The Reimbursement Turn: How a SaMS Category, Ten Algorithmic Lab Codes, and a Wellness-Visit RFI Would Teach Medicare to Buy AI Care
CMS’s CY 2027 proposed rules would give clinical software its own payment identity for the first time — and the freshly closed comment record shows physicians and AI advocates still divided on what Medicare should actually buy.
The 1,592-Page Question
On July 14, 2026, the Centers for Medicare & Medicaid Services (CMS) issued the CY 2027 Medicare Physician Fee Schedule proposed rule (CMS-1848-P), announcing payment and policy changes that would take effect on or after January 1, 2027.[1] The rule runs 1,592 pages, and its digital health content is not confined to a single section — it is spread across code valuation, a primary care Request for Information (RFI), the Quality Payment Program, the Shared Savings Program, and a separate interoperability RFI.[2] The public comment window closed September 14, 2026.[3]
Buried in that machinery is a first: CMS’s initial structural attempt to pay for software-based clinical services as their own category, rather than as a cost buried inside reimbursement for something else. As Wilson Sonsini’s health care practice observed, the consequence is a new reimbursement vehicle through which software and AI-enabled products could be paid as standalone services in the future.[4]
The timing makes the stakes sharper. Both PFS conversion factors would decline under the proposal — to $33.17 (-1.19%) for qualifying APM participants and $32.84 (-1.68%) for everyone else — because the temporary 2.50% conversion factor increase Congress provided for CY 2026 expires.[1] CMS is designing how to fund new technology at the same moment the overall payment pie shrinks. Payment design, not technology, will decide which AI tools reach Medicare patients.
Why the Fee Schedule Struggles With Software
The PFS pays each service based on relative value units (RVUs) reflecting three components: physician work, practice expense, and malpractice costs. Software expenses have historically been reimbursed inside practice expense — the non-physician cost bucket for equipment, supplies, and clinical staff time.[4]
The misalignment is structural, not incidental. Much of the value of medical AI comes from reducing or replacing physician interpretation, yet physician work remains the largest driver of payment under the PFS. The result is a pricing paradox:
“A fully autonomous AI service could therefore involve little to no physician work, producing little reimbursement under the ordinary logic of the fee schedule despite providing significant clinical value.”[4]
CMS’s main workaround to date — folding software value into practice expense, as it did for autonomous retinal imaging analysis (CPT 92229) — has produced generally low reimbursement. The deeper problem is that the practice expense methodology rests substantially on practice-cost survey data collected in the late 2000s, before software services of this kind existed, and CMS has historically declined to accept per-click software fees as a direct cost input.[4] A payment system priced for exam rooms and cotton swabs has no natural slot for an algorithm.
Enter SaMS: Software as a Medical Service
The CY 2027 proposed rules — the Physician Fee Schedule released July 14 and the Hospital Outpatient Prospective Payment System (OPPS) rule released July 2 — retire CMS’s old term “Software as a Service” in favor of Software as a Medical Service (SaMS), defined as software-based technologies that support clinical decision-making through algorithmic analysis. CMS expressly distinguishes SaMS from prescription digital therapeutics and from remote physiologic and therapeutic monitoring.[4] The renaming signals a decision to treat clinical software as a category of medical service with its own payment approach, not an IT line item.
On the hospital outpatient side, the OPPS proposal takes three concrete steps:[4]
- Designating 36 HCPCS codes as SaMS. These are specific, already-defined clinical services — AI-based retinal imaging analysis, coronary fractional flow reserve derived from CT, quantitative brain MRI analysis, algorithmic ECG risk assessment, fracture-risk modeling. Most already exist and are reimbursed today; the proposal changes how CMS categorizes, and in some cases pays for, each one.
- Creating a new status indicator, “O1,” to identify SaMS on outpatient claims. The indicator would function like the existing “S” — separate payment, no multiple-procedure discounting — while CMS asks whether something more restrictive is appropriate. The significance is visibility: before an agency can reprice a service, it has to be able to see it in claims data.
- Moving 21 SaMS codes currently paid in clinical ambulatory care into New Technology APCs, described as an interim step while CMS develops a longer-term payment methodology, which the agency suggests may better align payment with clinical outcomes.
Notably, CMS is not creating a single payment category for “AI” writ large. Medicare reimbursement remains tied to discrete clinical services, valued code by code. SaMS is a naming and accounting framework — the foundation on which per-service pricing decisions will later be built.
