Cohere Health digitizes clinical policies with Amazon Bedrock AgentCore
Amazon Web Services
Cohere Health
Arize AI
Cohere Health built Cohere Policy Studio using Amazon Bedrock AgentCore to digitize clinical policies for prior authorization. The system reduces policy digitization time by 30% and accelerates agent deployment from 3-4 months to 2-6 weeks, using multi-tenant isolation, unified tool access, and a modular skills framework.
Cohere Health, a clinical intelligence company, developed Cohere Policy Studio using Amazon Bedrock AgentCore to digitize clinical policies that govern prior authorization. The application leverages AgentCore Runtime's MicroVM isolation for multi-tenant security, AgentCore Gateway for unified tool access, and AgentCore Memory for session feedback loops. The team deployed agents using reusable Amazon ECR base images, wrote skills with clinical policy experts, and evaluated them using observability based on Arize AI. Results showed a 30% reduction in policy digitization time, from 2 hours 15 minutes to 1 hour 35 minutes per policy, and faster agent deployments, from 3-4 months down to 2-6 weeks. The system supports government regulations requiring API-based electronic prior authorization by January 2027 and AHIP commitments for 80% real-time approvals. Skills use dual-layer versioning with semantic versioning and Amazon S3 object versioning, and a staged deployment pipeline with CI/CD. The architecture enables clinical policy experts to author and refine skills directly, maintaining human oversight and governance.
- Abbreviations
- CMS = Centers for Medicare & Medicaid Services — Центры по услугам Medicare и Medicaid
- AHIP = America's Health Insurance Plans — Американские планы медицинского страхования
- API = Application Programming Interface — программный интерфейс приложения
- MCP = Model Context Protocol — протокол контекста модели
- ECR = Elastic Container Registry — реестр эластичных контейнеров
- S3 = Amazon Simple Storage Service — простой сервис хранения Amazon
- CI/CD = Continuous Integration / Continuous Delivery — непрерывная интеграция и доставка
- ML = Machine Learning — машинное обучение
Source: AWS ML blog —
original
