AIG Assist Delivers 40% Binding Lift and 55% Faster Underwriter Quotes Across Eight Lines of Business with Multi-Agent AI Underwriting

Title

AIG Assist Delivers 40% Binding Lift and 55% Faster Underwriter Quotes Across Eight Lines of Business with Multi-Agent AI Underwriting

Content

AIG, the global insurance organization, presented Q1 2026 earnings metrics for AIG Assist — a multi-agent AI underwriting system deployed across eight lines of business — showing a 30% improvement on quoting more submissions, a 55% reduction in time to quote for underwriters, and approximately 40% higher binding of submissions. The system is orchestrated on Palantir Foundry and uses Anthropic's Claude via AWS Bedrock. The deployment is built on Palantir Foundry orchestration with Claude via AWS Bedrock, allowing AI agents to handle document ingestion, risk evaluation, and pricing recommendations across multiple specialty and mainstream lines of business. AIG framed the system as a multi-agent underwriting layer designed to free underwriters from manual triaging and accelerate the most consequential step of the insurance funnel: binding the quote. Traditional underwriting flows at AIG required underwriters to manually review submissions, gather supporting documentation, apply judgment to risk factors, and prepare quotes — a workflow that constrained throughput and limited AIG's ability to quote on the long tail of submissions. The combination of slow turnaround and inconsistent follow-through reduced bind rates even when pricing was competitive. AIG Assist was deployed across eight lines of business in Q1 2026 and the Q1 earnings call reported the quantifiable lift. The platform's premise is that AI agents can take on the document ingestion and first-pass risk evaluation work that dominates underwriter time, leaving humans to handle judgment-heavy exceptions and broker relationships. The architecture is intentionally multi-agent so that each step of the underwriting pipeline can be specialized rather than relying on a single generalist model. The technical stack layers Palantir Foundry as the data and orchestration backbone, AWS Bedrock as the model gateway for Claude, and a multi-agent design that separates document ingestion, risk evaluation, and pricing recommendation into distinct agents. Anthropic's Claude is the underlying language model for the agents, chosen for its reasoning performance on long-context insurance documents and structured output reliability for risk classification. Operational impact delivered a 30% lift on submissions quoted, a 55% reduction in time to quote for underwriters, and approximately 40% higher binding rates across all eight lines of business running in Q1 2026. These improvements compound: more submissions quoted means more broker opportunities; faster quote turnaround means brokers route more business to AIG; higher bind rates mean more of that quoted business actually converts to written premium. AIG framed the Q1 2026 result as proof that multi-agent AI underwriting can shift the economics of specialty insurance at scale. The next reported phase is expected to extend the agent footprint to adjacent workflows — first notice of loss, claims triage, and broker servicing — with the same Palantir + Claude + Bedrock foundation.