Mattijs, you have been working with insurance data for more than ten years. What keeps it interesting?

“I actually started out as a test engineer, without a traditional IT background. That turned out to be a great way to learn. As a tester, you first have to understand how a system is supposed to work and then figure out whether it actually does. I was a bit of a sponge, so from there I moved into data analysis, ETL development, data warehousing and eventually migration delivery. What keeps it interesting is that no two migrations are the same. The systems change, the products change and the problems change. And now AI is changing the way we work again. It is probably the biggest shift I have experienced in my career.”

People hear ‘data migration’ and think: move data from system A to system B. Is that basically it?

“If only. (laughs) Moving the data is the easy part. Moving a million rows from one database to another is not the real challenge. Understanding what those rows mean is. An insurance system can contain decades of history. Products have changed, rules have been added and exceptions have accumulated. Some of that is documented. A lot of it isn’t. Crucial knowledge can still sit in the heads of a handful of people who have worked with the system for years – and are quite often enjoying their retirement. That makes migration sometimes a bit like archaeology. You open the box and try to understand what is really happening underneath. Then you have to translate that old world into a new system without losing its meaning. Migration is mainly a knowledge problem, not a plumbing problem"

"AI defines the new speed of migration" - Mattijs Augustijnen

"Our experts can spend much more of their time understanding the data, challenging assumptions and validatig the result" - Mattijs Augustijnen

And ‘close enough’ isn’t really an option?

“Definitely not. On one migration I worked on, we moved around six billion euros in savings products. At that point, you don’t just say: ‘Looks good to me.’ You need to prove that the result is correct. If there are 200 employees in the old system, I want 200 in the new one. If financial values are transferred, I want them to reconcile. We can even verify whether premiums calculated after migration remain consistent with the old situation. That is why reconciliation is so important. It gives Finance, Actuarial, auditors and ultimately the client confidence that nothing disappeared or changed unnoticed along the way.”

Where does AI change that process?

“Traditionally, you had an analyst investigating the data and writing specifications and an engineer translating those specifications into code. That created handovers, waiting time and opportunities to lose context. AI removes a lot of that friction. It can help us analyse databases, build transformation logic, generate tests and investigate issues. That means our experts can spend much more of their time understanding the data, challenging assumptions and validating the result. You notice the difference even in meetings. Someone asks a detailed question about the portfolio. In the past, I might have written it down, investigated it afterwards and returned with an answer later. Today, I can often explore the question while the meeting continues. Five minutes later, we can look at an actual case together. That changes the rhythm of a project completely.”

"The project starts moving at the speed of decision-making, not the speed of coding" - Mattijs Augustijnen

So AI mainly makes migration faster?

“It makes us faster for sure, but that is not even the most interesting part. It changes where we spend our time. If AI can take over some of the engineering work, our experts can focus more on analysis. We can investigate anomalies, test assumptions and identify problems much earlier. Instead of discovering something difficult close to go-live, you can find it during an early rehearsal when there is still time to do something about it. The budget and people’s effort move towards the part of the project where the real risk sits: understanding and proving.”

Does working faster have its downsides?

“It creates a new bottleneck. And strangely enough, that is good news. In the past, development took time. A difficult business decision could wait because the developers were still building other things. With AI, that buffer becomes much smaller. Once something has been specified, it can be built quicker. Suddenly, developments are waiting for the specifications from the business side. The project starts moving at the speed of decision-making, not the speed of coding. That exposes knowledge gaps much earlier, which is exactly when you want to find them.”

With AI doing more, what happens to the consultant?

“The consultant becomes more important where it matters. AI can write code faster than I can, no doubt about it. Sometimes it also spots something I have missed. I will give it an instruction and it comes back with a question I had not considered. And occasionally my reaction really is: ‘Damn. Good point.’ But AI does not own the migration. We do. You still need people who understand the business, the products, the source system, the target system and the data. You need someone who can sit with Finance or an actuary and understand why an exception matters. Somebody always has to take responsibility. Everything AI produces has to be directed, reviewed and validated by people. AI can do the heavy lifting, yet the expert still has to know whether what comes out makes sense.”

"Ai forces us to think carefully about accountability" - Mattijs Augustijnen

You’ve built your own data platform at thewave. What does that add?

“AI becomes much more useful when it understands the context it is working in. In a traditional project, knowledge is often scattered across specifications, meeting notes, emails and people’s heads. We wanted to bring that together in a structured and safe way. The platform connects to the data and captures what that data means: definitions, mappings, rules and decisions. That gives both our experts and the AI, which we use in a confined space, a much stronger foundation to work from. Migration is the first use case because it is such a demanding one. But the same principle can be used for reconciliation, reporting or data-quality work. Once you have properly understood and structured the data, you can do much more with it.”

Insurers are cautious about AI, particularly when sensitive data is involved. Do you understand that hesitation?

“Utterly and completely. Insurance runs on trust. You cannot experiment carelessly with sensitive data. That is also why I think the human-in-control principle is so important. AI works within a controlled environment and our experts remain responsible for what it produces. We review it, test it and reconcile the result. I don’t see AI as something that replaces accountability. If anything, it forces us to think more carefully about it. AI may produce the work faster, but I am still accountable for the outcome.”

"AI makes experts stronger. It doesn't make expertise unnecessary" - Mattijs Augustijnen

What ultimately makes a migration successful: better technology or better people?

“People. It’s people who do the trick, every time. You need technical expertise, business knowledge and people who genuinely understand the systems involved. You also need the right decision- makers around the table. Technology can make that team dramatically more effective, but it cannot compensate for missing knowledge or decisions nobody wants to make. The strongest combination is good people using good technology. AI makes experts stronger. It doesn’t make expertise unnecessary.”

Where is migration heading?

“I think migrations will become significantly faster, but also more transparent. Reconciliation and traceability will increasingly be built into the process rather than treated as something you do towards the end. There is another interesting consequence: at the end of a good migration, you don’t just have a portfolio in a new system. You also have a much better understanding of your data. That clean, structured and understood dataset can then support reporting, automation, data quality and new AI applications. So migration becomes less of a painful one-off project and more of a capability you can build on.”

Do you find yourself surprised by what AI can do?

“All the time. There are moments when it produces a specification, spots an issue or asks a question that is simply better than what I had in mind. We are occasionally getting beaten at our own game. And I love it.”