There are two Medicare stories about artificial intelligence running right now, and only one of them is getting attention. The attention is going to a question that is still theoretical: whether Medicare will eventually pay an algorithm the way it pays a physician. Becker's ASC Review reported on September 15 that HHS and CMS are drafting a payment category for AI tools that deliver care or generate diagnoses. That work is internal. Nothing has been published, proposed, or priced.
The other story is already live. Since January 1, 2026, a CMS Innovation Center model has been applying technology-assisted review to fourteen categories of service across six states. Epidural steroid injections are on the list. So are cervical fusion, arthroscopic debridement of the osteoarthritic knee, percutaneous lumbar decompression, and skin and tissue substitutes. If your practice bills Original Medicare in Texas, New Jersey, Oklahoma, Ohio, Washington, or Arizona, this is not a forecast about the future of AI. It is the claims environment you are submitting into today.
And in July, CMS proposed the first real payment architecture for clinical software in the outpatient rule. Thirty-six codes. A new status indicator. A candid admission from the agency that the whole framework is interim. Most practices spent the summer reading the physician fee schedule. This one moved with far less noise and will shape reimbursement for longer.
Key Takeaways
The Wasteful and Inappropriate Service Reduction Model, known as WISeR, has been live since January 1, 2026 and runs through December 31, 2031.
It applies to Original Medicare only, in Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington. CMS states it does not apply to Medicare Advantage.
Fourteen service categories are in scope for performance year one, including epidural steroid injections, cervical fusion, arthroscopic lavage and debridement for the osteoarthritic knee, percutaneous lumbar decompression, and skin and tissue substitutes.
Providers choose one of two paths: submit prior authorization before the service, or proceed without it and accept post-service, pre-payment medical review.
CMS states that all recommendations for non-payment are determined by appropriately licensed clinicians, and that existing appeal rights are unchanged.
Separately, the CY2027 OPPS and ASC proposed rule creates a Software as a Medical Service category with a new O1 status indicator and designates 36 HCPCS codes. CMS calls it a first step, not a settled methodology.
What is the WISeR model actually doing to claims right now?
CMS launched WISeR on January 1, 2026. The performance period runs six years, through December 31, 2031, in six states. It touches Original Medicare only, and CMS has been explicit that Medicare Advantage members are not affected by it.
The model gives providers two paths. You can submit a prior authorization request to a model participant before delivering the service. Or you can deliver the service without one and go straight into a post-service, pre-payment medical review. Read that carefully, because practices keep misreading it. Neither path is optional in the way that matters. If the service is on the list and the patient is in one of the six states, a review is happening. You are choosing when, not whether.
Participants conducting those reviews are paid a percentage of the expenditures associated with averted wasteful or inappropriate care. CMS adjusts that percentage based on performance measures, including provider experience.
Which of my procedures are actually on the list?
Fourteen categories for performance year one, and the spread across specialties is wider than most practices expect.
Pain management sees epidural steroid injections and induced lesions of nerve tracts. Spine sees cervical fusion, percutaneous vertebral augmentation, and percutaneous lumbar decompression for spinal stenosis. Orthopedics sees arthroscopic lavage and debridement for the osteoarthritic knee. Wound care sees skin and tissue substitutes, which is the line item most likely to catch a practice off guard.
The remaining categories are electrical nerve stimulators, sacral nerve stimulation for urinary incontinence, phrenic nerve stimulators, vagus nerve stimulation, hypoglossal nerve stimulation for sleep apnea, incontinence control devices, and diagnosis and treatment of impotence.
One operational detail matters here. CMS publishes the specific CPT and HCPCS codes in the model's Operational Guide, not in the provider fact sheet. Mapping your own codes is a deliberate step. It is not something to assume from a category name.
Is an algorithm denying these claims on its own?
No, and the distinction changes how you should respond. CMS states that all recommendations for non-payment are determined by appropriately licensed clinicians who apply standardized, transparent, evidence-based procedures. Technology accelerates the review. A clinician signs the outcome.
