Automating medical evidence ingestion for AI-driven appeals

Our client, Claimable, provides a GenAI-powered platform that generates customized, evidence-based appeal letters for denied health insurance claims. As the company expanded toward enterprise-scale customers, manually onboarding clinical policies and citations became a major bottleneck.

Client challenge

Manual ingestion processes caused inconsistent throughput, two-week onboarding cycles, and difficulty prioritizing high-quality clinical citations. Our client needed explainable, auditable GenAI workflows that could operate reliably within regulated healthcare environments.

Our solution

Turnberry Solutions partnered with Claimable to deliver a production-ready, AWS-native ingestion and classification pipeline in five weeks. The solution automated ingestion, normalization, and classification of clinical policies, while low-confidence cases were routed for human review to ensure accuracy.

The pipeline used AWS-native services, including Amazon Bedrock for LLM-assisted classification, AWS Lambda for ingestion, and Amazon SQS for scalable orchestration. Metadata and clinical criteria were stored in Amazon RDS for PostgreSQL, supporting rule-based matching, while a Promptfoo-based framework enabled prompt testing and smooth deployment

This engagement showed that GenAI, combined with deterministic controls, confidence scoring, and human oversight, can safely accelerate healthcare workflows. The solution eliminated manual bottlenecks, shortened onboarding cycles, and created a scalable foundation for Claimable’s growth.