CMS Picks Six Tech Firms to Run AI-Powered Medicare Prior Authorization Pilot

For millions of older Americans, the gap between needing care and getting it often comes down to a single bureaucratic hurdle: prior authorization. It is the insurance industry's permission slip, the step where a doctor's recommendation gets checked, sometimes delayed, sometimes denied, before a patient can move forward with treatment. Now that process is about to get an artificial intelligence upgrade inside traditional Medicare itself, and the stakes for patients could not be higher.

The Centers for Medicare and Medicaid Services has chosen six technology companies to administer a new AI-powered prior authorization pilot, STAT has learned. The program, called the Wasteful and Inappropriate Services Reduction model, or WISeR, is CMS's attempt to rein in waste and abuse in a program that spent more than $1 trillion in 2024. That is not a typo. One trillion dollars, in a single year, for a health insurance system that covers roughly 68 million Americans.

Starting in January, the pilot will run through 2031 in six states: New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington. Unlike Medicare Advantage, the privatized version of Medicare where insurers already use algorithms to screen claims, this pilot touches traditional Medicare, the fee-for-service program that has long been seen as the more straightforward, less filtered option for beneficiaries.

Why Private Insurers' AI Troubles Loom Large

Think of prior authorization like a toll booth on the road to treatment. For decades, a human reviewer staffed that booth, reading charts, weighing medical necessity, and deciding whether to lift the gate. AI promises to speed up that process, scanning claims faster than any person could and flagging patterns a human might miss. But speed is not the same as fairness, and that distinction is exactly what has landed private insurers in serious legal trouble.

UnitedHealthcare and Humana are both facing class action lawsuits over their use of AI to make coverage decisions for Medicare Advantage beneficiaries. Congress has also opened its own investigation into how insurers use algorithms to deny claims. The core allegation in these cases is unsettling in its simplicity: that AI systems, trained to find patterns of overuse, end up denying care that patients actually need, and that the people harmed are often the sickest, most vulnerable beneficiaries who have the least capacity to fight back.

That history casts a long shadow over WISeR. CMS insists its pilot targets waste and abuse rather than legitimate care, but critics argue the line between "wasteful" and "necessary" is far blurrier than policymakers like to admit. A knee replacement that looks unnecessary on paper might be the only thing standing between an elderly patient and permanent loss of mobility. An algorithm optimized to cut costs does not always know the difference, and the people writing the code are not the ones living with the consequences.

The political response has been swift. A congressional representative from Washington, one of the six pilot states, is introducing a bill this week aimed at stopping the Department of Health and Human Services from implementing WISeR at all. That a lawmaker is moving to block a federal health pilot before it even launches signals how contentious this issue has become, and how little trust remains after years of reporting on AI-driven denials in the private insurance market.

It is worth pausing on why this distinction matters so much. Medicare Advantage is run by private insurers who have profit incentives baked into every coverage decision. Traditional Medicare, by contrast, was built as a public insurance backstop, free from some of those competitive pressures. Introducing AI-driven prior authorization into that system, even framed as a narrow pilot targeting waste, raises a deeper question about what kind of program traditional Medicare is meant to be, and whether efficiency gains are worth the risk of replicating the same harms patients have already reported under private plans.

Supporters of the pilot would likely argue that smart oversight tools, used carefully, could actually protect patients by catching truly unnecessary procedures and freeing up resources for care that matters. That is a reasonable hope. But hope is not a safeguard, and the six states selected for this experiment will serve as the real-world test of whether AI can distinguish between genuine waste and essential treatment without repeating the mistakes already documented in Medicare Advantage.

Why It Matters

The debate over WISeR is really a debate about trust in automated decision-making at a moment when that trust is already fraying. Patients in New Jersey, Ohio, Oklahoma, Texas, Arizona, and Washington are about to find out, firsthand, whether a six-year pilot program can tell the difference between a wasteful claim and a necessary one. For a program spending over a trillion taxpayer dollars annually, getting that distinction right is not a technical detail. It is the difference between a system that protects people and one that quietly fails them when they are most exposed. Congress clearly has doubts, and given the pattern of lawsuits and investigations already surrounding AI in health insurance, those doubts deserve a serious hearing before this pilot reaches full scale.