
Medicare is now using AI on prior auth. WISeR is the tell.
Since January 1, Original Medicare has been running an AI-assisted prior-authorization pilot in six states. In July, the Senate voted to keep it. The direction of travel is not subtle.
What WISeR actually is
The Wasteful and Inappropriate Service Reduction (WISeR) model launched January 1, 2026 in New Jersey, Ohio, Oklahoma, Texas, Arizona and Washington, and runs through 2031. It applies prior authorization to 17 outpatient services CMS considers vulnerable to fraud, waste and abuse — in Original Medicare, which historically had almost none. The mechanism: AI screens the request; a licensed clinician must review any denial. That human-in-the-loop structure is the template every serious deployment converges on, and it is the same one we hold in our own healthcare builds.
The Senate just kept it alive
On July 16, the Senate rejected an effort to overturn WISeR. Whatever your politics on AI-assisted utilization review, the operational fact is this: the payer side of American healthcare is arming itself with AI at federal scale, with Congressional cover. Commercial payers were already ahead of CMS. The review clock on the other side of your claims is getting faster and cheaper every quarter.
The new asymmetry
Here is what that means in practice. A payer running AI screening can evaluate a request in minutes against published criteria. A provider assembling that request by hand — chart notes, failed-therapy history, medical-necessity evidence, portal by portal — takes days. When one side of a negotiation automates and the other does not, the manual side eats the delays, the denials, and the write-offs. Denials with named criteria are answerable; but only if someone has the hours to answer them, and most revenue-cycle teams are already underwater.
What to build before it expands
Symmetry. An agent on the provider side that assembles the evidence package against the payer’s own published criteria, cites the source page for every claim, drafts the appeal when the denial names its reason, and logs everything for audit. That is not speculative — it is the exact workflow shape we build in our Production Agent Sprint: running alongside your team in 21 days, evaluated against your own historical cases before it touches anything live. Six states today. Check the expansion trajectory and decide whether you want to build the muscle now or during the fire.
Put an Applied AI Scientist on your problem
Thirty minutes, no deck. Bring the workflow that hurts and leave with an honest read on whether AI can carry it — and what it would take to prove it in your environment.
Book a call with an Applied AI Scientist Explore Intelligent AI World