Newsletters
Risk Central
Digital EditionOnly 12% of Insurers Are Cashing In on A.I., but They Are Pulling Far Ahead
Just 12% of insurers have embedded A.I. across core operations, adopted it at scale and generated consistently high returns, while the rest remain stuck in disconnected pilots or broad experimentation without matching business impact, according to the IBM Institute for Business Value. Among this top tier, 69% report extensive A.I. use in fraud investigation, compared with just 21% of other insurers.
The findings come from a survey of 800 senior insurance executives across 31 geographies, alongside a consumer survey of 10,000 insurance customers across 17 geographies. Overall, the share of insurance A.I. projects delivering positive ROI reached 35% in 2025, and executives project that figure will climb to 45% in 2026 and 60% by 2029. Among the highest-performing insurers specifically, more than half of A.I. projects are expected to deliver positive returns in 2026.
Embedding, Not Experimenting, Drives Returns
The research grouped insurers into three stages:
- 55% are “getting off the ground” with low adoption and CIO-led pilots.
- 33% are “experimenting broadly” across many use cases without consistent returns.
- 12% have reached “high returns” with A.I. embedded as an enterprise capability.
What separates the top group isn’t how much A.I. activity is underway but how deeply it’s embedded in daily workflows, the institute found. High-return insurers report 62% extensive employee adoption of A.I. in claims, versus 18% among other insurers, and 47% in IT.
Scale matters too: 89% of insurers with more than $10 billion in gross written premium have reached production use of A.I., compared with 49% of those below $5 billion.
Yet the report found a mismatch between where insurers are investing and what customers want. Executives are concentrating spending on claims, fraud and IT, back-office functions, while consumers say they want better experiences and personalized policies.
Consumers are most receptive to A.I. in low-stakes interactions, with 63% preferring A.I. involvement in claims status updates and 60% in simple policy updates, but they want human judgment on consequential decisions such as pricing and claims payment. More than 90% expect transparency about A.I.’s role, and 40% support major penalties for insurers who misuse it, the institute found.
Architecture, Not Tool Count, Separates the Leaders
High-return insurers built their lead by layering a shared enterprise infrastructure for data, models, governance and orchestration above legacy systems, according to the report. While the average insurer runs 12 core applications with the oldest systems averaging 18 years old, and more than 80% of executives say IT complexity limits speed to market and drives up costs, only 29% of high-return insurers report complexity limiting speed to market and just 34% say it raises costs.
These insurers also distribute A.I. accountability across the C-suite rather than concentrating it in IT: they are more than twice as likely to name the CEO or COO as the primary decision-maker on A.I. strategy and nearly twice as likely to house their A.I. center of competence outside IT. They are also more likely to fund A.I. as an enterprise-wide capability, 41% versus 24% of the broader market, according to the report.
Trust is treated as a design requirement rather than a compliance afterthought among these insurers, with nearly 90% maintaining a written enterprise A.I. ethics policy, compared with 49% of the market overall. A.I. currently consumes 11% of insurer IT budgets, a share the report projects will reach 18% within three years, while the share of IT budgets dedicated to new capabilities rather than system maintenance is projected to rise from 29% today to 40% by 2029.
Obtain the full report here. &
