Starting January 2026, six US states launched the WISeR pilot, letting AI decide whether certain Medicare (the federal health program covering Americans 65 and older) outpatient services get reimbursed. On the surface this is a tech upgrade. In practice, it hands the keys to seniors' healthcare decisions to a third party with a built-in conflict of interest.
What this is
WISeR targets a dozen or so services flagged as high-fraud risk, including neurostimulator implants and epidural steroid injections. The workflow: doctors submit requests, AI returns a recommendation in seconds, a human physician "reviews" it.
A human check sounds reassuring, but there's a fatal flaw: third-party vendors are paid on a share of "expenditures avoided." More denials, more revenue. The AI's KPI (key performance indicator) is denial volume.
Medicare spends over $800 billion a year. It has avoided large-scale pre-authorization for six decades — not out of oversight, but because administrative costs typically cancel out the savings. This time, CMS (the Centers for Medicare & Medicaid Services) has decided to ignore that lesson.
Industry view
Supporters argue AI pre-authorization will compress fraud and lift efficiency, and CMS stresses human review as the "compliance backstop."
But the opposition deserves more attention. A September 2026 Electronic Frontier Foundation (EFF) survey found that reviewing physicians spend an average of only dozens of seconds on each AI judgment — AI denial recommendations effectively drive the final call. The Center for Medicare Advocacy has called for the pilot to be terminated immediately, saying the program is harming an already vulnerable senior group. KFF's 2025 data also shows nearly 70% of Americans already view pre-authorization as a burden on care.
We believe the real issue isn't whether the AI is accurate — it's that when incentives are misaligned, an "accurate" AI is more dangerous than a slow one. Technology neutrality is a false premise. The incentive structure is the underlying code.
Impact on regular people
For enterprise IT: This is a lesson any organization must learn before deploying AI in high-stakes decisions. Before model selection, the questions "who pays for wrong decisions" and "does the model's KPI align with business goals" come first.
For individual careers: If you work in approval, customer service, or risk control — roles where you "own the decision" — expect AI to slot you into a "review-and-sign" position, where time pressure erodes real judgment authority.
For consumer markets: This US case will be dissected by regulators and insurers worldwide. If the senior-care AI-denial model gets adopted by commercial insurers, products involving medical claims and health-insurance underwriting in China are likely to face similar disputes — worth tracking early.