more than half of all applications is still unknown [1]. And data management is often not in order. "If we put the chatbot on our website, all the web pages need to be up to date, otherwise the bot will say strange things," noted one respondent [1].

But who is actually supplying this?

Microsoft, Google and OpenAI deliver solid technology, proven, documented, and ready to deploy. For a municipality that wants to get started tomorrow, the choice is understandable. The models work, the contracts are often already in place, and the IT department knows the environment.

But one respondent from the TNO research put it bluntly: "In my experience at the Dutch government, we walked straight into the Microsoft trap. Now you notice we are at a kind of crossroads again." [1]

The figures confirm that picture. For 54% of applications, we know for certain they run on American technology. For 42%, the origin is unknown. Only 4% is demonstrably European [1]. OpenAI, Meta, Microsoft and Google dominate. European alternatives can be counted on one hand: DeepL and Mistral. And respondents indicate that those European options are not yet up to standard in terms of quality [1]. That is partly true, but quality grows with use. Without customers, alternatives will continue to lag behind.

Why this matters

The question is not whether Big Tech delivers good tools. It does. The question is whether we want to place everything there. In the current geopolitical climate, that is more than a technical choice.

Every euro that flows to Silicon Valley is a euro that does not reach Dutch AI companies, no revenue, no growth, no track record to win the next tender.

The Netherlands has the ambition to position AI and data as a third mainport, alongside Schiphol and the Port of Rotterdam [2]. That ambition requires launching customers: major clients willing to choose Dutch and European solutions, even when they are not yet perfect. The government can take on that role.

But procurement needs to play its part too. In regular ICT contracts, there is little room for innovative SMEs [3]. Requirements around revenue, track records and certifications create barriers that young companies can rarely clear, a chicken-and-egg problem the government itself can break.

Where things can be done differently

There are examples of organisations that consciously make different choices. The Nationaal Archief is developing an Archive Assistant built on Mistral, a French open-source model [1], a principled decision to retain control over the technology.

Tolkie, a social enterprise from Eindhoven, helps municipalities make their texts accessible to the 3.3 million people in the Netherlands with low literacy [4]. The tool runs on European servers. Multiple municipalities are already using it. It demonstrates that Dutch startups can deliver what the government needs.

One respondent summed it up: "Civil servants alone won't make it work. Entrepreneurs alone won't make it work either. You really need to move forward together." [1]

The crossroads

From 8 to 81 applications in a year. The government is no longer experimenting, it is building. Municipalities are collaborating, organisations are sharing knowledge, and increasingly the social challenge is taking centre stage rather than the technology.

The examples in this article show that Dutch and European alternatives exist. The Nationaal Archief is making a deliberate choice for Mistral; Tolkie is helping municipalities with accessible texts. The growth from 8 to 81 is a promising first step. The next step is for the Dutch startup industry to offer more home-grown services, with the government acting as a launching customer. That is how our ecosystem can take off and accelerate the ambition to become Europe's leading AI and data mainport.