An arms race for data and AI – quantum leap needed

The Capgemini Research Institute released its “World Property and Casualty Insurance Report 2024” in April, providing an update on underwriting capabilities, how insurers are being affected by organisational constraints and the uptake of technology.

The institute’s global industry leader for insurance, Adam Denninger, (pictured) sat down with GR to discuss some of its findings – particularly the adoption of aspects of artificial intelligence (AI).

One of the key discoveries has been that only 8% of underwriters have looked to embrace AI technologies such as predictive analytics and machine learning, or to integrate diverse data sources, including third-party data and data from connected devices.

This is a subject that Denninger is passionate about.

“People are embracing technology, but the question is how successful they have been in executing on it,” he said.

“Some lines have much further to go, so it’s possible that you have people in some lines who are going to take a quantum leap on it. I think Lloyd’s is in this process – they were very manual and very traditional, but all this investment that’s going on, as part of the market’s digital transformation is a twenty-five-year step.”

He added: “Is it going to be longer and harder to do it than they thought it would be? Absolutely. But are they spending the money on it? Absolutely. They’re working on it and the question is around how long it will take to achieve the transformation the market wants, which is a difficult question to answer.”

Market attitudes have shifted considerably, Denninger said, with widespread recognition that the various tools and forms of automation, usually grouped together as AI, can be used to make more decisions better and faster. The questions are around choosing the right applications for the right business.

There is an attitude at the top of organisations that technology advancement is something that should be increasingly embraced, Denninger emphasised. However, it is more difficult to see this cascading down into the lower levels, he warned.

This was a discord marked in the report by the fact that while some 62% of insurance executives believed AI would enhance their underwriting ability and reduce instances of fraud (largely on the claims side), this was undercut by only 43% of underwriters regularly accepting automated recommendations.

“The rank and file are often afraid of being replaced by a machine. But aren’t we all, to some extent? But there is also the issue about what if the AI gets something wrong?” he said.

“Let’s imagine that I outsource a decision to a generative AI tool, or we use something like that for underwriting, and that tool takes a decision that is wrong, and goes down the wrong path. You’re always going to be a little bit wrong when it comes to underwriting, but you don’t want to be dramatically wrong on a regular basis.”

A big issue around AI in this field, concerns ‘adverse selection’, Denninger explained. This quickly leads to underwriting losses, he warned.

“What this means is that if my pricing or underwriting models are inaccurate, then I’m going to under-charge for bad risk or overcharge for good risks in comparison to my competitors,” he said.

“That means I’m going to get, over time, all the bad risks that I’ll then be undercharging for them. That’s adverse selection.”

The current landscape, Denninger concluded, is an arms race around safety, sophistication, and underwriting.

Vast resources are being poured into data and technology projects, meaning also that companies have to continually invest, and at the heart of it is the quality, availability and analysis of data.

“What these executives are thinking about is how to get data—more data, better data—that can help them make better underwriting decisions,” he added.