How Cohere Health digitizes clinical policies using Amazon Bedrock AgentCore

Prior authorization, a crucial process in healthcare, remains a significant manual task due to policies being stored in unstructured formats. This content, central to day-to-day clinical operations, affects millions of patients annually and varies across clinical areas, geography, line of business, and health plans. Historically, health plans lacked a systematic approach to manage, analyze, and optimize these policies, causing a critical operational bottleneck that hindered the modernization of prior authorization operations at scale.

To address this issue, Cohere Health, a clinical intelligence company, developed Cohere Policy Studio using Amazon Bedrock AgentCore. This solution employs a multi-tenant agentic architecture with secure microVM isolation, unified tool access, and a managed agent runtime to accelerate policy digitization and scale capabilities while preserving transparency, version control, and human oversight.

The challenge lies in transforming static, unstructured policy content into structured, machine-readable data that AI systems can use consistently while still allowing medical professionals to review and maintain clinical judgment.

Cohere Health identified three key challenges in building an AI solution for prior authorization digitization: adhering to government regulations, such as CMS requirements for API-based electronic prior authorization by January 2027 and AHIP commitments for 80% real-time approvals; fulfilling the unique requirements of each line of business; and meeting technical architecture demands for ingesting multiple input formats and producing distinct policy representations for various downstream consumers.

To tackle these challenges, Cohere Policy Studio was built on AgentCore using the Agent Skills open standard, with three main tasks completed: deploying AgentCore Runtime with AgentCore Gateway and Memory for a full agentic system; configuring AgentCore Gateway to fetch tools and skills; and writing skills with clinical policy experts, evaluated using Cohere Health's standardized observability process based on Arize AI.