The 72-Hour Promise

Medicare’s AI-assisted prior-authorization experiment was supposed to make selected reviews timely and appropriate. Newly released records show why the real endpoint is not a decision. It is whether the patient reaches care.

Presented By Our Partners

Prior authorization is built around a peculiar kind of suspense. A clinician has already decided that a patient needs something. The patient may already be in pain. Yet treatment pauses while another organization decides whether Medicare will pay.

Artificial intelligence is often introduced into this interval with a promise that sounds almost self-evident: machines can review paperwork faster than people. But speed in an authorization system is not the same thing as speed to care. A request can move through software, receive a decision and still leave a patient stranded.

That distinction now has a revealing federal test case.

In January, the Centers for Medicare & Medicaid Services launched the Wasteful and Inappropriate Service Reduction model, mercifully shortened to WISeR. The six-state experiment applies to selected services in Original Medicare that CMS considers vulnerable to waste, fraud or abuse. Private participants use technologies including artificial intelligence and machine learning to help evaluate prior-authorization requests. CMS says clinicians with appropriate expertise must validate determinations.

The program’s public language emphasizes timely payment, safer evidence-supported care and less administrative burden. Newly released internal records describe a rougher beginning: missed turnaround targets, technical and communication failures, thousands of delayed requests and provider reports of patients waiting in pain.

The gap between those two accounts is not merely a story about a troubled software launch. It is a lesson in what healthcare AI must be required to prove.

The clock that patients actually feel

CMS designed WISeR to run through 2031 in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. Providers can submit a prior-authorization request before furnishing one of the selected services or proceed without advance approval and face post-service, prepayment review. The initial service set includes such items as skin and tissue substitutes, electrical nerve-stimulator implants and knee arthroscopy for osteoarthritis.

The policy rationale deserves a fair hearing. Unnecessary procedures can expose patients to infection, anxiety and out-of-pocket costs. Fraud also drains a public program. A well-designed review system could protect both patients and taxpayers.

But approximately 1,000 pages obtained by the Electronic Frontier Foundation through a Freedom of Information Act lawsuit show that the operational question arrived before the policy promise was settled.

The records, which cover the early months of the program, show many requests exceeding a 72-hour target. Reporting by Fierce Healthcare found that several hundred requests remained unanswered at the end of March and that one was 83 days old. The documents also show large denial volumes. EFF reports that two companies denied more than 20,000 requests during the first three months and that one participant denied more requests than it approved. CMS required that participant to submit a corrective-action plan.

These are early rollout records, not a final evaluation of a six-year model. They do not establish that every denial was wrong, that AI alone made the decisions or that all participants performed alike. They do establish something narrower and important: a system advertised as timely can become another source of delay, and a decision count cannot tell us whether a patient received appropriate care.

Featured Partner

Invest in the Infrastructure Behind Modern Medicine

As healthcare expands beyond hospital walls, the buildings and campuses supporting that shift are generating compelling returns for investors who move early. The Healthcare Real Estate Fund offers qualified investors direct access to a curated portfolio of medical office, outpatient, and specialty care facilities.

Learn More →

“Human review” is a beginning, not a safeguard by itself

The WISeR model does not formally hand the final word to an algorithm. CMS says participating companies must employ clinicians with relevant expertise to conduct medical reviews and validate determinations.

That requirement is necessary. It is also incomplete.

The public record still leaves basic questions about the underlying systems unanswered: What data were used to train or configure them? How were accuracy and bias tested? How frequently do reviewers disagree with machine-generated recommendations? Can a reviewer see the source and logic behind a recommendation? Are staffing levels sufficient to examine a growing queue rather than approve it by momentum?

A human signature at the end of an automated process does not reveal the quality of the review. Real oversight requires time, authority, usable evidence and a measurable record of disagreement and correction.

The released provider feedback gives the abstraction a human scale. Clinicians described patients calling their offices in tears while pain procedures remained unresolved. One wrote that three patients awaiting decisions about vertebral-augmentation procedures had cried at the bedside because no answer had arrived. These accounts do not prove population-level harm. They do show why a missed administrative deadline cannot be treated as a harmless software metric.

For someone in severe pain, the clock is clinical.

The incentive belongs inside the audit

WISeR participants can share in savings associated with spending the model averts. CMS also applies quality measures and does not reward denials later reversed on appeal. The arrangement is intended to align private expertise with Medicare’s effort to reduce inappropriate spending.

It also creates a conflict that cannot be managed by aspiration. When revenue is linked to avoided expenditure, a denial is both a coverage decision and a financial event. That does not make every denial improper. It means the audit must follow much more than gross savings.

At minimum, the public should be able to see:

• approval and denial rates by participant, service and patient group;
• median and worst-case decision times, not only averages;
• how often clinicians overturn automated recommendations;
• appeal rates and reversal rates;
• how many patients ultimately receive the requested service;
• how many abandon care after delay or denial;
• complications, emergency visits or other measurable consequences during the wait;
• and whether outcomes differ by disability, race, language, geography or digital access.

Without those measures, the program can count dollars that did not leave Medicare while failing to distinguish waste prevented from care deferred.

A decision is not completed care

Healthcare AI is usually evaluated at the moment where the machine acts. A model flags an image. A chatbot recommends a destination. An authorization engine approves or denies a request. These endpoints are attractive because they are easy to count and close to the software.

Patients live farther downstream.

The meaningful WISeR denominator begins with every request and ends with the disposition of the patient: Was the requested service appropriate? Was the decision timely? If denied, was an alternative offered and completed? If appealed, was the decision reversed? Did delay change the patient’s condition? Did the process reduce unnecessary care without widening disparities?

This is the discovery-to-delivery gap in administrative form. The “signal” is the clinician’s request. The algorithmic action is the review. Delivery occurs only when the patient reaches appropriate care, whether that means receiving the service, choosing an evidence-supported alternative or safely avoiding an unnecessary intervention.

An approval that arrives after the useful clinical window is not success. Neither is a fast denial that sends a patient into an opaque appeal. The endpoint is not throughput. It is resolution.

What patients and clinicians can ask now

For people affected by an AI-assisted authorization program, the most useful questions are practical:

1. What is the expected response time, and who owns escalation when it is missed?
2. Was the decision reviewed by a clinician with relevant expertise?
3. What coverage criterion was applied, and what documentation is missing?
4. How can the decision be appealed, and does an urgent-review route exist?
5. If the requested service is not approved, what covered alternative is available?
6. Who tracks the case until the patient receives appropriate care?

Clinicians and health systems should ask one more: Can we see our own request-to-completion data, rather than a dashboard that ends at authorization?

WISeR is still an experiment, and the records released so far are a partial view of its earliest months. CMS and the participating companies may improve performance. More documents and fuller outcome data could change the picture.

That is precisely why the standard should be fixed now. An AI system that influences access to Medicare care should not be judged by how modern its technology sounds, how many requests it processes or even how quickly it produces an answer. It should be judged by whether appropriate care becomes more reliable, more equitable and genuinely complete.

The 72-hour promise matters. What happens to the patient after the clock stops matters more.

Sources

• Centers for Medicare & Medicaid Services. WISeR (Wasteful and Inappropriate Service Reduction) Model. Accessed September 21, 2026.
• Electronic Frontier Foundation. New Records Reveal Problems with Medicare’s AI Prior Authorization Experiment. September 8, 2026. The page links to the underlying CMS records and provider feedback released through FOIA litigation.
• Trang B. Medicare’s AI prior authorization pilot was rushed and full of problems, new documents reveal. STAT. September 15, 2026.
• Muoio D. Delays, tech failures abound in CMS’ WISeR AI prior auth pilot, documents show. Fierce Healthcare. September 15, 2026.
• Centers for Medicare & Medicaid Services. CMS Interoperability and Prior Authorization Final Rule. Accessed September 21, 2026.

This article is for general education and does not provide medical or legal advice. Patients should contact their clinician and Medicare or their insurer for help with an individual authorization or appeal.

Free Daily Briefing

The Latest Longevity Science.
Delivered Every Morning.

Join researchers, physicians, and health professionals getting daily breakthroughs in AI-driven medicine, epigenetics, and longevity research.

Support the research that powers this editorial

Subscribe Free

No spam. Unsubscribe anytime. We respect your inbox.

Keep Reading on Healthcare Discovery

• The Daily Rounds: Longevity & Health Care Brief | April 24, 2026
• Cracking the Undruggable Code: How Daraxonrasib Just Rewrote the Rules for Pancreatic Cancer Survival
• The NAD+ Renaissance: How a Cellular Fuel Molecule Is Rewriting the Science of Brain Aging, Sleep, and Neurodegeneration