AI Just Walked Into Medicare's Prior Auth Office
For the first time, AI is screening prior authorization in traditional Medicare. The goal is cutting real waste. The catch is who gets paid, and for what.
Photo by Mikhail Nilov on Pexels
The WISeR (Wasteful and Inappropriate Service Reduction) model brings AI-assisted prior authorization to traditional Medicare for the first time, starting January 2026 in six states for three service types prone to overuse. Technology vendors screen requests, but a licensed clinician must review before any denial. The design's core tension is that vendors are paid partly on the savings from denied care, which is exactly the incentive that got Medicare Advantage AI denials investigated.
For the first time in Medicare's history, when an older American on traditional Medicare needs certain procedures, an artificial intelligence is going to weigh in on whether they get it. That is the WISeR model, a new CMS demonstration that puts AI-assisted prior authorization into original Medicare, starting in 2026. Prior authorization, the permission-slip dance that Medicare Advantage patients have endured for years, is coming to the other side of the program, and this time with algorithms in the room. I have a lot of feelings about this, and I think you should too.
What WISeR actually is
Let me give the model its due before I get nervous about it, because the fair version matters. WISeR is bounded and it is targeted. It is voluntary, it runs in a handful of states, and it aims at three specific service categories that genuinely are overused. This is not AI deciding whether grandma gets her heart surgery. It is narrow by design.
And the waste is real. Skin and tissue substitutes, one of the three targets, have been a poster child for runaway low-value spending for years. Reducing care that does not help patients is a legitimate goal, and I am not going to pretend otherwise. If WISeR were just smarter clinicians catching genuine waste, I would cheer. The trouble is in how it pays the people doing the catching.
Follow the money, because it is the whole story
Here is the decision path, and I want you to watch where the money enters, because that is the tell.
Here is the part that made me put my coffee down. As KFF lays out, the technology companies running these reviews are paid, in large part, based on a share of the savings from the care they deny. Read that twice. You are paying a company a cut of every no. We have seen this movie before, on the Medicare Advantage side, where AI tools were investigated for driving up denial rates and running with too little human oversight. Now the same basic incentive is being piloted inside traditional Medicare, the program people trusted precisely because it did not do this.
The safeguards, and what they leave on the table
To be fair, CMS did not walk in blind. It built real guardrails. But every guardrail here comes with a gap, and it is worth seeing them side by side rather than taking the reassurance at face value.
The whole thing rests on one load-bearing beam: human review before a denial. If that review is a physician carefully weighing each case, WISeR could genuinely trim waste while protecting patients. If it becomes a click-to-approve-the-denial button pressed under quota, then the human in the loop is decorative and the incentive wins. The difference between those two futures is not in the model card. It is in the enforcement, the audit teeth, and whether anyone is measuring how often the human actually overturns the machine.
If you build in patient access, WISeR is the most important experiment to watch this year, because it is a live test of whether AI utilization management can be done to patients well instead of just cheaply. I am rooting for the careful version. But I would watch one number above all others: how often the human reviewer disagrees with the AI. If that number is healthy, the oversight is real. If it rounds to zero, we will know the stopwatch won, and we should say so loudly while it is still a pilot and not yet the whole program.
- WISeR brings AI-assisted prior authorization to traditional Medicare for the first time, in six states from 2026.
- It targets three overused service types: skin substitutes, nerve stimulator implants, and knee arthroscopy for osteoarthritis.
- A licensed clinician must review before any AI-driven denial, and CMS added audits, penalties, and gold-carding.
- The core tension: vendors are paid partly on the savings from denied care, an incentive to deny.
- The waste is real, but so is the risk. Watch whether human review is genuine oversight or a rubber stamp under volume.
Frequently asked
What is the WISeR model?
A CMS Innovation Center model, running 2026 to 2031, that uses AI and machine learning plus clinician review to screen prior authorization for select, often-overused services in traditional Medicare. It is the first such model where technology companies are the only participants.
Which services and states does it cover?
Three service categories prone to overuse: skin and tissue substitutes, electrical nerve stimulator implants, and knee arthroscopy for osteoarthritis. It runs in six states: New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington.
Does AI get to deny care on its own?
No. CMS requires a licensed clinician to review before any denial based on the AI screen. The AI flags; a human is supposed to make the call.
So what is the concern?
Vendors are compensated partly based on the savings from services they deny. Critics note this mirrors the incentive behind Medicare Advantage AI-denial investigations, where algorithms raised denial rates. The safeguard is human review, and its quality is everything.
Is the underlying waste real?
Yes. Skin substitutes in particular have well-documented overuse, and reducing low-value care is a legitimate goal. The debate is about the method and the incentives, not whether waste exists.
Sources
Naveen Kumar
Healthcare engineering and product executive in Pittsburgh. 15+ years building AI-first patient access, a decade at Treatspace.