After your first few months at OCTO, what have your impressions been, and which priorities do you consider strategic in supporting the company’s next phase of technological evolution?

My first impression was that I had found an asset far rarer than the way it is usually described. Twenty-three years of insurance telematics do not amount only to 610 billion kilometres of driving data: they also include more than 13 million validated crashes, meaning events that have been labelled and verified against actual claims. In the world of artificial intelligence, a model is only as good as its labels, and an asset like this cannot be bought: it is built up one year at a time.

My second impression was that there is more friction than necessary between that asset and the market. This is natural after twenty years of growth: layered architectures, parallel platforms performing similar functions, and an infrastructure that is still partly physical.

I have put four priorities on the table, not ten. Simplify the architecture. Move the infrastructure to the cloud, because resilience and release speed are no longer negotiable. Redesign technology’s operating model: areas with a written mandate, declared capacity and a single point of entry for the business.

The fourth priority, which acts as a lever for the other three, is AI augmentation for our people. We need to enable our teams to use artificial intelligence across the platform’s knowledge base, with a shared method and human quality control at every stage. The goal is not to replace expertise, but to multiply capacity: this is what allows us to achieve the other three priorities faster than we otherwise could.

I chose them according to one criterion: how much friction can we remove between an idea and its release?

Technological transformation is not only about tools and platforms; it also involves processes, organisation and ways of working. How important are governance and collaboration between technology and the business functions to the success of projects?

They are crucial, but not in the way this is usually understood. The governance I have seen work is not a committee that grants approval: it is a set of verifiable commitments between areas. This is why the working model we challenged across the entire technology function is not about the organisation chart, but about what each area promises to deliver to the others.

Three principles hold it together. Those who build, operate: the people who make decisions are also the ones who maintain the outcome, and that alone aligns most incentives. Capacity first, then the gate: if we declare how much capacity is available, prioritisation becomes an explicit choice rather than an invisible queue. Enable, do not authorise: technical standards provide constraints and guidance, not an approval gate. Anyone who departs from them explains why in writing, and the decision remains traceable.

I have taken a simple approach to the relationship with the business. For every recurring interface with sales and delivery, there is one person with a mandate, not merely a place in the diary. And there is one backlog, rather than two roadmaps that turn out to be different at the end of the quarter. Because in projects, the cost is almost never the work itself: it is the waiting.

Artificial intelligence, cloud and automation are accelerating technological transformation. What is your vision for the future of technology, and how can these innovations contribute to the evolution of OCTO and connected mobility?

I will start with a point that receives little attention. The European Data Act is turning access to vehicle data into a commodity: until yesterday, the competitive advantage lay in gaining access to the data; from tomorrow, it will lie in knowing how to interpret it. And that is exactly the field in which insurance telematics has operated for twenty years.

I apply a simple criterion to artificial intelligence: the KPI is not AI adoption; it is the cost of the claim, the time required to settle it and the quality of risk assessment. If a model is not linked to a verifiable outcome, it remains a demo. Here, that verifiable outcome exists and has a name: a validated crash, a closed claim. I also see the new European regulatory framework for artificial intelligence as an accelerator in this respect: anyone performing insurance scoring has always needed to be able to explain their models.

There is another area I have made a priority and that is worth explaining: applying artificial intelligence to our own work. We are bringing the platform’s knowledge – code and documentation accumulated over twenty years – into a knowledge base that models can query, with human quality review at every stage. The aim is not to write code faster, but to understand quickly a system that no single person knows in its entirety.

The cloud is the foundation for all of this.

A more personal question: in the way you lead innovation, what principle do you consider non-negotiable and seek to instil in your team every day?

Accountability. Not formal accountability, but real accountability: the person who makes the decision is also the person who maintains the outcome. When those two responsibilities are separated, elegant decisions emerge whose cost will be borne by someone else – and those are the most expensive decisions of all.

Two things I ask for every day follow from this. The first is timely dissent. When I challenged the technology function’s new working model, I wrote one thing clearly: read it, challenge it, then we will freeze it. A doubt left unspoken today returns as friction in six months, by which point it costs ten times as much. The second is measurability: a standard that cannot be checked automatically is not a standard; it is an aspiration.

And then there is something that concerns me personally. I continue to get hands-on: during my first few weeks, I built a prototype myself to unblock an analysis that had stalled. Not because a CTO needs to write code, but because I do not believe technology can be led from slides. The people working with you immediately understand whether you have grasped the problem – and all your credibility rests on that.