More than half of insurers now say AI is delivering measurable process efficiencies, but data management remains the dividing line between carriers that scale AI successfully and those that don't, according to a new study from professional services firm EXL.
The survey found that 54% of insurers are achieving process efficiencies at notable scale with AI, and 51% are using AI to target and attract new customers. Nearly half - 46% - have fully deployed AI in actuarial and underwriting functions. Insurance also leads all industries in moving AI pilots into production, with 62% reaching scale.
Data is the biggest bottleneck
Despite those gains, data problems persist. More than half of insurers - 56% - say their data is a challenge to AI success, and 38% cite data silos as the top barrier. Only 24% of insurers consider themselves leaders in data management maturity.
"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," said Rup Goswami, insurance growth office lead at EXL. "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."
The gap between AI leaders and laggards in the study is largely a data story. EXL found that 91% of leaders rate themselves ahead in data management maturity, versus 61% of laggards. More than 80% of laggards still operate in siloed environments that limit what AI can do.
The data-AI loop
Insurers are ahead of other industries because they've rejected the sequential "data-first" approach, Goswami said. Instead, they improve data quality and deploy AI at the same time - a two-way relationship between AI for data (normalizing, enriching, reconciling siloed records) and data for AI (cleaner inputs that reduce processing cycles, prompt engineering and token costs).
The functions furthest along - fraud detection, customer servicing, risk management, actuarial and underwriting - share a common trait: they run on structured, high-volume data pipelines, which signals where carriers should prioritize data investment next, he said.
Carriers that want to improve their pilot success rate should focus on areas where data is already cleanest, and pair each new pilot with an executive sponsor who owns the business outcome, Goswami said. They should also design for production integration from day one rather than retrofitting after a proof of concept succeeds.
Why this matters for insurance professionals
Underwriting and claims teams are among the primary beneficiaries because these functions already have structured data pipelines. For professionals in those roles, the practical takeaway is to build data hygiene with these processes before extending AI to new areas. Aligning with organizations that treat data as a business priority rather than an IT afterthought positions professionals on the side of the 62% pilot-to-production leader, not the laggards. Training on AI for Insurance Courses and AI Data Analysis Courses can help professionals bridge the data competency gap that separates the leaders from the rest of the field.
Your membership also unlocks:
