Since January this year, the U.S. Centers for Medicare & Medicaid Services (CMS) has been piloting the WISeR model — Wasteful and Inappropriate Service Reduction — in six states. It lets private tech companies use AI to handle prior-authorization reviews for a slice of traditional Medicare services. A lengthy Ars Technica report published September 25, drawing on contractor data, physician complaints and congressional questioning, pushed the pilot's denial rates and revenue-sharing arrangement into public view.
Who's Reviewing, and How's It Going
The pilot runs in New Jersey, Ohio, Oklahoma, Texas, Arizona and Washington, through 2031. Roughly a dozen services fall under the review, including neurostimulation, epidural steroid injections, cervical spinal fusion, skin substitutes and treatments for urinary incontinence.
The three contractors have fared differently. Virtix had processed 6,096 applications as of March 30, approving 2,863 and denying 3,233 — a denial rate of roughly 53%. It was placed on a corrective action plan in June for missing the required 72-hour turnaround; the plan closed on August 14, and the company now says its average processing time is 1.18 days. Innovaccer wasn't ready at launch and at one point auto-approved every application; it did not respond to the report's requests for comment. Zyter spent months with mismatched data between Medicare Part A and Part B claims, though CEO Sundar Subramanian says both are now “fully functional.”
Turnaround times have been the sharpest complaint. The report cites one application that sat unresolved for 83 days, delaying a surgery by nearly two months. One Ohio practitioner wrote in feedback that they had watched three patients cry at bedside while waiting for prior authorization.
Pay Tied to Money “Saved”
The most contentious issue is how contractors get paid. Under CMS documents, a contractor keeps 25% of the spending “avoided” by a denial. That share is scaled by a quality score: 85 to 100 earns the full rate, 60 to 84 earns 95%, and anything below 60 still earns 90%. Even the lowest quality tier only shaves off a tenth of the payout, while denials translate directly into revenue.
At a hearing, Senator Patty Murray pressed Chris Klomp, nominee for HHS deputy secretary, on whether contractors have a financial incentive to deny claims. Klomp answered, “My understanding is no.” Murray then produced CMS documents she said showed contractors are incentivized to “deny as many claims as possible,” and said CMS plans to expand the pilot to cancer treatment — something she intends to fight. Representative Suzan DelBene separately accused the administration of withholding related documents.
Procedural questions have piled up too. The Government Accountability Office (GAO) determined in May that the Trump administration did not follow required procedure when setting up the program. CMS documents show the next wave of expansion could include cancer treatment, air ambulance transport, advanced imaging such as MRI, cardiac catheterization, pacemakers and implantable defibrillators, and genetic and molecular testing. Neither CMS nor HHS responded to requests for comment before the report's deadline. Virtix said it understands the delays affect patients but maintains its decisions follow CMS's “long-established” coverage standards rather than rules the company set on its own.
From Commercial Insurance to Traditional Medicare
AI-driven prior authorization isn't new in the U.S. — it has mainly shown up in privately run Medicare Advantage plans. In 2023, UnitedHealth and other insurers were sued over algorithm-driven denials of rehabilitation care, with plaintiffs arguing the tools' determinations were overturned at a high rate on appeal. Ars cites figures showing only about 11.5% of Medicare Advantage denials are appealed, but roughly 80% of those appeals succeed.
Put those two numbers together and the practical effect of prior-auth denials becomes clear: the overwhelming majority of patients who get denied never appeal. WISeR carries that same approach from commercial insurance into traditional, government-funded Medicare, testing it first on older patients in six states. CMS has not yet published data on how many of the 3,233 denied applications in the pilot were appealed, or how many of those appeals were overturned.
Sources: Ars Technica, CMS program documents, CocoLoop, U.S. Government Accountability Office; verified the pilot state list, Virtix's approval and denial counts, the 25% revenue share, and the quality-score tiers.