WISeR pilot: AI approval system was rushed and error-prone

The Medicare pilot program WISeR, short for Wasteful and Inappropriate Service Reduction, began in January 2026. It requires providers to obtain prior authorization before performing specific procedures or ordering products, such as skin substitutes and epidural injections for pain. The experiment spans New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington, with a planned duration until 2031.

The Electronic Frontier Foundation secured more than a thousand pages of documents and data through a public records lawsuit. These materials reveal a rushed rollout riddled with errors. One WISeR technology vendor warned the Centers for Medicare and Medicaid Services that expecting a finished product by the fixed implementation deadline was unrealistic. As a STAT report from September 15, 2026, describes, the result was delays in service delivery.

What it means for patients

Prior authorization programs are not new to the American healthcare system. What is new is the degree of automation in decision-making and the speed of implementation. When an algorithm issues approvals or denials for pain treatments or products used on hard-to-heal wounds, every delay translates directly into patient waiting time, not administrative costs.

The documents also show an information asymmetry. The agency knew about the vendor's warnings and still did not move the deadline. In practice, public oversight of such a program rests on court-ordered disclosures, because applications and correspondence with contractors are not published on an ongoing basis.

The scale of the disclosure matters on its own. This is not a single document but more than a thousand pages of correspondence, notes, and materials that the agency released only under court pressure. The picture that emerges is consistent: a political deadline coupled with technology that was not ready. In a program whose horizon reaches 2031, fixes after launch are not the exception but the rule. What stays open is who monitors, during that time, how many approvals and denials the algorithm issues, and on what basis.

In the background, another dispute is playing out over how Medicare pays for care technologies. The agency is expanding the ACCESS model, in which it pays companies for managing chronic diseases, while at the same time proposing to limit remote patient monitoring, which we cover in separate articles. The common thread is one: more and more decisions about services are made in code, and fewer and fewer of them are first tested on paper.