AIG Uses Agentic AI for Underwriting

- AIG’s "agentic AI ecosystem" is built in partnership with tech companies Palantir and Anthropic, leveraging a "digital twin" strategy that maps every data element across the organization to improve risk assessment. - For its North America private and non-profit business, the AI Underwriter Assistance tool has slashed the underwriting timeline from three or four weeks to less than one day and increased the bind-to-submit ratio from 15% to 20%. - CEO Peter Zaffino has positioned the AI initiative as a core business transformation focused on high-value activities like underwriting and claims, rather than a peripheral project for expense savings. - AIG's excess and surplus unit, Lexington Insurance, has already surpassed 370,000 submissions in 2025, advancing toward a goal of 500,000 by 2030, with projections to increase the bind rate from 2% to 6% in that time. - The AI models enhance fraud detection by analyzing vast, unstructured datasets—including claims forms, emails, and even social media—to identify subtle patterns and inconsistencies that traditional underwriting methods often miss. - This type of at-scale AI deployment requires significant leadership focus to overcome implementation barriers; an MIT report notes 95% of AI pilots fail not because of technology, but due to a narrow focus that overlooks the need to reinvent processes and integrate siloed data. - The company's 2030 roadmap includes creating an advanced orchestration layer to coordinate multiple AI agents, envisioning them as a "workforce multiplier" that can handle complex knowledge assistance, auditing, and decision support tasks. - Beyond underwriting, AIG used its generative AI platform to analyze and prioritize accounts during the integration of a renewal rights portfolio from Everest, completing the evaluation in a fraction of the standard time.