The Ten Codes That Stopped Being Lab Tests
Clinical laboratory tests are paid under a separate system, the Clinical Laboratory Fee Schedule (CLFS), and payment is available only when the test is performed by a laboratory certified under the Clinical Laboratory Improvement Amendments (CLIA). CMS’s proposal draws a new line in that system: standalone algorithmic analyses of existing lab data — genomic reanalysis, AI-based digital pathology review — are not laboratory tests once the underlying data has been generated, and they do not require a CLIA-certified laboratory to perform.[4]
So ten algorithmic laboratory analysis codes would move off the CLFS. On the PFS side, CMS proposes no fixed national rate at all — local Medicare Administrative Contractors (MACs) would determine payment, with all future algorithmic lab-analysis codes likewise contractor-priced and a comment solicitation on which additional codes should be reclassified.[2] The practical consequences are threefold: MAC pricing discretion replaces CLFS methodology, beneficiary cost-sharing attaches, and budget-neutrality requirements apply.[2]
The core logic is a two-step unbundling. CMS is separating the data-generating step (still a lab test, still CLFS) from the algorithmic interpretation step (now a medical service). Payment follows the interpretation. For companies building AI diagnostics, this classification is close to existential: which fee schedule you live on determines your national rate, your exposure to cost-sharing, and whether you have a national rate at all.
Should an AI Company Deliver Your Wellness Visit?
Alongside the payment mechanics, the PFS rule contains an RFI titled “Redesigning Primary Care to Make America Healthy Again,” seeking comments on three topics: reconsidering relative primary care payment in the PFS; understanding the payment implications of including technology in primary care; and establishing prospective primary care payment in the Medicare Shared Savings Program and potentially Original Medicare broadly.[1]
Within that RFI, CMS cites ambient AI scribes and clinical decision support tools by name, points to adoption data, and then asks a set of questions whose direction of travel is hard to miss:[2]
- What requirements “may create barriers to an AI technology company delivering AWV-related services while employing or contracting with appropriately licensed physicians?”
- Should technology-enabled organizations, including AI companies, participate in Annual Wellness Visit (AWV) delivery models directly or through partnerships with enrolled providers?
- What kinds of outcomes should CMS hold technology companies accountable for?
- Should CMS create “technology-enabled care management” codes, and should payments be tied to outcomes?
The Annual Wellness Visit is a revealing place to start. It is standardized, preventive, and low-acuity — the least dangerous room in the house to test whether a technology company employing licensed physicians can deliver a Medicare-covered service. Nixon Law Group’s read of the rule is that CMS is drawing a hard line between technology it pays for as a practice input and AI/technology companies it might treat as care delivery partners — and digital health companies will need to decide which side of that line they want to be on.[2]
The rule also proposes a new Merit-based Incentive Payment System (MIPS) improvement activity crediting clinician use of AI — not a payment rule, but a signal of where CMS wants behavior to move.[4]
What the Comment Record Says
With the window now closed, the filed comments sketch the fault lines. The Bipartisan Policy Center (BPC), a bipartisan think tank, urged CMS to anchor any AI payment in evidence rather than enthusiasm: “CMS should link clinical AI payment to demonstrated value, including evidence that a tool improves clinical outcomes, lowers total cost of care, or meaningfully expands access.”[5]
BPC’s comment also flagged the structural limits of every alternative to fee-for-service for AI. Provider organizations contract with multiple payers, each with different administrative structures, making it hard to redesign care around tools that benefit only a subset of patients; savings are difficult to attribute to any specific tool because AI-driven care is ongoing rather than episodic; and high upfront implementation costs fall on the deploying organization before any return materializes.[5] On the evidence itself, BPC noted that ambient scribe technology shows promise for reducing documentation burden, but its impact on efficiency metrics — clinician time saved, patient throughput, coding and financial outcomes — “remains mixed or inconclusive.”[5]
The American Medical Association (AMA), the physicians’ professional association, focused its opposition on the remote monitoring proposals that ride along in the same rule. CMS would require that clinical staff furnishing remote physiologic monitoring (RPM) or remote therapy monitoring (RTM) be direct employees of the billing practice — barring contracted third-party staff — while cutting practice-expense RVUs for several codes and offering an alternative that would bundle all 17 RPM and RTM codes into four new G-codes, which the AMA says would sharply reduce code specificity and implement significant cuts all at once. The AMA plans to oppose the employee-only requirement, and the Relative Value Scale Update Committee is moving its review of the codes up by a year, urging CMS to postpone the reductions and allow time to collect data.[6]
The BPC comment highlights the same RPM proposals from the access side: barring vendor delivery could disrupt services for practices that contract with third parties precisely because they lack monitoring capacity in-house, and the agency’s new ACCESS model — which has accepted more than 150 technology vendors — may not absorb displaced services, because its payment rates encourage participants to use AI in place of human-led care management, while remote monitoring improves outcomes only when someone can detect and act on the data, a response that is still largely human today.[5]
A separate RFI inside the same rule takes aim at the AMA’s own franchise: CMS solicits comment on the “harms or challenges associated with AMA’s monopoly over CPT-4 licenses” and asks whether private competition could supplement the CPT-4 coding standard, noting that the reliance on CPT-4 as the national code set is an HHS regulatory interpretation rather than a statutory requirement.[2] CMS is simultaneously questioning who codes the services and how the services are paid — the two levers that determine a technology’s economics.
