WISeR pays AI vendors a share of averted Medicare spending

When a Medicare contractor keeps a share of the spending it prevents, the incentive is not subtle. That is the arrangement inside WISeR, the Centers for Medicare and Medicaid Services pilot that launched in January and runs for six years in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington. Vendors use AI plus clinical review to screen prior authorization requests on a defined list of traditional Medicare services, and, as Ars Technica lays out, the compensation is tied to averted expenditures. Reporting from KFF Health News pins the vendor share at 10 to 20 percent of those averted dollars.

The covered list is narrow but revealing: electrical nerve stimulator implants, epidural steroid injections excluding facet joint injections, percutaneous vertebral augmentation for compression fractures, and percutaneous image-guided lumbar decompression for spinal stenosis. These are procedures with legitimate overuse concerns and real patients who need timely answers. Early reporting says approvals that used to take about two weeks are now landing in the four to eight week range, and doctors interviewed by KFF called the process horrendous. One Oklahoma radiologist reportedly documented four times that a patient lacked numbness and still had the WISeR application denied on the ground that numbness ruled out the spinal procedure.

The political stakes turned concrete on July 16, when the Senate voted 46 to 50 along party lines to preserve WISeR. According to STAT, Democrats led by Ron Wyden of Oregon, Patty Murray of Washington, Maria Cantwell of Washington, Richard Blumenthal of Connecticut and Kirsten Gillibrand of New York had wanted to end the pilot, arguing that a payment structure familiar from Medicare Advantage, where investigators previously found 13 percent of denials should have been approved, was being imported into traditional Medicare.

The honest caveat is that WISeR is still early. The reporting does not name which AI vendors CMS actually selected, does not give a clean baseline denial rate for the targeted procedures, and does not spell out how the timeliness and provider-experience adjustments practically reduce a vendor's payout. Whether the model saves money without harming patients is, right now, an argument between economists and physicians rather than a settled number.

What is already actionable is the operational picture. Providers in the six pilot states are the ones who need to staff up documentation and appeals work, because a contractor whose upside is the value of what it blocks will not unblock more than the performance adjustments compel. WISeR runs through December 2031, long enough for whoever wins that operational contest to become the reference model for AI in payer decisions across the rest of Medicare.