WISeR is the clearest live test of AI running a regulated decision in American healthcare. It is a CMS Innovation Center model that runs from 1 January 2026 to 31 December 2031 in Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington, and adds prior authorization to selected services in traditional Medicare. CMS describes the participants as companies that make medical necessity recommendations using enhanced technology like AI, and says all recommendations for non-payment are determined by licensed clinicians.
Price on the decision, by all means. But pay for the decision that holds up, not for the direction it points.
The part that matters for anyone pricing autonomous work is how the vendors are paid. In CMS's own words, participants receive a percentage of the expenditures associated with averted wasteful, inappropriate care, adjusted for performance on process measures including provider experience. Put plainly, the vendor earns more when less care is paid for. CMS adds that the model does not change Medicare coverage or payment policy, and that providers with a record of compliance may in future be exempt from review.
On 27 October the Electronic Frontier Foundation published more than 800 pages of records obtained through its lawsuit against CMS. EFF reports that CMS sent requests for information or corrective action plans to at least four of the six vendors in the first months of the pilot, for problems including determination letters not issued on time, turnaround times that exceeded guidelines, denials that appeared inaccurate or inconsistent with applicable criteria, and a provider portal that did not work. The records also show CMS acknowledging an extremely condensed timeline for the launch.
The story drew fresh coverage this week. Tech Times, reporting on the records on 3 November, quotes Senator Gillibrand saying AI has no place making the final call on Americans' medical care, and a CMS deputy administrator saying that there are a lot of things about WISeR that are not working. KFF Health News had earlier reported physicians describing weeks-long delays and denials that misstated a patient's own record.
None of this is an argument that software cannot run utilization review. It is an argument about what the software is paid to do. A model rewarded for averted spending will find spending to avert, and a human reviewer downstream of it inherits that tilt. EFF's own point is that human review protects patients only if reviewers have the time and authority to overrule the machine, which the records do not show.
We say Layer H backs companies priced on the decision, not the seat. WISeR shows the failure mode of that idea. Pricing on the decision is sound when the price follows the quality of the decision. It breaks when the price follows the direction of the decision, because then every party in the chain, the regulator, the clinician and the patient, has reason to distrust the output, and the license to operate becomes the thing at risk.
Sources
- 01CMS: WISeR (Wasteful and Inappropriate Service Reduction) Model
- 02CMS: CMS launches new model to target wasteful, inappropriate services in Original Medicare (2025-06-27)
- 03EFF: New records obtained by EFF expose more problems with Medicare's AI experiment (2026-10-27)
- 04Tech Times: Medicare's AI prior authorization pilot had 4 of 6 vendors flagged in first weeks, records show (2026-11-03)
- 05KFF Health News: Darius Tahir on delays and errors in Medicare's WISeR pilot (2026-06-23)
Tell us what you are building. Every note gets a real read.