Zurich pilots generative AI to spot claims trends
06/10/2025
Jonathan Davis, data science lead at Zurich UK, has revealed we're currently piloting generative artificial intelligence to identify claims trends more effectively.
Article taken from interview held and posted by Insurance Post.
He explained Zurich is using AI to study the countless documents around a claim in detail and identify trends in order to anticipate problems before they happen.
The pilot will explore whether the trends identified can enable colleagues to actively help customers mitigate risk.
When asked what the objective of the pilot is, Davis said: “I want [the pilot] to surprise me and our colleagues across the business. It has to be able to find things we haven’t found before.”
He also pointed to the ability to be proactive on emerging issues as a goal of the pilot.
Davis added: “I know it’s not a percentage or a clear number, but that’s how I think it will work in practice.”
Data ingestion
Davis also highlighted that Zurich’s generative AI data ingestion tools are already live and being deployed across the business.
He explained before generative AI it was very challenging to train a model to process the various forms of unstructured data Zurich would receive but now the technology has the ability to understand and extract the relevant data, then present it to colleagues.
The high quality of the data output was highlighted by Davis as a primary benefit of the technology. He shared that this is often measured through positive feedback from business users.
Davis said: “We have an AI tool that helps bring the right documents to the relevant claims handlers. We’re finding that they’re a lot happier now that they’re getting the right documents at the right times. They’re also getting quicker, which has been really positive.”
He shared a similar example with underwriters and noted that they are far more satisfied when they don’t have to “fix” the data that comes to them.
“It makes people’s lives easier and it makes it possible to make decisions much faster,” Davis added.
"When someone in the group builds something, it’s used across the group, and we’re finding that really pays off."
This comes after research from Hyperexponential revealed that 79% of underwriters are worried about burning out, partly due to the sheer number of manual tasks they have to carry out day-to-day.
Group scale helps
Davis shared while it is challenging to go from an AI pilot to production, the shared resources provided by Zurich’s group approach to AI implementation have been “transformative”.
Zurich’s approach is to build in-house and then reuse the technology across the group, also known as “build once, use many”.
Davis explained: “This means that when someone in the group builds something, it’s used across the group, and we’re finding that really pays off.
He shared the number of uncertainties that have to be resolved can make going from pilot to production is more challenging, but being able to reuse the tool across the group makes it more effective, as it helps with both monitoring and keeping up with the pace of change.
