When Haaga-Helia University of Applied Sciences introduced AI into IT service management, the approach was straightforward: start small, stay practical, and build something useful from day one. Rather than chasing a perfect solution, they focused on fitting AI into their existing environment, keeping it simple to use, easy to manage, and built to grow.
"It’s better to start with a limited scope and see how it works in practice. You learn much more from real usage than from planning everything in advance." "
Sampo Tyllilä
Matrix42 Platform Administrator, Haaga-Helia University of Applied Sciences
Haaga-Helia supports around 11,000 students and staff across multiple services, with Matrix42 at the core of service delivery. As AI became more relevant in IT support, the team wanted to explore how it could improve their service experience.
One decision guided their approach from the beginning. AI should not be a separate tool. It needed to be part of the existing system, embedded directly into the service portal. Reaching that point required preparation. The knowledge base was not ready to support AI. Instructions existed in different formats, many were image-based, and much of the content needed updating. The team also had to ensure consistency across Finnish and English materials.
Security and data protection added further complexity. Contracts, access rights, and data visibility had to be clearly defined before anything could go live. This took time but was essential to build trust in the solution. The team also expected adoption to be gradual. The goal was not immediate efficiency gains, but to create a safe and controlled way to learn.
When the opportunity to join a Matrix42 AI pilot came up, Haaga-Helia moved forward. The preparation phase quickly became the most important part of the project. AI was introduced directly into the service portal so users could access it in a familiar environment. To keep things manageable, the first phase was limited to IT-related questions for staff only.
Most of the effort went into improving the knowledge base. Image-heavy guides were rewritten as text, and all content was made available in both Finnish and English to support better AI responses. At the same time, strict controls were put in place. The AI only had access to selected content based on user roles. This ensured that sensitive or irrelevant information was not exposed. Another priority was reliability. Instead of focusing on perfect answers, the team tested how AI behaves in difficult situations. Could it be pushed into giving unsafe or misleading responses?
"When the AI gives a different answer each time you probe a boundary, that shows a weakness. Ours stayed consistent. That mattered more than technical accuracy."
Joonas Kröger
Head of End-User Services, Haaga-Helia University of Applied Sciences
This approach helped build trust in the system early on.
At the same time, the support team began to adjust their way of working. Instead of only solving tickets, they started focusing more on improving the knowledge that AI uses. This shift is still ongoing, but it is a key part of making AI useful in the long run.
The Results
The pilot confirmed that a careful and structured approach works. Even with a limited scope, the AI proved to be stable and consistent. It handled edge cases well and did not produce unexpected or risky responses. There have been no major issues or negative reactions from users.
Usage is still growing. Around 700 staff members have access, but only a smaller group actively uses the service. This is expected at this stage. The team sees this phase as building a foundation rather than measuring success through volume.
The preparation work also delivered clear benefits. The knowledge base is now more structured, easier to maintain, and more useful across all support channels. Most importantly, the team has confidence in the solution. AI is no longer an experiment.
Haaga-Helia is now preparing to expand its use of AI. The next step is to open the service to a larger audience, especially students. This will increase usage and create more opportunities to automate common requests. The long-term goal is clear. AI should become the first point of contact for users who do not need immediate human support.
"We want the AI assistant to be so good that it becomes the preferred option over calling the helpdesk. "
Joonas Kröger
Head of End-User Services, Haaga-Helia University of Applied Sciences
The team is also exploring how AI can support service agents. This includes suggesting solutions based on previous cases, summarizing tickets, and helping with routine tasks. The aim is not to replace people, but to remove repetitive work and free up time for more complex issues.
Haaga-Helia’s experience highlights a few practical lessons:
Ready to start your own AI journey?
This case shows that getting started with AI does not require a large transformation. It starts with preparation, a clear focus, and a willingness to learn.
The technology is already there. The next step is to make it work in your environment.
Talk to your Matrix42 contact to explore how you can begin your own AI pilot of contact us here.