In this series, master's degree AI students talk about their experience at companies in the Lifeport region, gained through the SME Datalab-East. From quick wins to strategic insights, their stories show that artificial intelligence is not future music, but can already make a difference for regional SMEs today.
MKB Datalab-Oost is an initiative of Radboud University and the University of Twente that helps SMEs in the eastern Netherlands exploit the potential of artificial intelligence and data. In doing so, AI and Data Science students carry out AI projects (6-10 weeks) or work as a data buddy (40 hours), on site.
More and more SMEs in the Lifeport region are discovering how students from MKB Datalab Oost can help them with smart, practical data applications.
From Excel sheet to intelligent decision tree
With her background in Artificial Intelligence and Information Sciences, Rivka van der Horst wanted to experience what she could do for an entrepreneur in practice. Through the Datalab, she got the chance to explore that at DBL Finance in Duiven.
Rivka van der Horst, alumna AI at Radboud University, became hooked on technology at an early age. As an eight-year-old, she put together her first websites by copying HTML code from game pages and adapting it to her own taste. That curiosity grew into a bachelor's degree in Artificial Intelligence and a master's degree in Information Sciences in Nijmegen.
During her master's, she applied to SME Datalab Oost. She considered a role as a data buddy, but the team saw more in her and paired her with a full AI project at DBL Finance in Duiven.
Matching lease cars
DBL Finance is the sole trader business run by entrepreneur Dick Poot. He helps private individuals and small business owners find the most suitable car lease deal.
Rivka explains: “Until recently, Dick did everything manually, using just an Excel file, a set of fixed criteria and a great deal of knowledge stored in his head. During an inspiration session organised by MKB Datalab Oost –made possible by Rabobank in collaboration with MKB Duiven – Dick won a grant for an AI project. The original idea was to build a machine-learning model that automatically predicts which leasing company is the best fit for his clients.”
Decision tree
Rivka soon realised that machine learning was not at all suited to what Dick needed. “He works with strict criteria, not probabilities,” she says. “His decision-making process is purely rule-based. The latest AI systems, such aslarge language models (LLMs), are brilliant with language, but far less suited to working reliably with figures and strict conditions.”
That is why she threw in the towel. In 80 hours, she mapped the problem, came up with a solution and built a locally running tool that does exactly what Dick needs. "I figured out all the rules he had in his head and recorded them in an Excel file. The code automatically reads those rules and adapts immediately if it changes anything. This way, when changes are made, the tool always stays up-to-date without having to programme a single line."
All criteria met
Because much of the knowledge was only in Dick's head, it was quite a puzzle for Rivka to get all the criteria in focus. She had extensive discussions with him, delved into previous lease applications and checked each time whether her decision tree made the same choices as Dick himself. "It was really detective work at times," she says. "But precisely that process: making everything explicit that someone normally does intuitively, I found incredibly fun."
She also wrote clear documentation so that DBL Finance's web builder can later easily link the tool to the website. In the future, this will allow an application to be processed automatically through the decision tree, without Dick having to look at it manually.
Full project responsibility
The project was exciting, Rivka admits. Studying AI consists of many separate assignments that often have limited links to practice, so you learn how to write code, but less how to use that knowledge for real applications. "As a result, it sometimes feels like you mainly solve problems, without knowing whether you can use them in practice. Only when you work for a real entrepreneur do you discover what you can really do."
Rivka also discovered that technical skills are important, but that it is at least as important to listen well and to ask thorough questions in order to get the right information. Especially with an entrepreneur who has a lot of knowledge in his head and has not recorded everything explicitly. She also noticed that she enjoys taking responsibility for an entire project.
"As soon as you work for a real entrepreneur, you really notice how much knowledge you already have in-house"
Rivka van der Horst
Technically and communicatively strong
Since graduating, Rivka has been working at KPMG as a consultant in Governance, Risk & Compliance Technology. There, she helps companies set up their systems in a safe, controllable and compliant way.
The experience at DBL Finance still helps her. She now better understands what a complete project looks like, from customer demand to solution.
She noticed that this project allowed her to look beyond her own task: "This project made me see much better how all parts of a project interlock. It also made me realise the importance of being both technically and communicatively strong."
"Thanks to Rivka, there is now a stable and reliable foundation"
Dick Poot
Stable and reliable basis
Client Dick Poot: "This project makes me realise better how you slowly grow along with the matter that influences your decisions every day. You work with it almost unconsciously. By consciously thinking about the various decision trees, I got a better grip on the entire process.
Thanks to Rivka, we now have a stable and reliable basis. It is immediately usable in daily practice and forms a solid foundation to both further develop the work and support the growth of my company."
Tip for other SMEs
Like other students, Rivka discovered that AI is not always the best solution. Sometimes a simple, transparent method is better, cheaper and more reliable.
This is immediately its most important advice to SMEs: because large language models (AI trained on huge amounts of text) seem so accessible and powerful, they are sometimes too quickly seen as the solution. In practice, they are by no means always the best, simplest or cheapest choice. That is why it is smart to have someone with knowledge look at them before you start working with them.
