Manual triage of submission emails and attachments were time-intensive and inconsistent
Underwriters at a US mutual insurer were spending significant time manually pulling data from emails, PDFs, Excel files, and Word documents just to triage a single submission. Risk evaluation was inconsistent across the book. They needed AI that worked the way underwriters do – fast, thorough, and structured.
Built GenAI-powered Underwriting Assistant to process submission emails and extract insights from PDFs, Excel, and Word
Combined submission data, underwriting guidelines, and external web sources into a unified risk summary
Evaluated AWS Bedrock Knowledge Base and OpenSearch for near-real-time inference
Entire solution built on AWS serverless for event-driven, scalable, cost-efficient architecture