Successful AI implementation means improved combined ratio and margin, not layoffs

The re/insurance industry spent $70 million on AI in 2023 and $300 million in 2024 yet productivity has not followed the spend.

So says mea Platform CEO Martin Henley, who told Intelligent Insurer’s Re/insurance Outlook USA Conference on May 12 in New York that the ultimate metrics of AI success should be improvements in margin and combined ratio, rather than document automation or headcount reduction.

Henley cited the efficacy gap in the re/insurance model where, for every dollar of premium for a typical carrier or cedant, 65 to 70 cents is apportioned to losses – claims adjusting; 12 to 15 cents for acquisition – brokers’ commission; 12-14 cents for operations – underwriting workflow, policy administration and claims handling; leaving only 3-5 cents for margin.

“So, in most years, in most lines, your operations are costing more than your underwriting margin. And this is not a soft number, this is an S&P number, a McKinsey number, a III [Insurance Information Institute] number, and it’s held roughly constant for the last 15 years,” Henley said.

He quoted the Accenture underwriter survey which found that underwriters spent 40% of their time on administration, 30% on sales support and negotiation and 30% on actual underwriting. 

“Two-thirds of your underwriter’s time is not spent on what you’re paying them to do… underwriting, and this is a $17-$32 billion every year efficiency gap in the industry,” he said.

Henley added that underwriters are therefore not able to model AI liability across the portfolio, unable to model cyber aggregation, and not able to close the social inflation gap reality with the actuarial models being used. The best people, he said, are “reconciling spreadsheets”. 

Henley advocated AI should handle repeatable “always-on” tasks such as submission ingestion, triage and endorsement processing. He said best talent should be freed from spreadsheets to focus on understanding mutating risks, managing aggregations and making complex underwriting judgements.

He said an emphasis on people managing risks and AI handling the repeatable leads to loss and expense ratio improvement. 

“These two combine and compound, four points on combined ratio conservatively. As we know, four points on combined ratio is $200 million on a $5 billion book, $800 million on a $20 billion book,” Henley added.

Fundamental to this he said was closing the loop between operational capability and the combined ratio “and you can only do tis if the technology is built specifically for insurance, from the first line of code. Not adapted, not retrofitted, but built”. 

Henley said the mea Platform language model has been trained only on insurance and its knowledge graph understands the language of underwriters, brokers and claims handlers. He said fact that it was built for insurance explains why the platform has $200 million of GWP running through it.

By optimising the end-to-end workflow through specifically built AI, underwriters stop being the bottleneck in the system and start becoming the judgement layer the system delivers to, Henley said. 

“If you remove the repeatable from the underwriters they can write 40% more in-appetite GWP at no extra effort,” he added.

Carriers and brokers who successfully make this AI shift will be the ones to define the next decade of insurance economics, Henley concluded.

Did you get value from this story?  Sign up to our free daily newsletters and get stories like this sent straight to your inbox.