Marsh tests AI tool to model insurance risk in real time
Marsh is testing a new AI-powered platform called Project Leapfrog with select clients to help companies rapidly assess how business decisions affect their insurance coverage and risk exposure. The tool, currently in alpha testing, combines large language models, actuarial analytics and loss simulation technology to provide directional guidance on corporate risk strategy.
The broker unveiled the project at RISKWORLD 2026 in Philadelphia, positioning it as part of a broader suite called Risk Companion that launched this week. Risk Companion consolidates Marsh's separate analytics engines for casualty, property, cyber and directors and officers liability into a unified architecture powered by AI.
Real-time scenario modeling
Project Leapfrog lets clients and brokers test insurance program structures, adjust assumptions and compare outcomes instantly rather than waiting days for manual actuarial modeling. A company considering an acquisition, for example, could ask how adding millions of customer records would affect its cyber exposure and policy coverage. The tool would assess the client's current security controls, evaluate policy wording and rerun exposure scenarios in real time.
John Davies, Marsh's commercial director, said the company intentionally limits the tool to tightly defined tasks. "It's a guide, not a guarantee," Davies said. "It still needs to be checked. It's really about reducing manual workflow for the client."
Data advantage and governance trade-offs
Marsh's scale gives it an edge in building these models. The company draws on exposure, premium and claims data accumulated across its global client base, which feeds the AI systems with patterns and historical information that standalone tools lack.
The deployment of AI in risk strategy raises governance concerns. Even specialized AI systems can produce hallucinations and inaccurate outputs, and historical claims data can embed bias into models. Davies acknowledged these risks but said Marsh is designing the tool to reduce reliance on AI judgment for critical decisions.
Client interest has been strong, particularly among organizations already experimenting with generative AI tools. Marsh executives said 20 to 30 clients volunteered to beta test the platform at the conference, suggesting demand for this type of application.
The broader shift reflects how brokers and carriers across the industry are integrating generative AI into underwriting, risk analysis and client advisory functions. The goal, according to Marsh executives, is not to replace human expertise but to accelerate analysis that previously required days of manual work.
Learn more about AI for Insurance and AI for Executives & Strategy.
Your membership also unlocks:
