Accelerating Healthcare Decisions with AI Agents

‍Overview:

In this AWS Podcast episode, Gigi Yuen, chief data and AI officer, and Kenji Fujita, staff AI platform engineer at Cohere Health, discuss how advanced artificial intelligence (AI) is transforming healthcare decision-making for payers, providers, and members. The conversation focuses on Cohere Review Resolve, an agentic AI-powered solution built on Amazon Bedrock AgentCore. Review Resolve uses clinically trained AI to boost efficiency for medical necessity reviews, improve accuracy through automatically surfaced clinical guidelines and insights, and provide clinical intelligence to guide reviewers.

Trust, reliability, and security remain core principles in Cohere Health’s work. Rather than presenting AI as a standalone tool, the team explores the critical shift toward evaluation‑driven development and the strategic integration of clinically trained experts. Cohere Health designs its solutions with strong privacy protections, continuous human oversight, and expert involvement at every stage. This approach helps ensure the technology supports safer patient outcomes, meets regulatory expectations, and strengthens operational efficiency across the healthcare ecosystem.

Key themes from the conversation:

1. Building trustworthy clinical AI

• Human oversight is a foundational principle, ensuring patient safety and clinical accuracy throughout development and deployment.
• Integrating clinically trained experts early aligns technology with standards of care and meets regulatory requirements.

2. Accelerating deployment and operationalizing AI

• A modular, secure architecture supported by cloud platforms such as Amazon Bedrock AgentCore enables rapid, scalable deployment while protecting sensitive health data.
• Built-in governance and integration features help streamline system updates, minimizing disruptions and maintaining compliance.

3. Driving measurable improvements

• Cohere Review Resolve, built on AgentCore, drives a 30-40% reduction in medical necessity review processing times, streamlining documentation analysis for complex clinical decisions.
• The company's AI-assisted prior authorization solutions support 85% real-time approvals, reducing delays and friction, while any cases that cannot be automatically determined are routed to a clinician for review.

4. Embracing evaluation-driven development

• Success requires establishing thoughtful metrics–such as accuracy, speed, and cost–before deployment to ensure the technology drives meaningful improvements.
• Expert review remains essential. AI is never used to make final decisions on complex cases, but to surface to reviewers automated, clinically appropriate information and guidelines to streamline the review process.