Carrier AI Roadmaps in 2026: The Sequencing Most Get Backwards

Most transformation roadmaps are built in the wrong order, and the cost of that shows up years later.

Lucy Pilko of AXA XL closed the Insurtech Insights fireside with the line that reframes everything else: “The job of commercial insurers is to underwrite the economy. The economy isn’t static. Shame on us if we’re not pivoting with it.”

Her argument runs like this. Clients do not have more budget for risk transfer. But the world keeps getting riskier. Those two facts cannot both be absorbed by pricing. So the insurer’s job becomes helping the client manage down their total cost of risk, reducing what has to be transferred at all, which frees capacity for the risks a client genuinely cannot avoid.

The part worth chewing on is the sequencing. She puts transforming the client-facing mission first, and the carrier’s own operations second.

Most roadmaps invert that. Year one, replace the policy admin system. Year two, unify the data platform. Year three, retire the mainframe. Year four, assuming the sponsor is still in the seat and the budget survived two planning cycles, the client finally feels something.

Here is what that costs. Every year spent on internal modernisation before any client-facing change is a year where nobody outside the building can tell you did anything. Renewal conversations sound the same. Brokers describe you the same way. And the modernisation itself gets built against today’s assumptions about what clients need, because nothing in the sequence forces you to go and find out. By the time year four arrives, you have a faster version of an older answer.

Run it the other way and the internal work still happens, but it inherits a specification from the field. You learn which parts of the client’s cost of risk you can actually influence, and that tells you which systems matter and which were never the constraint. The efficiency gains land in the places that turned out to be load-bearing.

Neither order is free. Client-first means working around legacy systems longer than anyone enjoys, and some things genuinely cannot be done until the platform moves. The point is not that internal work does not matter. It is which one sets the brief for the other.

So the question for your own roadmap is uncomfortable and short: is it organised around your efficiency, or your client’s cost of risk?

If nobody outside the building would notice the first two years, that is the answer.