UTF.ai: A Sovereign AI Writing Assistant for the Government

With "Breaking Barriers," the Tilburg-based startup aims to establish an AI Trust Center and strengthen its sovereign AI stack.

When you write policy papers or documents in the public sector, there are countless requirements you must meet. These include corporate style guidelines, legislation, as well as language standards and internal protocols within your own department. Documents are often shuffled back and forth between departments to ensure the texts meet the required standards. To simplify this process, the Tilburg-based startup UTF.ai developed the AI solution Schrijfkader.

We spoke with Juul van Kessel (CEO and co-founder) about his platform and his goals within the program.

“Call us when you figure this out”

Van Kessel studied Cognitive Science and Artificial Intelligence. At the end of his studies, he had the opportunity to work with co-founder Giulio Tosato on a co-creation project with the City of Tilburg to develop a solution that eventually became Schrijfkader. The AI writing assistant evaluates a document as it is being written—assessing its structure, writing style, and language level. The tool also checks whether the document complies with the Open Government Act (Woo). The AI makes suggestions and provides supporting rationale. The writer then decides for themselves whether to follow the advice.

Before they started building, they called Dutch municipalities to find out if such a tool could be scaled. “They said, ‘Please, figure this out and call us when you’ve done so.’ This encouragement led to the creation of their own company: UTF.ai. “I wanted to start a business in the most opportunistic, competitive, and impactful sector out there.” That turned out to be AI. UTF.ai now develops multiple AI solutions for governments, including the transcription tool SlimGesprek, which automatically converts conversations into reports, and the AI Trust Center.

We want every government agency to be able to work with us with complete confidence.

Juul van Kessel, CEO and co-founder of UTF.ai

But for a startup, working in the public sector isn’t easy. You have to prove that you’re trustworthy. “It’s difficult to launch a product because you have to meet many requirements related to security, privacy, and compliance,” says Van Kessel. “That involves multiple certifications, more than fifty documents, and everything has to be perfect. The government can’t afford for anything to go wrong.”

UTF.ai already counts several national and regional governments among its clients. To further build trust, UTF.ai will use the expertise gained through the Breaking Barriers program to establish what is known as an AI Trust Center. This will be a single location where it is clearly visible which AI standards the startup adheres to and how compliance is verified. For Van Kessel, the Trust Center is crucial during his company’s current growth phase. After all, the stronger the evidence of compliance, the greater the trust—and that means more clients. “We want every government organization to be able to work with us with complete confidence.”

Sovereign AI

For a government organization, it’s important to know where data ends up and what regulations govern it. We’re talking about citizens’ data, but often also about policies that aren’t yet in effect. That’s sensitive information. “In practice, anyone in the government who purchases software almost always ends up running it on the servers of large American tech companies. That has become the standard, even if a solution is hosted within the organization itself. We leave that choice up to the customer: we’re transparent about where and how data is processed, and we offer a fully European, sovereign environment as an alternative,” says Van Kessel.

To make this happen, UTF.ai built a sovereign AI environment at the European STACKIT. “I think we’re one of the first to be able to offer this to the government in this way, even when it comes to scale.” The Breaking Barriers program is being used to validate this European AI stack and strengthen it where necessary. This includes support for sovereign hosting, logging, and continuous monitoring, as well as integrating these features into the Trust Center.

To be an AI alternative

So far, the options for AI solutions within the government have been limited. “You often hear that Copilot is the only option that meets the requirements. But many processes aren’t possible with it. And on top of that, it’s an American provider, simply because there are no alternatives.”

UTF.ai aims to be just that: an alternative. The company is already active at all levels of government and aims to accelerate its growth by leveraging the Trust Center as a burden of proof and the European AI stack as a solid foundation for AI. “We’re leading the way in AI for the government. Now we need to ensure that our trust and our foundation grow stronger,” Van Kessel concludes.