How Titeca trained its entire organization on Microsoft Copilot

Project Overview

Who is Titeca?

Titeca is an accountancy group headquartered in Roeselare, with offices in Kortrijk, Ghent, Waregem and Oudenaarde, among others. The group has around 220 employees.

Titeca was already well ahead when we started. There was an internal A.I. working group, close to a hundred Copilot Pro licenses already in use, and the organization was in the middle of a shift from SME to bigger company.

The challenge

That's exactly where the problem sat. At most organizations, the question is how to get people using A.I. at all. At Titeca it was different: the licenses were being used, just without direction.

Three things came out of the discussions in July 2025:

• Employees were experimenting a lot, but lacked a clear framework and solid training.
• There was confusion about what Copilot agents can and can't actually do. Microsoft markets them as powerful tools, while in practice they have real limits.
• There was no governance: who's allowed to build something themselves, when do you bring in a partner, and who's responsible when someone leaves.

On top of that came an organizational challenge. Titeca works with different pillars and support functions, each doing a different kind of work. One training for everyone would miss the mark for half the room.

Project Delivery

The approach

We didn't start with a big rollout. We started with a test.

1 · A try-out with eighteen ambassadors

In August 2025, we ran a try-out session for eighteen employees across the different pillars, all people who already had a Copilot Pro license and were already doing something with it. They got the session first and gave live feedback through Mentimeter.

That's a small investment with a big payoff. The official trainings could then be adjusted based on what turned out to be too long, too short or too vague in practice.

2 · Inspiration sessions for the wider organization

Short online sessions of about an hour, organized per pillar. Goal: understand what A.I. is, what it can't do, what the risks are, and how to use it responsibly. No skills yet, just a framework.

3 · Hands-on workshops per department

Three-hour half-days, with a fixed structure: introduction, learning to prompt, using Copilot inside the Microsoft tools, exercises, and concrete next steps to keep going independently.

These workshops weren't split by office, but by job group. There were separate sessions for the different pillars, and within the support functions, even separate ones for IT, for People & Culture, and for office and management assistants. Every group got examples from their own work.

4 · Advanced users and governance

For people who wanted to go further, we added sessions on Copilot Studio and agentic A.I.: how to build an agent that supports a process, and where the limits are. Alongside that, we worked on the framework: who's allowed to build, who's allowed to publish, and when to bring in a partner.

How the program ran

Sessions took place at the offices themselves: Roeselare, Kortrijk, Ghent, Waregem and Oudenaarde. Groups of about ten to fifteen people, because a hands-on workshop doesn't work at a bigger size.

By February 2026, Titeca had run fifteen to twenty Copilot trainings. Sign-ups came in steadily, and a few sessions filled up completely.

For fall 2026, a follow-up series of nine sessions is planned: inspiration sessions for new colleagues, basic workshops for the new regional offices, and advanced sessions on Copilot Studio.

Results

What participants say

Titeca's L&D lead sent us the measured feedback from participants after the fall program:

• Satisfaction with the content: 81 percent on average.
• 84 percent of participants would recommend the training to a colleague.

The comments that came with it were just as useful. The training was seen as accessible and well delivered, and the interactive, hands-on approach was strongly appreciated. The main point to watch: there's a real gap in Copilot knowledge within a group like this. For some, it can go deeper; for others, the level is just right. The client's own advice was to keep explicitly inviting participants to say whether the pace should go faster or slower.

That's exactly the kind of feedback you improve a program with, and it's why the sessions are split by job group instead of by office.

What it delivers in day-to-day work

Trainings are rarely measured in time saved. At Titeca, that saving did surface, in a conversation with the team that works on credit files and valuations.

Task	Before	After
Building a valuation model	about 3 hours	about 10 minutes
Building a credit file	2 to 3 days	about 2 hours, once all documents are ready

Valuations happen several times a week. Credit files about once or twice a month.

Worth being upfront about: these are figures straight from employees, not from a formal baseline measurement. And the credit file runs on an agent the team built itself, with our guidance. That's not a lesser result, though — it's actually the point: after the training, people build things themselves.

Conclusion

Titeca went from close to a hundred licenses without direction to an organization where A.I. training happens by job group, where people build their own agents, and where there are clear agreements on who's allowed to build and publish what.

The program isn't finished, and that's exactly how it should be. Nine new sessions are planned for fall 2026, including a series for new colleagues and the new regional offices. There's also a proposal on the table to train internal A.I. ambassadors, so knowledge keeps growing between our sessions.

Do you also have licenses being used without direction? Then your next step isn't another tool — it's a program. Get in touch with Flowkify for a conversation about A.I. training and governance in your organization.