Prior authorization — the requirement that a payer affirm coverage before a service is delivered — has long been a Medicare Advantage practice. On 1 January 2026 it arrived in traditional Medicare, carried by artificial intelligence. The Wasteful and Inappropriate Service Reduction (WISeR) Model is a six-year CMS Innovation Center test in which six technology companies review selected services in six US states, paid from the spending their reviews avert. It has already survived a Senate repeal vote. This page explains the model from its primary documents — the Federal Register notice and the CMS model page — and maps the questions it raises for health systems everywhere. As of August 2026.
What is the WISeR Model?
WISeR is a payment-model test run by the Centers for Medicare & Medicaid Services (CMS) under Section 1115A of the Social Security Act, announced in a Federal Register notice on 1 July 2025 and effective 1 January 2026 for six years, structured as two three-year agreement periods 2. Its stated target is spending "that could be reduced or eliminated without adversely affecting quality of care or health outcomes" — waste CMS estimates at roughly a quarter of total US health spending 2.
The design is unusual in two ways. First, the participants are technology companies rather than providers or payers — CMS calls WISeR "the first Innovation Center model in which technology innovators are the only model participants" 1. Second, the review they run is technology-assisted: the same AI-enabled prior authorization machinery that commercial payers already use, now applied inside original Medicare with CMS-defined guardrails. For the broader taxonomy of these systems, see what agentic AI in healthcare means and the agentic AI definition in our glossary.
Which services and states does it cover?
The model runs in four Medicare Administrative Contractor (MAC) jurisdictions — a MAC is the regional contractor that processes Medicare claims — covering New Jersey, Ohio, Oklahoma, Texas, Arizona and Washington 2. Each of six participants covers one state: Cohere Health (Texas), Genzeon (New Jersey), Humata Health (Oklahoma), Innovaccer (Ohio), Virtix Health (Washington) and Zyter (Arizona) 1.
The service list is deliberately narrow: seventeen categories tied to existing national or local coverage determinations, selected for "existing evidence of potential FWA" — fraud, waste and abuse 2. It includes skin and tissue substitutes, electrical nerve stimulators (vagus, phrenic, sacral, hypoglossal and deep brain stimulation), epidural steroid injections for pain, cervical fusion, percutaneous vertebral augmentation, arthroscopic lavage and debridement for the osteoarthritic knee, incontinence control devices, and impotence diagnosis and treatment 2. Inpatient-only services, emergency services, and anything whose delay would pose substantial patient risk are excluded 1. KFF puts the targeted spending at services CMS associates with between $1.9 billion and $5.8 billion in low-value care in 2022 34.
How does the review actually work?
The notice describes three pathways 2. A provider may submit a prior authorization request directly to the model participant, or to the MAC which routes it onward. Submission is voluntary — but a claim filed without one is flagged for pre-payment medical review, so opting out means review anyway, just later. Requests meeting all coverage, coding and documentation rules receive a "provisional affirmation," and the claim is then paid in full when billed correctly.
The guardrails matter as much as the pipeline:
- A licensed clinician behind every non-affirmation. "All recommendations for non-payment are determined by appropriately licensed clinicians" applying standardized, transparent, evidence-based procedures 1. The technology can accelerate affirmations; a human must own the denials — the same human-in-the-loop principle we treat as the floor for clinical agent oversight.
- Unlimited resubmission and peer-to-peer review. A non-affirmed request can be resubmitted without limit, with the option of a peer-to-peer discussion to inform the new determination 2.
- Appeal rights preserved. A denied claim follows the existing Medicare appeals process, unchanged 2.
- Coverage itself untouched. WISeR changes when review happens, and by whom — the underlying coverage, coding and payment rules are the same ones a MAC would apply 1.
CMS is also exploring "gold carding": exempting providers who sustain a provisional-affirmation rate of 90 percent or higher from prior authorization altogether, focusing scrutiny on outliers 2.
Why is the payment design controversial?
Participants are paid "a share of averted expenditures" — a percentage of the spending their reviews prevent, adjusted by performance measures that include provider experience 12. Critics read that as a pay-for-denial structure: the more care goes unapproved, the more the reviewer earns 3. The counterweights CMS points to are the clinician requirement, quality-adjusted payments, audits and monitoring of downstream quality 2 — but whether those checks hold under a denial-linked revenue model is the empirical question the six states are now testing. The Medicare Advantage record is why skeptics start uneasy: nearly 50 million prior authorization requests went to MA insurers in 2023, and among the small fraction of denials that were appealed, roughly 82 percent were overturned 34 — a pattern consistent with substantial over-denial. Bias is a live concern too; a model trained on historical claims can inherit historical inequity, which is why subgroup performance audits belong in any evaluation of the model's outputs.
Where does the political fight stand?
Opposition arrived early. Members of Congress wrote to CMS urging a halt before launch, citing the for-profit participation and Medicare Advantage's record 4. In May 2026, Senate Democrats led by Senator Ron Wyden introduced a Congressional Review Act resolution to overturn the model's authorizing rule 6. The decisive vote came on 16 July 2026: the Senate rejected the motion to proceed on S.J.Res.198 by 46 votes to 50 5. WISeR stands — and with the CMS model page listing it as active with six participants 1, it is now the clearest federal signal of AI's place in Medicare utilization management. We log developments like this in our global AI-in-health regulation tracker, alongside the state laws now regulating AI in payer utilization review.
What should leaders take from WISeR?
Three things, whether or not you operate in the six states.
First, the review layer is becoming algorithmic across payers, and the providers responding in kind — with revenue-cycle and coding agents and EHR-integrated agents that assemble documentation before submission — are treating prior authorization as a machine-to-machine exchange. Evaluate those tools with the same rigor as any clinical system: our guide to evaluating an agent before deployment applies almost verbatim.
Second, operational exposure is measurable now: affirmation rates, decision times and resubmission loops belong on the same dashboards as denials — a governance task your AI committee should own, with hospital operations downstream of it.
Third, accountability questions are unsettled. When an AI-assisted review delays care that a clinician later shows was necessary, the allocation of responsibility among participant, MAC and CMS runs into the same doctrine gaps we map in liability when clinical AI errs — and physician sentiment, already skeptical of payer AI in our survey tracker, will shape how loudly those cases surface. For the wider deployment context, see AI in healthcare statistics.
Sources and method
This page is built from the model's primary documents: the Federal Register notice of 1 July 2025 (90 FR 28749), which defines the service list, the review scenarios, the clinician requirement and the payment methodology 2, and the CMS Innovation Center model page, which lists the six participants, their states, and the model's active status 1. KFF's February 2026 analysis 3 and the Georgetown Center on Health Insurance Reforms' assessment 4 supplied the Medicare Advantage comparison figures and the critique of the incentive design. The legislative record — the May 2026 resolution and the 16 July 2026 Senate vote — is drawn from Healthcare Dive's reporting 6 and the official GovInfo bill-status data for S.J.Res.198 5. We revisit this page every ninety days and sooner when CMS publishes model data or Congress acts. Current as of 1 August 2026.