Municipality of Wageningen
Within the Municipality of Wageningen, the council clerk’s office plays an important role in preparing council meetings. Before proposals appear on the municipal council’s agenda, clerk staff check the documents for structure, language use, and compliance with guidelines.
“A lot of time went into that,” Kools now knows. He visited the municipality two times to talk with employees and gain insight into their work. He also had frequent contact with Joost Gerrits, who works as a council advisor for the municipality. The project ran from August to December 2025 and originated from an earlier study on the possibilities of AI within the municipality.
Kools explains: “Many of the changes in those documents are relatively simple: missing headings, incorrect numbering, or spelling mistakes. A template that is automatically followed would be ideal. The clerk’s office indicated that they would rather spend more time on substantive support for the municipal council’s documents.”
So the question to Kools was clear: Can artificial intelligence help automatically generate part of that feedback?
System prompt
Kools advised his clients to create an agent within Microsoft Copilot, integrated into the municipality’s Microsoft environment. He wrote the system prompt: fixed instructions that the language model receives at the beginning of every conversation.
“I guided this language model very specifically with context and instructions,” he explains. For example, Kools specified which guidelines the municipality follows and what the agent should and should not provide feedback on. He based these instructions on conversations with employees from the council clerk’s office, municipal guidelines, and examples of documents along with the feedback that accompanies them.
The result is a practical tool: an employee can upload a document in Copilot and immediately receive feedback that aligns with municipal templates and writing rules. “Technically it wasn’t extremely complicated for me,” he says. “But I really had to understand how the municipality works and what the clerk’s office needs. By talking with employees, I realized how much knowledge from my studies I could apply.”
A tool
Although the technology can do a lot, Kools emphasizes that AI remains a tool. “I also pointed out the risks to the employees of the Municipality of Wageningen. Such a model can sometimes make things up. The text may look very convincing, but that doesn’t automatically mean everything is correct,” he says.
“The system works based on statistical patterns in large amounts of data. That’s why it remains important for people to stay critical and remain responsible for the final result.”
A valuable collaboration
The practical experience Kools gained through MKB Datalab-Oost during his studies proved valuable. “It gave me insight into a world I didn’t know very well yet. At the same time, I saw that my knowledge can really mean something for an organization.”
Within the municipality, the proposed agent has been positively received. Council advisor Joost Gerrits says: “The added value of this collaboration is that we were able to develop an AI agent at low cost, while learning from the process ourselves and being able to adjust along the way based on feasibility. With a commercial procurement process, this would not have been possible.”
The use of AI remains a topic of discussion within the municipality. Kools thinks that’s a good thing. “Employees are open to it, but there is also healthy skepticism. The question is always: how much control do you want to hand over to AI? It’s important to make clear that it is a tool. Responsibility always remains with people.”
For Kools, the collaboration left him wanting more. “I enjoyed combining technical knowledge with conversations with people in an organization. Explaining how complex technology works and exploring together what is possible. That’s something I’d like to do more often in the future.”
What is MKB Datalab-Oost?
MKB Datalab-Oost introduces entrepreneurs to AI and data-driven business in an accessible way. It does this through inspiration sessions and masterclasses, as well as by engaging in small-scale AI projects that are tailor-made in close collaboration with regional entrepreneurs at an attractive rate. MKB Datalab-Oost also offers Data Buddies: students who assess the opportunities within a company to start working with AI projects. The initiative is a collaboration between Radboud University and the University of Twente. It invites entrepreneurs in the eastern part of the Netherlands to learn more about AI and the possibilities it can offer them. The projects are carried out by AI and Data Science students from Radboud University, who are supported by their lecturers.
