The federal government is placing a high-stakes wager on artificial intelligence to solve America's cardiology crisis. ARPA-H has committed $62.7 million to develop autonomous AI agents capable of providing 24/7 heart failure care—a bold attempt to address the fact that nearly half of U.S. counties have no practicing cardiologist.

The four-year Advocate program represents a fundamental shift in how clinical AI could function. Rather than serving as decision-support tools that assist physicians, these AI agents are designed to operate as independent members of the care team, monitoring patients continuously between clinic visits and intervening when necessary. If successful, they would become the first FDA-authorized autonomous AI tools for clinical care—a regulatory milestone no company has yet achieved.

Six Teams Enter a Winner-Takes-All Race

ARPA-H has divided $33.7 million in first-year funding among six teams, each tackling different aspects of the challenge. Kaiser Permanente received $16.3 million and will deploy its solution across 21 medical centers and more than 260 clinics. Duke University secured $15.5 million to test AI agents across five health systems using Epic and Oracle EHR platforms.

Three AI vendors are competing to build the actual autonomous agents: Atman Health ($7.7 million) is developing a voice-based interface, Tempus AI ($9.5 million) is focusing on continuous health change detection, and Updoc ($9 million) rounds out the commercial competitors. Stanford University also received funding, while Johns Hopkins University Applied Physics Laboratory will serve as independent evaluator.

The program borrows DARPA's competitive model, where funding continues only if teams meet aggressive milestones. Non-performers face elimination, creating pressure to move fast toward the two-year deadline for FDA submission.

Targeting America's Cardiovascular Care Desert

The urgency behind the program stems from stark geographic disparities. Counties without cardiologists show higher rates of cardiovascular disease, even as heart disease remains the leading cause of death nationwide. ARPA-H's vision is for AI agents to fill this gap, particularly in rural areas where specialist access is limited or nonexistent.

The program is structured around three focus areas: developing patient-facing AI agents, creating supervisory safety systems to oversee autonomous decisions, and establishing deployment strategies that can scale across diverse healthcare settings. Teams must also work with FDA to develop an entirely new regulatory framework, since no precedent exists for authorizing AI tools that act independently rather than supporting human decision-making.

What Happens Next

Duke's testing across multiple EHR systems and Kaiser Permanente's large-scale deployment will provide critical real-world data on whether autonomous AI can safely manage complex chronic conditions. Heart failure patients in these pilot programs will be the first to experience AI-powered cardiovascular monitoring that operates continuously, potentially catching deterioration before it becomes life-threatening.

The program's success or failure will likely determine whether agentic AI becomes a standard component of clinical care or remains confined to decision-support roles. For the three competing AI vendors, FDA authorization would establish market leadership in an entirely new product category—and potentially open pathways for autonomous AI across other medical specialties.

Sources

  1. ARPA-H to invest $62M to build agentic AI agent for heart care — Healthcare Dive