The Advanced Research Projects Agency for Health (ARPA-H) is soliciting proposals to develop two agentic artificial intelligence assistants for clinical care. It hopes to set a novel regulatory precedent with the Food and Drug Administration (FDA) for the use of generative AI in high-risk settings.
To date, only predictive AI technologies have been approved by the agency. President Donald Trump’s ARPA-H is taking a big swing at creating and deploying generative AI to extend the healthcare workforce.
The Agentic AI-Enabled Cardiovascular Care Transformation (ADVOCATE) aims to develop the first FDA-authorized agentic AI technology that can provide 24/7 specialty care. The program will select teams of innovators to create a patient-facing AI agent for cardiovascular disease, which results in 200,000 deaths per year, though treatments are inexpensive and widely available.
The research agency hopes the agents can supplement clinicians by scheduling appointments, adjusting medications and supporting diet and exercise. It will connect to the electronic health record, integrate with wearables and be available to patients 24/7.
The agent should be able to perform clinical tasks that a cardiologist could do over the phone, such as determining whether a patient is in heart failure or increasing a medication, Haider Warraich, M.D., program manager at ARPA-H, said.
A cardiologist himself, Warraich knows the impacts of heart disease well. Having worked in government at the FDA, he also understands the time and expense of setting regulatory precedent. Warraich wanted to couch the initiative in a high-risk and well-defined context like CVD.
While low-risk medical devices often operate in a regulatory grey zone, ARPA-H aimed to create a device that would sit clearly within the FDA’s authority to regulate, so that the agencies can work together to create a precedent.
The FDA and ARPA-H can work together to determine what data regulators need from generative AI and how to oversee it in an ongoing fashion. Under statute, ARPA-H can reimburse the FDA to accelerate the development of health technologies.
“We are focused on heart failure,” Warraich said. “Part of it is that the idea is that you need an intended use for your technology. The FDA cannot authorize an AI doctor, but they can authorize a technology that improves outcomes for patients with heart failure.”
The agents for cardiovascular disease will be co-developed with health systems and, after two years, will be deployed locally to test their impact on the system and gather data. The whole project, including FDA approval, is slated to last 39 months—just over three years.
Warraich wanted to move fast but also gather the necessary data for regulation. He also doesn’t think anything currently exists on the market that meets the ADVOCATE vision.
“I don't make regulatory policy,” Warraich said. “I don't make any payment models. But what we can do is basically us[e] novel technology development as a forcing function.”
The more difficult piece of the project, Warraich said, will be developing the supervisory agent, the second of the two systems the ADVOCATE program will fund.
The supervisory agent will attempt to solve a gaping issue in the world of healthcare AI, which is the lack of ongoing monitoring for continuously learning AI systems. It will be disease-agnostic so that it can eventually be used in a variety of contexts.
“Even though we haven't mandated it, it would be really nice for this also to be open source, as opposed to the cardiovascular disease agent technology, which is almost certainly going to be proprietary technology that's been designed for FDA authorization,” Warraich said.
ARPA-H will select teams of innovators within six months, in June 2026. After a year of innovating, the agency will eliminate some research groups and select the most promising technologies in a “down select” process.
“Agentic AI, which can autonomously plan and execute specific tasks, represents a massive opportunity for democratizing healthcare access,” ARPA-H Director Alicia Jackson, Ph.D., said in a statement. “ADVOCATE will allow patients to ask questions about their heart health and get reliable support and clinical interventions, helping make this AI-powered future real for CVD and other chronic diseases and paving the way for FDA-authorized agentic tools for many other conditions.”
The American Heart Association, which has its own AI evaluation center for cardiovascular solutions, said it is supportive of ADVOCATE.
“The American Heart Association is supportive of the ADVOCATE funding opportunity from ARPA and all funding opportunities that rigorously design and test different approaches for how AI can improve heart and stroke health for all individuals,” a spokesperson said in an emailed statement to Fierce Healthcare.
The email clarified that AHA does not view AI as a replacement for cardiologists or healthcare professionals.
John Whyte, CEO of the American Medical Association, said AI can help physicians reach more patients, and physicians support efforts to improve access to CVD treatment.
“At the same time, using autonomous AI to deliver cardiovascular care without clear physician oversight raises serious concerns about patient safety,” Whyte caveated in a statement to the media. “Heart care is complex, and mistakes in diagnosis or treatment can be life-threatening.”