The data: The insurance industry has been an early mover on AI adoption, outpacing all industries but one among those studied in a BCG survey.

But progress has stalled at the pilot stage: the majority of insurers (67%) are testing genAI programs, while only 7% have scaled them—near the bottom of all covered industries. Another 27% haven't started at all.

Digging into the data: Use cases vary. One insurer cited in the study uses custom OpenAI GPT models to draft messages to claimants. Others are equipping employees with AI copilots for operations and customer support. But most insurers remain stuck in siloed projects, with budgets under $5 million. The 7% who've scaled are spending $25 million or more.

Digital baggage is preventing full deployment of AI initiatives. Most insurers are wrestling with data fragmented across underwriting, claims, billing, and distribution, and systems that don't integrate well with external sources—technical debt that predates the AI era. Richer, better-quality data is a prerequisite for the efficiencies that insurers seek. Insurers also cite other concerns: data security and privacy risks, limited in-house expertise, and doubts about genAI accuracy.

Our take: Insurers should modernize their data architecture and tie their technology investments to business outcomes. Opportunities abound, from claims automation to embedded distribution.

Insurers that make serious, sustained investments in innovation and customer experience will outcompete their peers. And the winners won't be those with the most experiments, but those who commit to scaling the ones that work.