Capgemini: 35% of insurers use AI

Insurance companies are not debating whether to use AI anymore. That part is basically over. The real problem is that most of them still have AI sitting in demos, pilots, and side projects instead of inside the actual machinery of underwriting, claims, fraud detection, and customer service. That is the punch line from Capgemini’s 2026 World Property and Casualty Insurance Report, published this week. (capgemini.com) ### What is the actual news? The new report says only 35% of property-and-casualty insurers are linking AI work to real business outcomes. Nearly 60% are still in exploration or proof-of-concept stages, and 42% are not tracking AI metrics at all. So the headline is not “insurers are igno(capgemini.com)a meaningful way. (capgemini.com) ### Why is 35% such a big deal? Because insurance is one of those industries where a pilot is easy to start and hard to operationalize. You can build an AI tool that summarizes a claim file or flags suspicious submissions. But turning that into something trusted by adjusters, underwriter(capgemini.com)AI to outcomes, the sector is still early in the part that matters. (capgemini.com) ### Where is the bottleneck? Turns out the bottleneck is less the model and more the operating model. Capgemini says 72% of AI spending is going into technology and infrastructure, while only 28% goes to change management. That is like buying a new engine and never redesigning the car around it. The result is predictable — lots of capability, not enough adoption. (bfsi.economictimes.indiatimes.com) ### What is getting in the way? Legacy systems are the biggest drag. Capgemini says 81% of insurers cite them as a major constraint. Then come data-quality problems at 74% and regulatory complexity at 61%. Insurance runs on old policy systems, fragmented claims platforms, and rules that require explainability and audit trails. AI does not replace those constraints — it crashes into them. (bfsi.economictimes.indiatimes.com) ### Who is actually getting this right? A small group is. Capgemini calls them “intelligence trailblazers,” and they make up about 10% of insurers. These companies are not just deploying more tools. They are embedding AI into (bfsi.economictimes.indiatimes.com)ins between 2021 and 2024 than the broader field. (capgemini.com) ### Why does insurance have this problem more than some other sectors? Because insurance decisions are messy. A claim is not just a document-processing task. An underwriting file is not just a prediction problem. These workflows mix regulation, judgment, fraud risk, customer experience, (capgemini.com)ound the edges. The winners are redesigning work so human experts and “synthetic execution” operate together. (capgemini.com) ### So what changes next? The next phase is not more AI experimentation for its own sake. It is governance, measurement, and workflow redesign. Insurers need to decide where human judgment is essential, where automation is safe, and who owns the handoff between the two. Without that, AI stays impressive but peripheral. With it, AI starts to look less like a feature and more like operating infrastructure. (capgemini.com) ### Bottom line? The insurance industry has crossed the curiosity stage. But most firms have not crossed the execution stage. Capgemini’s report matters because it shows the gap clearly — AI interest is now common, but AI advantage is still rare. (capgemini.com)