What to Watch
A final rule is expected by November 1, 2026, with most changes taking effect January 1, 2027.[6] The likely survivors are the structural pieces: the SaMS terminology, the 36-code designation, the O1 status indicator, and the ten lab codes’ move off the CLFS. The contested pieces are the RPM staffing restriction and the G-code bundling, where CMS now faces opposition from both the AMA and access-oriented commenters. The AI-delivered-wellness-visit question is an RFI, so its answer will arrive in the preamble text of future rules rather than this one.
The unresolved question is the one the fee schedule was never built to answer. A payment system that prices physician time is now absorbing technology whose value is measured in physician time saved. The AMA’s CPT Appendix S taxonomy already sorts AI services into assistive, augmentative, and autonomous categories based on the role the software’s output plays in clinical interpretation.[4] CMS has now named the payment category those services will live in. What it has not yet decided — and what the next several rule cycles will decide — is what a unit of algorithmic analysis is worth when no physician is in the room.
References
[1] [PRIMARY] “Calendar Year (CY) 2027 Medicare Physician Fee Schedule Proposed Rule.” Centers for Medicare & Medicaid Services Newsroom Fact Sheet. July 14, 2026. https://www.cms.gov/newsroom/fact-sheets/calendar-year-cy-2027-medicare-physician-fee-schedule-proposed-rule
[2] [SECONDARY] “What Does the CY 2027 Medicare Physician Fee Schedule Proposed Rule Mean for Digital Health Companies?” Nixon Law Group. July 15, 2026. https://www.nixonlawgroup.com/resources/what-does-the-cy-2027-medicare-physician-fee-schedule-proposed-rule-mean-for-digital-health-companies
[3] [PRIMARY] “Medicare and Medicaid Programs; CY 2027 Payment Policies Under the Physician Fee Schedule and Other Changes.” Federal Register. July 16, 2026. https://www.federalregister.gov/documents/2026/07/16/2026-14327/medicare-and-medicaid-programs-cy-2027-payment-policies-under-the-physician-fee-schedule-and-other
[4] [SECONDARY] Daniel, Jodi, and Kayam, Ty. “CMS Proposes Payment Frameworks for ‘Software as a Medical Service.’” Wilson Sonsini Goodrich & Rosati, via JD Supra. July 23, 2026. https://www.jdsupra.com/legalnews/cms-proposes-payment-frameworks-for-7552931/
[5] [ADVOCACY] Gaumer, Zachary. “BPC’s Comment on the CMS 2027 Physician Fee Schedule (PFS) Proposed Rule.” Bipartisan Policy Center. September 2026. https://bipartisanpolicy.org/testimony-letter/bpcs-comment-on-the-cms-2027-physician-fee-schedule-pfs-proposed-rule/
[6] [ADVOCACY] “2027 proposed Medicare fee schedule: What physicians need to know.” American Medical Association. September 2026. https://www.ama-assn.org/practice-management/medicare-medicaid/2027-proposed-medicare-fee-schedule-what-physicians-need-know
AI automated and generated content. Human reviewed.
Listen to the Bears Discuss This

Episode 26 of AI Policy for the Bears
The Weekly AI Policy Brief
AI policy moves fast. Get the week’s developments, new trackers, and podcast episodes in one short email — free, ad-free, no spam.
Leave a Reply