Introduction: AI Moves Beyond the Supporting Role
For years, artificial intelligence in healthcare has played a supporting role. It flagged suspicious scans, drafted clinical notes, and summarized patient charts. Physicians remained firmly in the loop. Today, however, federal health officials are reaching for something far more ambitious — autonomous AI agents that diagnose patients, adjust medications, and act without waiting for a clinician to approve each step.
This shift marks a turning point. Rather than assisting doctors, agentic AI is designed to replace certain clinical touchpoints entirely. Moreover, the push is coming not from a private tech startup but from the U.S. federal government itself.
One HHS Strategy: A Coordinated Federal Push
On June 25, leaders from across the Department of Health and Human Services gathered online to present findings from a December request for information on accelerating AI adoption in clinical care. The event drew more than 7,000 public comments — a signal of the industry’s deep interest and concern.
The gathering reflected HHS’s coordinated approach, internally called “one HHS.” This strategy builds on an AI roadmap the department released late last year. Together, these efforts aim to align agencies ranging from the FDA to the Office of the National Coordinator for Health IT around a shared vision for clinical AI.
Thomas Keane, MD, national coordinator for health IT, opened the event by highlighting how AI already helps physicians detect strokes on imaging in seconds — work that was effectively impossible two decades ago.
“Technology has really transformed clinical care, and a big part of that technology is AI,” Dr. Keane said.
Yet the tools he described still assist a physician. The most striking vision of fully autonomous AI came from a different corner of the federal government entirely.
ADVOCATE: The ARPA-H Autonomous Cardiology Program
What Is ARPA-H?
The Advanced Research Projects Agency for Health (ARPA-H) is a federal agency modeled on DARPA, designed to fund high-risk, high-reward health research. Cardiologist Haider Warraich, MD, serves as a program manager there and is overseeing one of the agency’s most ambitious bets.
Introducing the ADVOCATE Program
Dr. Warraich leads a program called ADVOCATE — short for Agentic AI-Enabled Cardiovascular Care Transformation. Its goal is to develop the first FDA-authorized agentic AI system capable of delivering continuous, around-the-clock cardiovascular care. Currently, the program focuses on patients with heart failure and those recovering from heart attacks.
The rationale is stark: nearly half of all U.S. counties have no cardiologist at all. For millions of patients, the choice is often between waiting weeks for a specialist or going without care. ADVOCATE aims to fill that gap with software that never sleeps.
What the AI Agent Actually Does
Real-Time Integration With Patient Data
The ADVOCATE AI would integrate directly with the patient’s electronic health record (EHR) and data streams from wearable devices. Crucially, it would do so in real time. The system would not batch-process data overnight — it would monitor patients continuously and respond to changes as they happen.
A Wide Clinical Scope
The agent’s clinical scope goes well beyond scheduling appointments or sending reminders. According to Dr. Warraich, the technology would also:
- Support diagnoses based on EHR data and wearable inputs
- Triage patients based on symptom severity and risk
- Prescribe or adjust medications without referring every decision back to a physician
- Offer diet, lifestyle, and preventive care guidance
Dr. Warraich summed up the ambition plainly: the system should be able to “do everything a clinician can over the phone.”
Closing the Loop Without a Clinician
A New Model of Care Delivery
Traditional AI tools in healthcare operate in an open loop — they surface a recommendation, and a human decides what to do next. ADVOCATE is designed to close that loop entirely.
“This technology will be able to close the loop, instead of directing them to a clinician,” Dr. Warraich said. “It’ll be able to take the actions that the patient needs.”
A Template for Chronic Disease Care
Although cardiology is the initial focus, Dr. Warraich frames ADVOCATE as a blueprint for much broader use. “We are building a template that would be applicable to any patient with chronic disease,” he said.
Furthermore, he described the work in expansive terms — building “the entire stack that will be needed to make clinical agentic AI a reality for every American.” This positions ADVOCATE not just as a cardiology tool but as foundational infrastructure for the next era of healthcare delivery.
A Supervisory Agent Layer
The program funds a second, supervisory AI agent designed to monitor the primary clinical AI after deployment. This oversight layer would flag situations where human intervention is needed — adding a safety net to the autonomous system. Consequently, the model envisions AI watching AI, with humans stepping in only when the supervisory agent raises a concern.
Regulatory Guardrails: FDA Prepares for Agentic AI
Risk-Proportionate Oversight
The FDA is preparing regulatory frameworks designed specifically for autonomous clinical AI. Rick Abramson, MD, director of the FDA’s Digital Health Center of Excellence, said the agency is preparing for tools that operate with “greater autonomy and greater agency.” The goal, he said, is “regulation that is proportionate to risk.”
Beyond Approval: Post-Market Surveillance
Dr. Abramson stressed that FDA oversight cannot stop at the point of approval. Clinical AI tools, he noted, can “behave very differently in the post-market setting from how they behaved pre-market.” Therefore, the FDA is developing ongoing monitoring frameworks to track AI performance after it reaches patients.
He added that the agency plans to “release some ideas to the public for stakeholder comment in very short order,” though he declined to share specific policy directions still under development.
What Health Systems Need to Know Now
Accountability Remains Unresolved
ADVOCATE is a research and development program, not a product available for clinical use today. ARPA-H has outlined a multiyear process with early winnowing rounds before any technology enters clinical trials. Still, the signal to health systems is clear: autonomous clinical AI is coming, and Washington intends to shape the regulatory landscape before it arrives.
The harder questions begin after deployment. Chief among them is accountability: when an autonomous AI agent makes a wrong call, who is responsible? Some health systems are already wrestling with this question in the context of existing AI tools. Agentic AI raises the stakes considerably.
“Clinicians will have a lot of concern about technology that is so powerful,” Dr. Warraich acknowledged.
Building Trust Between AI and Clinicians
The agencies involved in the June 25 event openly acknowledged that they are still working through the trust question together. Specifically, the challenge is how to let an AI act on a patient’s care without losing the confidence of the clinicians it is designed to extend.
This balance — between autonomy and accountability — will define how agentic AI enters U.S. healthcare. Federal leaders are building the regulatory scaffolding and the technology at the same time. For health systems, the message is clear: the era of AI that acts on its own is no longer a distant hypothesis. It is a federal program with a timeline, a budget, and a blueprint.
