Federal health leaders are signaling that artificial intelligence is entering a new phase in U.S. healthcare. The question is no longer whether AI belongs in clinical care, but how quickly it can be deployed safely, governed responsibly and integrated into everyday workflows.

During a recent InterSystems webinar, AI-Ready Data for Learning Health Systems: From Interoperability to Data Utility — An Enterprise Strategy for Trusted AI and Data Utility, senior officials from across the U.S. Department of Health and Human Services (HHS) outlined what amounted to a coordinated federal roadmap for AI adoption. Their message was consistent across agencies: success will depend less on building new AI models than on creating the data infrastructure, governance frameworks, regulatory pathways, and implementation support needed to bring AI into routine care.

The discussion brought together leaders from the Office of the National Coordinator for Health Information Technology (ONC), the Food and Drug Administration (FDA), the Advanced Research Projects Agency for Health (ARPA-H) and the Administration for Community Living (ACL), offering one of the clearest pictures yet of how HHS intends to move AI from promise to practice.

AI Is Already Changing Clinical Care

Opening the session, Thomas Keane, M.D., MBA, National Coordinator for Health Information Technology, grounded the policy discussion in a real-world clinical example.

Keane described interpreting stroke CT scans just months ago in a Philadelphia hospital using AI-assisted imaging software that not only visualized cerebral blood vessels but also highlighted suspected areas of ischemia, calculated cerebral blood flow and blood volume, and directed clinicians toward subtle pathology that might otherwise have taken longer to identify.

The technology, he explained, enabled rapid identification of a tiny blood clot, activation of the stroke response team and timely intervention that restored the patient's neurological function.

For Keane, the story illustrated something much larger than a technological advance. AI is already changing how medicine is practiced by compressing the time between diagnosis and treatment and helping clinicians make faster, better-informed decisions.

He pointed to three areas where AI is already demonstrating value: clinical decision support, imaging and triage; administrative functions such as documentation and routine workflows; and patient-facing tools that help individuals better understand and manage their own health. Taken together, those applications represent what Keane described as a more accessible, affordable and patient-centered healthcare system.

Rising Costs Are Making AI Harder to Ignore

While the clinical potential of AI continues to expand, Mark Atalla, Pharm.D., MBA, Deputy National Coordinator for Health IT Policy at ONC, argued that the economic realities facing healthcare are making adoption increasingly urgent. "Our why is very simple," Atalla said. "The cost and quality of care."

He pointed to CMS Office of the Actuary projections showing U.S. healthcare spending reached approximately $5.3 trillion in 2024, while Medicare enrollment is expected to grow from 69 million beneficiaries in 2025 to 78 million by 2032. Medicare expenditures alone are projected to increase from roughly $1.2 trillion to $2.1 trillion during that period.

Against that backdrop, Atalla argued that healthcare cannot solve its workforce and cost challenges simply by adding more people. He cited diabetes as a prime example. More than 115 million Americans have prediabetes, roughly 40 million are living with Type 2 diabetes, and long-term outcomes continue to deteriorate despite decades of investment.

He highlighted research showing an AI-powered prediabetes coach performed comparably to human coaches, suggesting AI can extend clinical capacity without sacrificing quality. Rather than replacing clinicians, he said, AI offers an opportunity to augment today's care delivery model while bending the long-term cost curve.

HHS Is Aligning Around One AI Strategy

Perhaps the strongest message of the webinar came from Arman Sharma, Deputy Chief AI Officer at HHS, who described an unusually coordinated federal strategy designed to align regulation, reimbursement, scientific research and implementation under what he called a "One HHS" approach.

Sharma traced the administration's AI strategy through three priorities: accelerating innovation, building the infrastructure needed to support AI and strengthening U.S. leadership in AI security and diplomacy. But he argued that the conversation has moved beyond innovation alone.

"It's not enough just to build new AI tools," Sharma said. "They actually have to be adopted."

That shift in emphasis may prove to be one of the most significant developments for health systems. Federal agencies are increasingly focusing on the practical barriers that prevent AI from reaching clinicians and patients, including governance, implementation support, workflow integration and clearer evaluation standards.

Sharma also emphasized that none of those goals can be achieved without greater data liquidity. Making health information more accessible to patients and providers, he said, remains foundational to successful clinical AI.

ARPA-H Is Building the Foundation for Agentic AI

The federal vision is also beginning to take shape through research investments. Haider Warraich, M.D., cardiologist and program manager at the Advanced Research Projects Agency for Health (ARPA-H), described the agency's Advocate initiative as an example of how AI is moving beyond isolated applications toward continuous patient support.

The program is funding development of AI agents designed to help patients living with heart failure by coordinating appointments, providing lifestyle guidance, supporting medication management, monitoring patients through wearable devices and integrating directly with electronic health records.

The broader goal, Warraich explained, is to create continuous care rather than episodic care. Recognizing the safety concerns surrounding increasingly autonomous AI systems, Advocate also includes clinician oversight mechanisms and randomized clinical trials designed to evaluate both effectiveness and patient safety before widespread adoption.

Extending the Caregiving Workforce

AI's potential extends beyond hospitals and physician offices.

Mary Lazare, Principal Deputy Administrator at the Administration for Community Living (ACL), and Kelly Cronin, Deputy Administrator for Innovation and Partnership at ACL, described how AI could help address one of healthcare's fastest-growing workforce challenges: caregiving. With approximately 58 million Americans over age 60, 63 million family caregivers, and persistent shortages of direct care workers, demand continues to outpace available resources.

ACL's Caregiver AI Prize Challenge seeks technologies that reduce caregiver burden while helping older adults and people with disabilities remain safely in their homes. Rather than replacing human caregivers, the initiative is focused on extending their reach through practical, affordable AI tools that emphasize privacy, transparency, safety and human oversight. The agency also hopes the program will establish implementation best practices that organizations can adopt more broadly.

FDA Signals a More Adaptive Approach to AI Oversight

As AI capabilities continue to evolve, regulators are also preparing to evolve with them. Richard Abramson, M.D., MHCDS, FACR, Director of the FDA's Digital Health Center of Excellence, said stakeholder feedback has consistently called for greater clarity around how increasingly sophisticated AI tools will be regulated.

Rather than relying solely on traditional premarket review, FDA is moving toward a lifecycle approach that recognizes AI systems may change over time and behave differently after deployment. Abramson also emphasized the importance of risk-proportionate regulation, acknowledging that clinicians interact with AI very differently today than they did only a few years ago.

Equally important, he said, will be continued coordination across FDA, CMS, ONC, ARPA-H, ACL, professional societies, and international regulators to ensure a more consistent regulatory framework.

The Conversation Has Shifted

Taken together, the presentations revealed an HHS strategy that is considerably more coordinated than previous federal AI efforts.

ONC is focused on data interoperability and governance. FDA is reshaping regulatory oversight. ARPA-H is investing in next-generation clinical applications. ACL is exploring how AI can strengthen caregiving and community-based services. While each agency has a distinct mission, all are moving toward the same objective: making AI practical, trustworthy and scalable across the healthcare system.

For health system executives, the implications are significant. Federal leaders are no longer discussing AI primarily as an emerging technology. They are increasingly treating it as core healthcare infrastructure — something that will shape clinical operations, reimbursement models, workforce strategy and patient engagement for years to come.

As Atalla observed in his closing remarks, the trajectory of AI in healthcare may be inevitable. The pace of adoption, however, will depend on the decisions healthcare organizations make today.