Case studies

Faster claims intake for a regional property and casualty carrier

An AI intake assistant that reads first notice of loss documents and routes each claim to the right team on the first pass.

Industry: Insurance
Client: Regional property and casualty insurer, United States
Service: Custom AI and Agent Engineering

Challenge

Claims handlers were spending the first day of every claim reading emails and attachments to decide where it belonged. Misrouted claims bounced between teams, and the backlog grew after every storm.

Approach

Mapped the intake process with handlers and built an evaluation set from past claims with known correct routes
Built a retrieval and extraction service that reads emails, forms, and photos and proposes a route with its reasons
Kept a handler approval step for low-confidence claims and logged every decision for audit
Integrated with the claims system through its existing API and ran in parallel for a full month before cutover

Outcome

Client names are withheld. Figures shown are illustrative while client approval is pending.

Less time from first notice to adjuster assignment: 62%
Routing accuracy on the evaluation set: 94%
From pilot approval to parallel run: 3 weeks

Stack

Azure OpenAI
Python
Azure AI Search
Guidewire ClaimCenter API
Power BI

More work