Insurance leaders are confident about their progress on artificial intelligence, but new research from KPMG suggests their organizations have changed far less than that confidence implies.
In KPMG International's report Unlocking AI value in insurance, 44% of respondents placed their organization in the top quartile for AI transformation, and none said they were significantly behind. Yet no organization in the survey had fully redesigned sales and distribution or underwriting around AI. Only 3% had done so in policy servicing or claims.
KPMG surveyed insurance leaders at organizations with 500 or more employees across 20 countries and six regions between May 20 and 29, 2026. It used a discussion-board format that combined structured questions with open-ended responses, and it has not disclosed how many people took part. The report also draws on 53 insurance respondents from KPMG's Transforming the Enterprise 2026 study.
Most leaders see a threat in standing still. Some 77% said failing to redesign their enterprise architecture for AI would hurt their competitiveness within five years, and 68% saw moving too slowly as a bigger risk than moving too fast.
Their use of AI tells a more modest story. According to the survey, 71% use AI mainly for content generation and routine task automation, while 29% run end-to-end processes through AI agents or automation.
"The gap between activity and transformation is where the real opportunity and risk now sit," said Dr. Frank Pfaffenzeller, global head of insurance at KPMG International. He said many insurers were still focused on efficiency rather than on how AI could change what kind of insurer they become.
Some 92% of respondents said AI was helping them raise productivity and cut operating costs, while 25% were using it to grow through new products and services. Close to half of AI budgets go to operational and back-office efficiency, compared with 5% to 10% for new products and revenue models.
Measuring the payoff is another weak point. Only 11% said they had a very clear view of the return on their AI investment, and 23% reported limited clarity or none. KPMG cautioned that this leaves insurers judging progress by activity and adoption rather than by results such as lower costs, faster cycle times or growth.
The findings are consistent with the Capgemini Research Institute's World Property & Casualty Insurance Report 2026, published in May. It found that 60% of P&C insurers were still at the exploration or proof-of-concept stage, and that only 10% had successfully scaled AI.
Only 11% of KPMG's respondents said their data foundations and governance were ready to take AI beyond pilots. A further 55% described themselves as moderately ready, 21% as partially ready and 13% as not ready. They cited fragmented and poor-quality data, unclear ownership and legacy systems.
Workforce readiness is also low. Although 54% said they offer effective AI training, only 8% rated their staff as highly proficient with AI tools. Technology leaders own AI in 45% of organizations, and only 15% have fully built AI governance into strategic planning.
Respondents still expect significant changes to jobs. By 2029, 72% expect underwriting to run on a hybrid model with fewer people and redesigned roles. Some 36% anticipate significant role elimination in claims and 33% in policy servicing.
Matthew Smith, global lead for insurance strategy and transformation at KPMG in the UK, said the insurers most likely to make progress would be those that combine trusted data, clear accountability and a prepared workforce with a long-term outlook.
The zero is the number that matters most for intermediaries. With no carrier in the survey having rebuilt distribution around AI, and with budgets aimed mostly at internal costs, the carrier side of submissions, quoting and placement is likely to stay largely manual for now.
The underwriting forecast suggests that won't last. If most carriers expect smaller, hybrid underwriting teams by 2029, brokers can expect fewer underwriters to deal with and more decisions triaged by automated systems. Agencies with clean data and processes that connect easily to carrier systems will be the first to benefit when that shift arrives.
