Inside Howden Re's partnership with Dylogy: why AI still needs a human reader

Reinsurance contracts run to thousands of pages which has historically meant hours of manual review to catch a non-compliant clause, an inconsistent exclusion, or a change buried in a new contract version has meant hours of manual review. Since 2025, Howden Re France has been using Wording Analyzer, a tool built by French AI company Dylogy, to handle more of that groundwork.

We sat down with Margaux Déchant, Reinsurance Legal Expert, Howden Re, to talk about what prompted the partnership, what it has actually changed day to day and why responsible AI use matters in an industry built on trust. 

What prompted Howden Re France to start working with Dylogy?

Reinsurance is a knowledge-intensive business built on expertise, judgement, and the ability to interpret increasingly complex information. Our aim in working with Dylogy was to give our teams tools that let them work more efficiently and analyse information more effectively, so they end up delivering better outcomes for clients. We see AI as something that supports our brokers and specialists, not something that replaces the advisory side of what they do.

Why has AI become such a live issue in insurance and reinsurance right now?

It can speed up how quickly information moves and help people get through large volumes of data. But insurance and reinsurance are still relationship-driven, advisory businesses at their core. Clients rely on expertise and trusted counsel for the decisions that matter most. AI can support that, rather than replace it.

In practice, what has Wording Analyzer actually changed for your teams?

The obvious change is time. A review that used to take the better part of a day can now take a matter of minutes, so people are covering more contracts and spending more of that freed-up time on the clauses that actually need specialist judgement, rather than the mechanical parts of the review.

The other change is less visible but matters just as much. We can now benchmark clause usage against our own reference library and against the market, which gives clients something concrete to work with when they're negotiating with reinsurers. The tool also flags gaps or inconsistencies in a submission, though someone on our side still has to decide what to do about them.

Was this a case of bringing in an off-the-shelf AI product?

No, and that was deliberate. We started from the view that generic AI does not meet the demands of reinsurance work. From summer 2025, our teams worked closely with Dylogy to adapt the tool to how a reinsurance broker actually operates, testing changes directly with the people using it. Dylogy's speed in responding to that feedback was a big part of why the relationship has continued and why we are now extending it. Their experience supporting clients outside France, including in the UK and Bermuda, also mattered, since it meant they could support a wider international rollout rather than a single-market pilot.

Some in the industry argue AI will disintermediate parts of the value chain. Do you see that risk here?

There may be pockets of the market, particularly for highly standardised risks, where automation plays a bigger role over time. But in commercial, specialty and reinsurance business, clients are dealing with increasingly complex challenges that need specialist expertise and nuanced advice, whether that's structuring a programme, negotiating terms, or making the case to a reinsurer. That's still down to people, not the tool.

What's next, both for this partnership and for AI in reinsurance more broadly?

Having seen the results in France, we're rolling the partnership out progressively across Howden Re's other jurisdictions. Longer term, we're also looking at extending it into claims analysis and coverage assessment, which is another area where Dylogy has relevant experience.

More broadly, I think we're still early in our understanding of exactly what AI can do for this industry, and I'd be wary of anyone who claims to know exactly how it will develop. The organisations that do well with it will be the ones that use it alongside industry expertise, not instead of it.