The Advanced Research Projects Agency for Health (ARPA-H) awarded contracts to six teams Wednesday to build an FDA-authorized, autonomous AI tool for treating heart failure, committing $33.7 million in the first year of a four-year, $62.7 million program. The initiative targets a stark gap: nearly half of U.S. counties have no practicing cardiologist.
Recipients include Kaiser Permanente, Duke University, Stanford University, and health AI vendors Atman Health, Tempus AI, and Updoc. Teams will build a patient-facing AI agent that functions as a "digital member of the clinical care team," a second supervisory agent to oversee it, and a deployment plan for scaling the technology across health systems.
How the competition is structured
The Advocate program, first announced in January, splits work into three focus areas. Under the first, three health tech companies compete to build an AI agent that will be submitted for FDA authorization within two years. Atman Health will use up to $7.7 million to advance its large language models with a voice-based interface for heart failure patients. Updoc received $9 million to develop a conversational AI assistant, while Tempus AI will use $9.5 million to extend its existing patient health app with continuous monitoring features.
Under the second focus area, Stanford researchers will build a system to catch unsafe recommendations from the patient-facing AI tool. Under the third, Duke University and Kaiser Permanente will map deployment strategies. Duke received $15.5 million to create a testing platform across five health systems using Epic and Oracle electronic health records. Kaiser Permanente will use $16.3 million to deploy the agents in its Epic EHR workflows across 21 medical centers and more than 260 clinics.
ARPA-H's funding model mirrors the Defense Advanced Research Projects Agency: total award amounts depend on hitting aggressive research milestones, and teams that miss them can be dropped from the program. The Johns Hopkins University Applied Physics Laboratory will serve as an independent evaluator.
A regulatory first for clinical AI
The FDA has not yet approved an agentic AI tool for clinical care. Program leaders said they'll work alongside the agency throughout the program to develop a regulatory framework for this class of clinical AI. If successful, the effort could create an approval pathway for future applications of clinical-grade AI Agents & Automation in medicine.
The patient-facing agent is designed to provide 24/7 autonomous cardiovascular care between visits to real clinicians, fielding questions independently and only engaging human providers when necessary. This push into AI for Healthcare comes as the Trump administration has advocated for expanded use of clinical AI.
Why this matters for healthcare professionals
For clinicians and health system administrators, the Advocate program signals that agentic AI is moving from concept toward FDA review on a concrete timeline. The involvement of Epic EHR workflows at Kaiser Permanente and Duke's multi-system testing platform means the tools will face real-world interoperability challenges, not just lab benchmarks. If any team succeeds, the regulatory framework that emerges will shape how autonomous clinical AI gets evaluated, deployed, and reimbursed across cardiology and beyond.
Your membership also unlocks:
