Can generative AI (GenAI) deliver real value in general insurance today? That was the question under discussion during a ‘power lunch’ discussion at the 2025 Insurtech Insights Europe conference where Guidewire’s Laura Drabik spoke to Tom Wilde, CEO of Indico Data and Terry Buechner, global insurance core systems lead at AWS about the practical application of AI in insurance.
Contextualising where AI is adding value and where it’s in danger of becoming overhyped, Wilde noted that last year about 25 million people globally identified as software engineers. When GenAI arrived, it essentially enabled anybody able to type at a computer to become a programmer. That’s the scale of the disruption, he said, and it’s what is so profound about the advancement of AI.
“I think the challenge is how can you control that? Because you wouldn’t, as an insurance company, declare overnight that everybody is allowed to write software to change pricing models,” he said. As a result, the question for the market is how to make sure that this technology is pointing in the right direction, which means ensuring that the right controls are in place around it.
Where GenAI really excels is in areas including summarization, he said, including summarizing underwriting models and claims models with unprecedented results. “Some others are that, for the first time, we can take an unstructured object like underwriting guidance and turn that into a programmatic endpoint where now software can talk to a document and vice versa. We were never able to do that before, this is truly a breakthrough.”
Offering his perspective, Buechner also emphasized the capacity of AI to ingest and summarize documents at scale. The hype around AI is how it can be used to transform unstructured and structured document-intensive processes like claims and underwriting, he said. The hype is there for a reason and where it really fits is in the claims space, whether that’s verifying IDs for the first-notification-of-loss or being able to provide first-call resolution.
It’s clear that when used effectively, GenAI will result in faster, more transparent claims processes, he said. While the same applies on the underwriting side in terms of document ingestion and summarization, where he has seen some “fuzziness” is with regard to the move towards a fully automated underwriting process.
While that may happen, it’s likely to be a few years yet down the road, particularly in commercial lines where there are significant complexities involved. “There’s still the need there for a person to be involved,” he said, “for human judgements and the experience that comes with being in the insurance industry for many years. But it’s a very fast-moving area so that may change. But it’s not going to change overnight.”
Drabik highlighted research from Deloitte which revealed that 76% of insurers have implemented generative AI in at least from business function. “But the business benefits that we’re chatting about are only attainable if insurers can actually scale up successful initiatives…. And only 15% have actually scaled effectively.” The challenge for insurance companies today is how to scale up from pilots to enterprise-wide adoption which begs the question - what separates carriers who successfully scale up from those who get stuck in endless proof of concept?
It is easy to get stuck at that point, but there are routes out, Buechner said. He cited a recent study from Boston Consulting Group which revealed that only 26% of the companies surveyed have developed the necessary set of capabilities to move beyond proofs of concept and generate tangible value. And realistically, he said, that figure is probably significantly lower in the context of the insurance industry.
“So what can we do to try and scale it up?” The first step, he said, is to start with a project that embraces the basics of where GenAI excels. Think about the problems that you’re trying to solve or the overarching processes you’re looking to transform, then work backwards from that in order to build up from that foundation. “Have clear goals in mind and have measurable success factors in mind,” he advised. “A big part of this is about having a test-and-learn culture… And that’s not really part of the traditional DNA of insurance.
“So, build in that culture of test-and-learn where, not only do we build upon the capabilities [unlocked] by starting small and building upwards from there, but we’re also able to pull back if it’s not working because the technology has changed or the regulations have changed. We have to be able to scale up when something is working well but also to scale back down when it isn’t.”
