CMS Just Said the Quiet Part Out Loud. The Current System Will Not Work.

Three days ago in Las Vegas, the administrator of the Centers for Medicare and Medicaid Services stood in front of 24,000 healthcare leaders at the HIMSS Global Health Conference and said something that no one running a $1.7 trillion federal program is supposed to say publicly.

The current system will not work.

Not "the current system needs improvement." Not "we face challenges." Dr. Mehmet Oz, the CMS administrator, used the phrase multiple times during his keynote. He followed it by laying out a vision that would have been dismissed as science fiction two years ago: agentic AI in the hands of every Medicare beneficiary in the United States before the end of this administration. (Fierce Healthcare)

That is not a research initiative. That is not a pilot program. That is the largest payer in the world saying it intends to put autonomous AI tools directly into the hands of 67 million Americans to help them choose doctors, select Medicare Advantage plans, manage prescriptions, and navigate a system that even its own administrator admits is broken. (Healthcare Finance News)

And the timing of what happened around that stage in Las Vegas last week is what makes this more than a policy announcement. It is a signal that the healthcare industry just crossed a line it cannot walk back from.

The Week Everything Converged

On March 10, two days before Oz took the stage, Open Evidence recorded one million clinical consultations between NPI verified physicians and its AI system in a single 24 hour period. One million. In one day. The platform is now used daily by more than 40 percent of practicing physicians in the United States. It is backed by Google, Nvidia, and Sequoia at a $3.5 billion valuation and has raised more than $300 million. The New England Journal of Medicine called it the best AI tool for medical information. More than 100 million Americans this year will be treated by a doctor who used OpenEvidence to inform their clinical decision. (PR Newswire)

Let that sink in. The doctors have already adopted. They did not wait for CMS. They did not wait for the FDA, which has now approved more than 1,000 AI products in healthcare. They did not wait for hospital IT departments to finish their governance discussions. They went to Open Evidence because the volume of medical research now doubles every five years and no human can keep pace with it manually. (Chief Healthcare Executive)

Isaac Kohane, chair of biomedical informatics at Harvard Medical School, captured the tension precisely during a panel at HIMSS. He said the industry is going both too slow and too fast at the same time. Health systems are having substantive discussions about governance and safety, which has slowed adoption outside of revenue cycle management and ambient documentation. But that slowness has pushed physicians to adopt tools on their own, outside institutional oversight. Disruptive elements are going in without evaluations and without oversight, he warned.

This is the part the industry needs to sit with. The physicians are already using AI at scale. CMS is now pushing to put it in the hands of patients. And the institutional layer in between, the hospitals and health systems and payers, is still debating governance frameworks.

The $50 Billion That Cannot Solve the Problem

Oz was not speaking in abstractions. He pointed to the Rural Health Transformation Program, a $50 billion federal investment over five years to support rural hospitals. Then he said the money alone cannot fix what is structurally broken. Rural hospitals cannot recruit behavioral health professionals to rural areas regardless of funding. The physician shortfall is projected to reach nearly 100,000 by 2030. You cannot buy practitioners who do not want to live where the patients are.

This is where the agentic AI argument moves from aspirational to practical. If you cannot staff a behavioral health clinic in rural Mississippi, but you can deploy an AI agent that reminds a diabetic patient in that same community to pick up insulin, suggests meal plans, and flags when a biometric reading requires a telehealth visit, you have not replaced the clinician. You have extended the reach of the system into a geography it physically cannot serve today.

Oz drew the comparison explicitly. Banks are doing this, he said. Telecom companies are doing it. You can buy a mortgage with agentic AI giving you advice. You should be able to use the same technology to help you pick which Medicare Advantage plan to use or which doctor to go to. (Chief Healthcare Executive)

The comparison to financial services is the quiet part of the argument. Mortgage lending AI is regulated, tested, and deployed at scale. It handles high stakes decisions involving hundreds of thousands of dollars. The regulatory infrastructure exists. The compliance frameworks exist. If the financial services industry can trust AI to advise on a 30 year mortgage, the argument that healthcare cannot trust AI to help a 72-year-old choose between two Medicare Advantage plans starts to feel less like caution and more like institutional inertia.

