A Medicare pilot program that uses AI to approve or deny prior-authorization requests denied 53% of the cases reviewed by one contractor in its first months of operation. WISeR — the Wasteful and Inappropriate Service Reduction model — launched in January 2026 across six states and covers roughly a dozen procedures, including nerve stimulation, epidural steroid injections, and cervical spinal fusions. Federal documents released by the Electronic Frontier Foundation show contractors are paid a share of the spending they avert, tying vendor revenue directly to denial volume.
The denial numbers come from a March 30 weekly report on Virtix, one of the vendors running WISeR reviews. Virtix reviewed 6,096 pre-authorization requests, approved 2,863, and denied 3,233. WISeR is designed to return decisions within 72 hours; the released documents show at least one request pending after 83 days, and one healthcare provider surveyed by the government said a patient's surgery was pushed back nearly two months due to zero communication from the vendor.
The Centers for Medicare and Medicaid Services placed Virtix on a Corrective Action Plan in June for noncompliance with the 72-hour window. Virtix told Ars Technica the CAP ended August 14 and said it has reduced its average turnaround to 1.18 days for prior authorization and 1.17 days for prepayment review determinations, within WISeR's three-day requirement. The company said its determinations rely on National and Local Coverage Determinations set by CMS, not by Virtix.
Key facts
- 01WISeR vendor Virtix denied 3,233 of 6,096 prior-authorization requests reviewed by March 30, a 53% denial rate.
- 02At least one request was still pending after 83 days, against a 72-hour target for WISeR decisions.
- 03CMS planning documents describe WISeR pay as a 'share of averted expenditures' — vendors are compensated based on denied care.
- 04The pilot launched in January 2026 across New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington, and is scheduled to run through the end of 2031.
- 05The Government Accountability Office determined in May that Trump officials did not follow proper procedure in setting up WISeR.
A second vendor, Innovaccer, was not ready for the January launch and asked the government to delay the rollout. When CMS declined, Innovaccer told federal officials it would auto-approve requests as a stopgap.
“[A]uto-affirming is the only path available that avoids creating a backlog of unprocessed prior authorizations and claims while we finalize, validate, and deploy the full rules-based solution”— Innovaccer, WISeR vendor, in a letter to government officials
A third vendor, Zyter, spent months mishandling data because it did not distinguish between Medicare Part A (inpatient) and Part B (outpatient) claims. Zyter CEO Sundar Subramanian told Ars Technica the platform is now 'fully functional across Medicare Part A and Part B claims' and that the company is working with federal partners to improve the provider experience. Innovaccer told Stat News its technology is 'fully live.'
Provider survey responses collected by the government describe patients waiting weeks with no way to reach a human. One Ohio provider, writing in all capital letters about patients seeking kyphoplasty for spinal compression fractures, said they had watched three patients cry at bedside waiting for prior-auth responses and called the experience 'a disgrace to the human race.' Another Ohio provider wrote that a three-to-four-day delay for pain procedures 'is already difficult for vulnerable patients, but when providers cannot obtain answers for weeks, the situation becomes unacceptable.'
The design of the payment model has drawn direct scrutiny. At a September 16 confirmation hearing, Senator Patty Murray pressed Chris Klomp, deputy CMS administrator and the administration's nominee for deputy HHS secretary, on whether contractors earn more by denying care.
“Do the contractors in the model—who are the private companies conducting the prior authorization assessments—make more money if they deny care? Just 'yes' or 'no,'”— Patty Murray, US Senator (D-Wash.)
Klomp answered, 'My understanding is no.' CMS's own planning documents, included in the tranche released by EFF, describe WISeR as a 'novel payment approach where the model participants are compensated based on a share of averted expenditures' — meaning vendor compensation scales with the dollar value of care they block.
Congressional efforts to force disclosure of more WISeR documents have stalled. Representative Suzan DelBene called a committee vote last week to compel additional records; Republicans voted it down.
In May, the Government Accountability Office determined that officials did not follow required procedure in setting up WISeR, raising a legal question that has not slowed the rollout. The pilot is scheduled to run through the end of 2031, and CMS has signaled plans to expand it.
Prior authorization has been a fixture of private Medicare Advantage plans for years, but traditional Medicare has largely operated without it. WISeR extends that gate to a population that has not encountered it before, using automated systems that vendors themselves have described as not fully deployed, while paying those vendors more when they say no.
AI-driven claims adjudication is one of the most commercially attractive AI use cases in healthcare precisely because denials convert directly to insurer margin. WISeR imports that incentive structure into traditional Medicare and layers it on top of vendors that, per their own filings, were not ready to run it. The lasting policy question is not whether AI can process a prior-auth request in under 72 hours — Virtix's revised numbers suggest it can — but whether a payment model that rewards averted expenditures can produce clinically defensible decisions at scale. The 53% denial rate, and the survey responses behind it, are the first data points on that question, and they will shape how far the model gets to expand before 2031.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
