AIA

Submitting and processing medical insurance claims required significant time and effort from both customers and insurers, involving manual paperwork, meticulous documentation, and expert review. Claims adjudication required reviewing multiple documents in various formats — multilingual, handwritten, with specialized medical terminology — making the process slow and labor-intensive. Fraud, waste, and abuse (FWA) detection was also a persistent challenge across markets.

AIA developed a three-tiered Gen AI system: (1) a Gen AI-powered OCR solution that extracts key information from photographed invoices with 97% accuracy, supporting multiple languages and handwriting recognition; (2) a Gen AI summarization layer that condenses complex medical reports for claims assessors, cutting review time per case in half; and (3) an AI-driven FWA detection system using 20+ risk indicators calibrated by an encoder-decoder anomaly detection model. The system was built in-house and rolled out across 4 major markets.

End-to-end claims straight-through processing improved from 22% (June 2020) to 73% (December 2024), and auto-adjudication rates rose from 41% to 75%. E-payment adoption reached 99%. In Korea, Gen AI reduced the claims process from two days or more to under 25 minutes. The OCR solution contributed to AIA achieving a top Net Promoter Score in Hong Kong for ease of use and customer satisfaction.

• In-house Gen AI development enabled AIA to build OCR that outperforms standard solutions by handling non-standard formats, multilingual content, and handwriting — capabilities off-the-shelf tools couldn't provide.
• A three-tiered automation approach (document ingestion, adjudication assistance, fraud detection) multiplies efficiency gains across the entire claims lifecycle rather than optimizing a single step.
• Migrating 90%+ of compute to the cloud was a prerequisite that enabled the rapid, cost-effective AI innovation that followed.