The Trust Problem Is Real but Misdiagnosed

Here is where the opinion diverges from the optimism on the stage.

Only 31 percent of Medicare enrollees aged 65 and older trust AI a great deal or a fair amount to access their medical records and provide personalized health advice, according to a KFF survey. Oz acknowledged this directly, saying no one has gotten to them with the use case of why it will transform their life for the better. (Healthcare Dive)

He is right about the diagnosis, but the prescription deserves scrutiny. The trust deficit among seniors is not irrational. It is informed. These are the same Americans who have watched their insurance premiums rise, their provider networks shrink, 2.6 million Medicare Advantage enrollees lose their plans at the end of 2025, and their healthcare system become more confusing with every passing year. Telling them that AI will fix it requires more than a keynote. It requires the AI to actually work in their hands, on their terms, without requiring them to become technologically fluent overnight.

The DOGE advisor Amy Gleason described a concept where patients scan a QR code and bring their health data to their provider. That is a perfectly reasonable technical workflow for a 35 year old who manages their entire life from a smartphone. It is a completely different proposition for an 80-year-old in rural Arkansas who still gets their Medicare statements by mail.

The gap between what is technically possible and what is operationally deployable across 67 million Medicare beneficiaries is not a software problem. It is a design problem. And it is the kind of design problem that Silicon Valley has historically struggled with because the user base does not look like the engineering team.

What the HIMSS Formula Gets Right

Hal Wolf, the HIMSS president and CEO, offered a formula during the conference that distills the entire tension into a single line. New technology plus old organization equals costly old organization. (Chief Healthcare Executive)

That formula should be printed and taped to the wall of every health system CIO's office in the country.

The healthcare industry has a pattern of adopting technology and then wrapping it in the same bureaucratic processes that the technology was supposed to eliminate. Electronic health records were supposed to reduce administrative burden. They increased it. Patient portals were supposed to improve engagement. Most patients use them to check lab results and nothing else. Telehealth was supposed to transform access. It did, briefly, during a pandemic, and then the reimbursement structures pulled it back.

The risk with agentic AI for Medicare is that the same pattern repeats. The technology arrives. The institutions layer their existing processes on top of it. The complexity increases. The cost increases. And the patient, the person this was supposed to help, ends up with one more system to navigate inside a system they already do not understand.

That outcome is not inevitable. But avoiding it requires something the healthcare industry has rarely demonstrated: the willingness to redesign the organization around the technology rather than redesigning the technology to fit the organization.

What Actually Changed Last Week

Three things happened in the span of five days that will not reverse.

First, the administrator of CMS publicly committed to deploying agentic AI to Medicare beneficiaries within this administration's timeline. That commitment will be tested, debated, and potentially delayed, but the policy direction is now on the record.

Second, Open Evidence crossed one million physician consultations in a single day, confirming that the clinical workforce has already adopted AI as a decision support tool at scale. The adoption happened from the bottom up, not from the top down. That pattern rarely reverses.

Third, Kohane's warning at Harvard's HIMSS panel articulated the governing paradox of this moment. The industry is simultaneously moving too slowly on institutional adoption and too quickly on unregulated individual adoption. The gap between those two speeds is where the risk lives.

The healthcare industry did not choose this timeline. The convergence of a $1.7 trillion payer endorsing agentic AI, a physician facing AI platform reaching escape velocity, and a regulatory body that has already approved more than 1,000 AI products did not happen because anyone planned it. It happened because the economics of healthcare delivery have become unsustainable, the workforce shortage has become structural, and the technology has become capable enough to matter.

The question is no longer whether AI will reshape healthcare. The question is whether the institutions between the technology and the patient will adapt fast enough to make the transition productive rather than chaotic.

Based on the last 30 years of healthcare IT, the honest answer is that it could go either way. And the difference will not be determined by the technology. It will be determined by whether the organizations deploying it have the discipline to redesign themselves in the process.

New technology plus old organization equals costly old organization.

The formula has been stated. The test starts now.

Neural Press Network covers the systems, platforms, and policy shifts reshaping how industries operate. Subscribe at neuralpressbiz.com to receive the next edition directly.