CMS Introduces Prior Authorization Model That Incorporates Artificial Intelligence in the Process

ACP is warning the model's implementation could have unintended consequences; says prior authorization reform should focus on transparency, evidence-based patient care

Sept. 12, 2025 (ACP) -- The Centers for Medicare & Medicaid Services recently introduced a new prior authorization (PA) model, the Wasteful and Inappropriate Service Reduction (WISeR) Model, which will bring artificial intelligence (AI) and machine learning technologies into PA processes for select Medicare services.

CMS explains that the WISeR Model, set to run for six performance years from 2026 to 2031, will work with companies that use enhanced technologies to improve the review process for specific services that have been deemed “vulnerable to fraud, waste and abuse.” While CMS has stated that licensed clinicians will make final decisions despite technology supporting the review process, ACP has raised concerns directly to CMS about the model's payment structure and potential unintended consequences.

“The model rewards companies based on the cost savings they achieve, which could create a financial incentive to deny services,” said Dejaih Johnson, ACP manager of regulatory affairs. “Inviting the private sector to play a more active role in program design and implementation must be accompanied by strong safeguards to prevent undue influence and protect clinical decision-making.”

Participating companies will operate in assigned geographic regions and receive payments based on their ability to reduce unnecessary or noncovered services and lower overall Medicare spending. The model will initially be implemented in six states -- Arizona, New Jersey, Ohio, Oklahoma, Texas and Washington -- focusing on skin and tissue substitutes, electrical nerve stimulator implants and knee arthroscopy for knee osteoarthritis.

“Prior authorization has not consistently shown meaningful cost savings, and there is growing evidence of the real and often irreversible harm it causes, including treatment delays and care denials,” Johnson said. “The best way to alleviate the harm and burden of prior authorization is not to reinforce it, but to eliminate it where there is strong evidence to do so.”

ACP warns that the model's approach, which outsources functions to AI and substitutes manual barriers with algorithm-driven processes, will not meaningfully alleviate existing burdens and could introduce additional challenges such as bias and opaque decision-making.

“Physicians should not have to delay care while waiting for determinations that they are otherwise qualified to make directly,” Johnson said.

According to Johnson, some private insurers have been criticized for using AI to deny care. “Prior authorizations have been the subject of much scrutiny through research demonstrating subsequent limited access to care in the aftermath of denials from the use of AI technologies,” she said. “These issues have also been the subject of recent Senate hearings, where recent research has demonstrated disparities in prior authorization decisions and practices, highlighting how insurers prioritize profit over patient care.”

Rather than implementing AI-driven processes without addressing structural flaws, ACP supports reform that starts with an evidence-based assessment of where PA is truly effective. Johnson emphasized that improvements must focus on transparency, evidence-based practices and patient-centered care.

The organization has supported CMS's efforts to modernize the process by promoting interoperability through electronic submissions and requiring quicker turnaround times. ACP also supports the bipartisan Improving Seniors' Timely Access to Care Act, which would codify improvements in Medicare Advantage.

Back to the September 12, 2025 issue of ACP Advocate