I read with interest HHS’s request for public input on how regulation, reimbursement, and research and development can accelerate the adoption of artificial intelligence in clinical care. As someone who began his career as a software engineer, then became a practicing physician, and later founded a clinical AI company, I welcome this moment. It is overdue, and it is also fragile.

The promise of efficiency over autonomous medicine

AI’s greatest opportunity in health care goes farther than diagnostics or autonomous medicine; it is efficiency. If deployed correctly, AI can return time, attention, and cognitive bandwidth to clinicians, resources that have been steadily drained by documentation, compliance, onboarding, and quality reporting requirements.

But realizing this promise requires clarity about one core principle. Technology is a tool, not the goal.

Clinicians currently suffer from an overload of technological and technical solutions that disrupt clinical workflows rather than being seamlessly embedded within them. Clinicians do not need revolutions that force them to relearn how to practice medicine. They need evolution, tools that fit naturally into how care is actually delivered. The most effective AI will be largely invisible. Ambient documentation that reduces charting. Automated onboarding that shortens time to productivity. Quality interventions that surface the right guidance at the right moment, without adding clicks, alerts, or administrative burden.

Aligning policy with clinical realities

From a policy standpoint, this has implications across all three levers HHS has identified.

Regulation should focus less on abstract definitions of AI and more on real-world impact. Does a tool reduce clinician workload? Does it integrate into existing workflows without introducing risk or friction? Regulatory clarity that distinguishes between workflow-augmenting tools and decision-making systems would reduce uncertainty and encourage responsible adoption.

Reimbursement should recognize efficiency as a clinical good. Today, many AI tools that meaningfully reduce burnout and improve consistency struggle to find a reimbursement pathway because they do not map cleanly to existing billing codes. Aligning incentives so that systems are rewarded for reducing waste, variation, and administrative overhead would accelerate adoption far more effectively than grants or pilot programs alone that are typically bound by a limited timeframe.

Research, development, and digital health collaboration

Research and development policy may be where HHS can have the greatest leverage. In particular, it is critical to make it easier, not harder, for digital health companies to work with health systems. Hospital IT teams are talented and mission-driven, but they are not structured to deeply specialize in narrow problems. Digital health companies are. They focus all of their effort, expertise, and attention on solving one problem exceptionally well.

Yet too often, startups face years-long procurement cycles, opaque data access rules, and incentives that favor homegrown solutions over best-in-class external tools. Facilitating standardized, secure pathways for collaboration between health systems and digital health vendors would benefit both sides, and ultimately, patients.

HHS has rightly emphasized public trust, safety, and outcomes. The path to all three runs through usability, workflow alignment, and measurable efficiency gains. If AI is allowed to quietly do the work clinicians never trained to do, while letting them focus on the work only they can do, adoption will follow naturally.

Ido Zamberg is founder and chief medical officer of C8 Health and a research fellow at McGill University Health Centre. He is a trained medical professional specializing in anesthesia and internal medicine. In addition to his clinical training, Dr. Zamberg is a computer scientist with a decade-long career as a developer at companies including HP and Autodesk. His research and academic publications are available on ResearchGate.