Most insurers are deploying AI in customer-facing and operational areas, but data inefficiencies remain a major barrier, according to EXL's latest AI study.
Over half of insurers, 54%, are
"What the data tells us about insurance is that the industry has stopped asking whether AI works and started asking how fast they can scale it. That shift in mindset is showing up directly in the results. No other industry in our research is simultaneously leading on cost reduction, revenue growth, margin improvement and pilot success rates," said Rup Goswami, insurance growth office lead at EXL. "The common thread is that the carriers getting the most out of AI have treated data infrastructure as a business priority, not an IT project. The ones that close that gap fastest will do it by putting AI to work on the data itself— cleaning it, connecting it and making it accessible—because in insurance, better data doesn't just improve your models, it lowers the cost of running them."
Data inefficiencies remains a top concern for insurers: 56% say their data is a challenge
to AI success, and 38% cite data silos as the top barrier. As one of the main obstacles to abstracting more value from
"Insurance companies are ahead of most industries precisely because they've rejected the sequential 'data-first' approach, more than any other sector, they improve data quality and deploy AI simultaneously, AI for data—using AI to normalize, enrich and reconcile siloed records—and data for AI, like cleaner inputs that reduce processing cycles, prompt engineering and token costs, are two sides of the same investment," said Goswami. "The areas where insurance is furthest along, like in fraud detection, customer servicing, risk management, actuarial and underwriting, share a common thread: they're functions with relatively structured, high-volume data pipelines, which tells carriers where to prioritize data investment next. The dividing line between AI leaders and laggards in the study is almost entirely a data story: 91% of leaders rate themselves ahead in data management maturity versus 61% of laggards, and more than 80% of laggards still operate in siloed environments that directly cap what AI can do."
Compared to companies from other industries surveyed, insurance companies deploy more AI pilots into production than any other industry, with insurers seeing 62% of AI pilots scale to production. Many still stop at the pilot stage, however, according to EXL.
"Insurance already leads all industries at a 62% pilot-to-production success rate, and the report is explicit about why: high-quality, accessible data is the single most-cited success factor, meaning carriers that have invested in breaking down data silos are the same ones successfully scaling pilots. The other differentiator is that insurance companies are more focused than any other industry on integrating AI directly into core business processes like underwriting, claims and pricing, rather than running AI as a standalone capability, which creates the operational momentum and executive urgency that moves a pilot into production," said Goswami.
"Carriers looking to improve their success rate should scope new pilots around the functions where their data is already cleanest, pair each one with an executive sponsor who owns the business outcome and design for production integration from day one rather than retrofitting it after a proof of concept succeeds," he explained. "Choose the right spots to get it right."
