Capgemini report says most P&C insurers still stuck in AI pilots

The real split in property and casualty insurance is no longer between firms that are testing AI and firms that are not. It is between the few that have turned AI into an operating model and the many still treating it like a lab project. Capgemini’s 19th World Property & Casualty Insurance Report said only 10% of insurers had advanced AI capabilities in broad use, while 60% were still in exploration or proof-of-concept mode.

That gap is bigger than a technology problem. Capgemini said 42% of insurers tracked no AI metrics at all, and more than half were not clearly measuring return on investment. In practice, that means a lot of carriers are spending on models, pilots and infrastructure without the basic governance needed to know what is working. The report framed the divide as an “intelligence trailblazer” gap, with the winners aligning strategy, talent, technology foundations and adoption at the same time.

The difference shows up in results. Capgemini said trailblazers were seeing up to 21% higher revenue growth and about 51% higher share-price growth over three years. They were nearly four times more likely to invest in change management beyond basic training, nearly three times more likely to have explainable AI infrastructure, and almost twice as likely to bake AI accountability into job descriptions. That points to a hard lesson for software buyers: the value is not in the demo. It is in whether the product can fit into underwriting, claims and service workflows without creating a trust problem or a compliance mess.

Capgemini’s 2026 report was based on interviews with 344 senior insurance executives, four employee surveys covering 809 workers, and responses from 1,113 policyholders. Even with that broad view across the market, only 40% of P&C leaders said AI was meeting their expectations. Capgemini also said the industry is still directing too much AI spend toward technology and infrastructure, and not enough toward the change management that makes scaled deployment stick.

This is not a new failure mode. In Capgemini’s 2024 report, 62% of insurance executives said AI and machine learning improved underwriting quality and reduced fraud, but only 43% of underwriters said they trusted and regularly accepted automated predictive analytics recommendations. That trust gap matters because the business pressures have only grown, with catastrophes, social inflation, geopolitical uncertainty and macroeconomic volatility all squeezing underwriting and claims performance. Capgemini’s message is blunt: the industry is moving from “AI promise to AI advantage,” but only if insurers stop buying experiments and start building systems that people can actually use.