Prior authorization has long been one of healthcare’s biggest friction points.
A physician recommends a treatment. The patient is ready to move forward. But first, someone has to determine whether the service meets the insurer’s coverage requirements.
Now, artificial intelligence is entering that process.
In January, the Centers for Medicare & Medicaid Services launched the Wasteful and Inappropriate Service Reduction, or WISeR, Model. The six-year program is testing whether technologies including AI and machine learning can help review certain services in Original Medicare that CMS considers particularly vulnerable to fraud, waste, abuse, or inappropriate use. (Source)
The goal sounds familiar: use technology to make an expensive, administrative process faster and smarter.
But prior authorization is not just another workflow.
It can determine whether a patient gets access to care.
What Is WISeR Actually Doing?
WISeR is currently operating in six states: Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington. It applies only to selected services, including skin and tissue substitutes, electrical nerve stimulators, and knee arthroscopy for osteoarthritis. Emergency and inpatient-only services are excluded. (Source)
Importantly, AI is not independently denying Medicare coverage.
Participating technology companies can use enhanced technology to assess whether requests meet existing Medicare coverage criteria, but CMS requires appropriately licensed clinicians to make all recommendations for non-payment. The model also does not change Medicare’s existing coverage policies. (Source)
That distinction matters.
But so does the role the technology plays before a decision ever reaches a human.
Faster Decisions Could Be Better Decisions
There is a strong case for modernizing prior authorization.
CMS estimates that requesting prior authorizations costs providers $20 to $50 per hour and consumes an average of 13 hours each week. Across a year, that translates to roughly 700 hours of administrative work per provider. (Source)
Technology could potentially change that.
An AI system can analyze documentation, compare information against coverage requirements, flag missing information, and help straightforward requests move through the system faster.
Under WISeR, requests submitted through participant electronic portals are generally subject to a 72-hour turnaround requirement. (Source)
For providers and patients accustomed to waiting on approvals, speed matters.
But speed is not the only metric that matters.
What Happens When the Algorithm Gets It Wrong?
Prior authorization sits at an unusual intersection of clinical judgment, financial incentives, and access to care.
That makes AI implementation particularly complicated.
A model could miss relevant context buried in a patient’s history. Training data could reflect existing disparities in healthcare. A system optimized to identify unnecessary spending could potentially behave differently from one optimized primarily around patient access.
And WISeR introduces another important consideration: participating companies receive a percentage of expenditures associated with care determined to be wasteful or inappropriate, with payments adjusted based on performance measures including provider experience. (Source)
That makes transparency and oversight especially important.
HealthTech companies operating in this space need more than accurate models. They need clear audit trails, explainable recommendations, reliable escalation processes, and mechanisms for clinicians to challenge technology-assisted conclusions.
The Bigger Prior Authorization Experiment
WISeR is only one part of a much broader modernization of prior authorization.
Beginning January 1, 2027, certain CMS-regulated health plans will be required to implement standardized prior authorization APIs designed to move more of the process away from phone calls, faxes, and manual workflows. CMS estimates its broader interoperability and prior authorization policies could save approximately $15 billion over ten years. (Source)
That means the question facing HealthTech is becoming bigger than whether AI can help make coverage decisions.
It is how we design technology that makes those decisions faster without making them less accountable.
Because when an algorithm sits between a patient and a procedure, efficiency cannot be the only measure of success.