Abe Sutton, director of the CMS Innovation Center, has framed the incentive design as rewarding accurate determinations rather than volume of denials, and clinicians retain their existing appeal rights under the model.
That is the design on paper. The practical question for your practice is narrower. Does your documentation establish medical necessity at the moment the claim goes out, or does it get assembled after a request comes back? A reviewer working from a standardized evidence framework finds what is in the note. Nothing more.
What is Software as a Medical Service, and why should a specialty practice care?
CMS issued the CY2027 Hospital Outpatient Prospective Payment System and Ambulatory Surgical Center proposed rule on July 2, 2026. It published July 7 and comments closed August 31. Inside it is the first serious attempt to define how Medicare pays for clinical software.
CMS proposes the term Software as a Medical Service for software-based technologies that support clinical decision-making through algorithmic analysis. The agency draws a deliberate boundary against prescription digital therapeutics and remote monitoring, which are paid differently. It proposes a new O1 status indicator in the outpatient setting carrying the same payment treatment as the existing S indicator, meaning separate payment without a multiple-procedure reduction. CMS also asked for comment on whether a T indicator would be the better fit.
The numbers are concrete. Thirty-six HCPCS codes designated as SaMS. Twenty-one moved from clinical APCs into New Technology APCs. Ten pulled off the Clinical Laboratory Fee Schedule. Covered examples run to AI retinal image analysis, CT-derived fractional flow reserve, quantitative brain MRI analysis, algorithmic ECG risk assessment, and fracture-risk modeling.
One caveat worth stating plainly, because it is the kind of thing that gets blurred in vendor marketing. The status indicator framework sits inside the hospital outpatient payment system. The rule covers both OPPS and ASC, but the SaMS provisions as proposed are an outpatient construct. Any claim that Medicare now reimburses ASCs for AI diagnostics should be read against that.
Where is the federal government taking clinical AI next?
On September 9, 2026, ARPA-H announced ADVOCATE, short for Agentic AI-EnableD CardioVascular CAre TransfOrmation. Up to $33.7 million in year one, and $62.7 million committed over four years. The stated goal is an FDA-authorized clinical AI capable of delivering certain cardiovascular care autonomously while pulling in the clinical team when needed.
Physicians have already formed a view on who should benefit financially. The Doximity 2026 Physician Compensation Report, drawing on nearly 23,000 U.S. physicians surveyed over the past year, found more than 40% believe doctors should capture most of the cost savings AI produces. Close to half said physicians rather than health systems or payers should be the primary beneficiaries. More than 65% already use AI daily or weekly.
Put those two things side by side and the direction is clear enough. Federal money is moving toward autonomous clinical AI. The payment plumbing is being built in parallel. And the people delivering the care have opinions about where the savings land.
What This Means for Your Practice
Three things, in order of urgency.
If you bill Original Medicare in one of the six WISeR states for any of the fourteen categories, map your codes against the model's Operational Guide this quarter. Not next year. Reviews are happening on claims you are submitting now, and the choice between prior authorization and pre-payment review is a cash flow decision. It deserves to be made deliberately rather than by default.
Second, move medical necessity documentation to the front of the process. Practices that document to the evidence standard at the point of service do not have to reconstruct it later under a clock. The ones that do not will feel it in days in A/R long before they feel it in denial rate.
Third, watch the final outpatient rule. CMS describes SaMS as interim by its own admission, which means the methodology will change. The codes your practice may eventually bill for AI-assisted diagnostics are being sorted into payment buckets right now.
None of this requires buying technology. It requires knowing which of your codes are exposed and running a process that holds up under someone else's review. That is a revenue cycle problem before it is an artificial intelligence problem.
Frequently Asked Questions
Talk to Cosentus
Cosentus has spent 25+ years inside specialty revenue cycle, and our R+A approach keeps a qualified human accountable for every decision our AI accelerates. If your practice bills in a WISeR state, we can map your exposed codes and show you where documentation runs thin before a reviewer finds it for you.
Start at cosentus.com/contact