Keith Magnuson, 83, used to walk six miles a day. Today, lumbar spinal stenosis has made it hard for him to walk 100 yards without severe pain. 

His physician recommended a minimally invasive outpatient procedure called MILD — a standard, Medicare-covered treatment for his condition. But thanks to an artificial intelligence prior-authorization program that launched in six states in January 2026, his care has been delayed for weeks.

Magnuson’s case, which Senator Maria Cantwell shared at an April 2026 Senate Finance Committee hearing, shows what is wrong with Medicare’s new AI system. The model, called Wasteful and Inappropriate Service Reduction — or, WISeR — admirably aims simultaneously to improve care for Medicare patients and to reduce costs for taxpayers. 

Yet the contractors running the algorithm’s reviews are paid a percentage of the savings they generate by averting care. A system in which the gatekeeper’s revenue grows when it says no is not a system that will distinguish well between surgeries that should not happen and surgeries that should.

WISeR has another problem, too. The algorithm reads claims data — imaging findings, diagnosis codes, procedural histories. It reviews the paperwork of medicine, not the practice of it. Whether a prior-authorization tool approves Magnuson’s procedure, it is deciding what is technically right rather than what is humanely right.

Most crucially, WISeR doesn’t prevent unnecessary surgery. For that, patients need careful clinical conversations that Medicare does not currently pay for.

AI Can’t Replace Conversations with Patients

I am a neurosurgeon. I have stood at the operating table for more than 3,000 spine surgeries, and I believe, along with much of the peer-reviewed literature, that surgical care in the United States is overtreatment-prone in ways that genuinely harm patients. A significant portion of the elective operations performed in this country are performed on patients who would do better without them.

WISeR is built to address this real and serious problem by requiring doctors to get permission from a payer before performing certain procedures. In principle, requesting permission is a reasonable check on overtreatment. In practice, it has become a cat-and-mouse game. 

Prior authorization lets through plenty of operations that should not happen, and it blocks or delays plenty that should. Artificial intelligence is not going to miraculously make the right call every time or magically make prior authorization a perfect system. Paying for a new algorithm to do what we already know is broken will only get us more of the same.

The financial gravity of medicine in the U.S. pulls every consult toward the operating room, and an AI gatekeeper at the back end does nothing to change the physics at the front. 

Consider another patient — a composite from my practice. Call her Margaret Kramer. Fifty-nine, sciatica for six months, an MRI read as supporting spinal fusion. She arrived in my office expecting to schedule the operation. Her imaging supported the surgery. Her diagnosis codes supported the surgery. An algorithm reviewing her file would have cleared her for fusion in minutes.

Technically, Kramer was a strong candidate for surgery. But given her active lifestyle, it wasn’t her first choice — and she had never been offered a real attempt at structured conservative care. I spent an hour with her, walked her through her MRI, and proposed three months of physical therapy first. Within 12 weeks she was pain-free, hiking again, back at work. She had not had surgery, and she would not need it.

Identifying the right treatment is not a paperwork problem. It is a payment problem. Surgeons are paid to operate. We are not paid, in any serious way, to spend an hour with a patient deciding she does not need an operation. The financial gravity of medicine in the U.S. pulls every consult toward the operating room, and an AI gatekeeper at the back end does nothing to change the physics at the front. 

The same dynamic exists across surgical specialties — in cardiac stents placed for stable coronary disease despite two decades of trials showing medication therapy works as well, in knee replacements where roughly one in three may be inappropriate. 

WISeR’s 13 active targeted procedures span these fields. Its failures will, too.

Lawmakers Can Repeal, Replace, or Reform WISeR

Congress is taking notice. Last month, the Government Accountability Office determined that the Centers for Medicare and Medicaid Services (CMS) implemented WISeR improperly, bypassing the congressional review it should have undergone. 

Within days, Senators Ron Wyden, Maria Cantwell, Richard Blumenthal, and Kirsten Gillibrand introduced a resolution under the Congressional Review Act to repeal the program entirely. Twenty senators have signed on. The 60-day clock to force a floor vote is running now and will conclude on August 8, 2026.

The center I am calling for — pay clinicians to think, fund the conservative care pathway, end the perverse contractor incentives — is not the position either side is currently defending.

Although the senators leading the repeal effort are right that WISeR delays needed care for seniors, they are not yet engaging with the harder question: What will they do to address the problem of overtreatment? Any policy fix that simply rolls back oversight — without putting something better in its place — leaves patients like Kramer on the conveyor belt toward operations they don’t need. 

Repeal alone is not reform. The senators have correctly identified what to stop; they have not yet said what to start. The administration is right that some intervention is needed, but it has chosen a tool that cannot do the work. The center I am calling for — pay clinicians to think, fund the conservative care pathway, end the perverse contractor incentives — is not the position either side is currently defending.

To his credit, the secretary of Health and Human Services, Robert F. Kennedy Jr., has called the delays from WISeR “unacceptable.” If the program is repealed, what replaces it must be built on different principles. If it survives, it must be substantially reformed.

Either way, the path forward is clear. Senator Cantwell has already proposed requiring a knowledgeable board-certified physician to review any denial, with a written clinical rationale, within 24 to 72 hours depending on the urgency of the case. 

Congress and CMS should end the percentage-of-savings contractor payment model that makes denials a profit center. 

And fund what actually reduces overtreatment: integrated specialty programs that pay clinicians for the hour-long conversations that surface patients like Kramer and protect patients like Magnusen. Bundled payments, shared-savings models, and Center for Medicare and Medicaid Innovation demonstrations have shown this can be done. WISeR is not it.

The most consequential decision in surgical care is not how to operate. It is whether to. That decision belongs in a consulting room — not in an algorithm trained on claims data. 

And not, certainly, in the hands of a contractor paid to say no.