WTW, an advisory, broking, and solutions firm, has been closely tracking the adoption of artificial intelligence (AI) in insurance, and has highlighted how carriers are now deploying related technologies at a more rapid pace.
Laura Doddington, WTW’s North American Head of Personal and Commercial Lines in Insurance Consulting and Technology, explains how insurers are moving from experimentation to widespread implementation of AI to improve operations, reduce costs, and provide more personalised customer services.
Doddington explains that the use of AI across marketing, pricing, underwriting, and claims is not new, but insurers are now quickly exploring and deploying generative AI technologies.
She commented: “We are no longer the laggards of the financial services sector. Insurers have increasingly dedicated AI R&D resources focused on identifying where the best opportunities are for its use across the insurance value chain, and, frankly, they have little choice!”
Over the years, pricing teams have relied on advanced statistical methods such as generalised linear models to predict claim frequencies, claim severities, retention, and conversion rates. Platforms like WTW’s Radar have helped insurers rapidly build, refine, test, and deploy these models. Doddington notes that while these approaches remain central because of their transparency and explainability, AI has added a powerful new dimension to analytics.
“Whether it’s the use of machine learning models in predicting claim amounts or liberating unstructured data to provide a significantly greater pool of information to derive insights from, AI in insurance is now mainstream,” Doddington added. “We’ve helped insurers use AI to provide decision support at the point of underwriting, triage claims, identify underwriting and claims fraud, and pinpoint claims that might unexpectedly jump without early intervention. It’s very exciting!”
Doddington warns that scaling AI is not without its challenges. She noted: “A key requirement will be the need to monitor the performance of models once they are implemented, an activity that will add an increasing burden on analytics teams as the size of the model estate grows.” Proof-of-concept projects must be industrialised responsibly, with governance controls to prevent untested or unethical models from influencing day-to-day operations.
Generative AI, particularly large language models, is an area where Doddington sees immediate benefit. She explained: “These models can parse unstructured underwriting and claims data, or call centre transcripts, and translate, summarise, and analyse text, speech, and video data.” This enables insurers to access previously untapped data, improving automation, speed, and customer satisfaction.
Doddington refers to practical applications. “A great example of this is how insurers are using machines to analyse photo and video footage of property damage—for example, following an auto collision—to instantly decide whether a write-off offer can be made at the point of first notice of loss.” She added: “Other insurers use machines to read external reports to provide underwriters with useful insights they may have missed, often in the form of scores.”
Looking ahead, Doddington highlights the potential of agentic AI, which can act autonomously and pursue goals without human intervention. “This is a significant difference from generative AI and other previous AI classes insurers have readily adopted. I struggle to see agentic AI fully replacing jobs in what is a highly regulated industry where advice, accuracy, and human judgement are critical, but the nature of work will certainly change.”
“I’m already hearing from insurers who are beginning to experiment with it in various low-risk applications. But the need for guardrails is essential. Humans must review and validate agentic AI output and ensure it is used ethically and responsibly.”
WTW’s Radar technology underpins much of this innovation. Doddington explains that with the launch of Radar 5, the most advanced version yet, insurers can combine decades of expertise with generative AI to transform pricing, portfolio management, and insight generation.
By combining statistical techniques, machine learning, and generative AI in one SaaS-enabled system, Radar 5 supports insurers in introducing AI with proper oversight and clarity. Doddington views it as setting a fresh benchmark for innovation in a regulated market.
