Part of the AI Regulation News hub.
CMS is soliciting comment on revaluing primary care because generative and agentic AI were not anticipated when the relative value units underlying the Physician Fee Schedule were defined
Medicare's fee schedule prices a physician's work using relative value units built around time and intensity. CMS has now written down, in a proposed rule, that the technology reshaping that work was not anticipated when those inputs were defined, and asked what to do about it.
Bottom line: A comment solicitation inside a proposed rule, not a proposal to change payment. CMS is seeking comment on how to re-imagine and improve the relative valuation of primary care services, both within office and outpatient evaluation and management services and through alternatives to fee-for-service.
Who this affects: Primary care physicians and the organisations that employ them, health systems modelling Medicare revenue, developers of ambient documentation and agentic clinical tools, and anyone whose business case assumes current evaluation and management valuation holds.
Issue date: Published 16 July 2026 at 91 FR 43842. Comments must be received by 14 September 2026, referring to file code CMS-1848-P.
What changed: Nothing in payment. What changed is the record: CMS has stated in the Federal Register that generative and agentic AI were not anticipated when the relative value units and the time and intensity of physician services were defined.
Analysis: Note which direction the argument runs. This is not framed as AI reducing physician work and therefore justifying lower payment. It is framed around primary care being revalued upward in support of preventive care, with AI as one reason the existing measurement basis no longer fits. The two readings imply opposite outcomes, and the comment period is where that is contested.
Primary sources: CY 2027 PFS proposed rule, official GPO text (91 FR 43842) · Same document, GPO PDF
- Instrument
- Medicare and Medicaid Programs; CY 2027 Payment Policies Under the Physician Fee Schedule and Other Changes to Part B Payment and Coverage Policies; Medicare Shared Savings Program Requirements; and Medicare Prescription Drug Inflation Rebate Program
- Authority
- Centers for Medicare & Medicaid Services, Department of Health and Human Services
- Citation
- 91 FR 43842, pages 43841 to 44557, published 16 July 2026
- Regulatory identifiers
- 42 CFR Parts 400, 405, 406 and others; file code CMS-1848-P
- Jurisdiction
- United States, federal
- Status
- Proposed rule. Open for comment.
- Bindingness
- None on this point. The AI material sits in a comment solicitation about future valuation policy, not in a proposed payment change.
- Issue date / next deadline
- Published 16 July 2026. Comments must be received by 14 September 2026.
- Policy framing
- Presented as part of the HHS Secretarial priority described as Make America Healthy Again, with the stated aim of shifting toward preventive rather than reactive care
- Editorial Note
- Informational analysis for working professionals, not legal advice. Confirm how any rule applies to your situation with qualified counsel.
- Primary source
- https://www.govinfo.gov/content/pkg/FR-2026-07-16/html/2026-14327.htm
The sentence that matters
The Physician Fee Schedule prices physician services using relative value units, and those units rest on assumptions about the time and intensity of the work involved. CMS notes that primary care services delivered in the physician office were the locus for continuous, coordinated and comprehensive care when the PFS was established in 1992.
It then records what has changed. The definition of primary care as a beneficiary's first point of contact with the health care system has been increasingly challenged by broader access to medical information through digital resources, and CMS says this may accelerate as adoption of technology reshapes access to increasingly sophisticated sources of medical information before beneficiaries ever show up in the doctor's office.
The operative statement follows: advances in technology, including generative and agentic artificial intelligence, are poised to transform both beneficiary experience and the role of the primary care clinicians, and these were not anticipated when defining the relative value units and time and intensity of services delivered by physicians, which are essential inputs to the PFS.
That is an agency writing into the Federal Register that a core input to how it pays physicians predates, and does not account for, the technology now reshaping the work.
What CMS is actually asking
The solicitation is specific about the frame. CMS says it is seeking comment on how to re-imagine and improve the relative valuation of primary care services, and that it is focused on understanding how that re-imagination might occur within the current construct of office and outpatient evaluation and management services, as well as via alternatives to fee-for-service payment.
It names two such alternatives: outcomes-based payment, and expansion of prospective primary care payment, noting that the latter has long been a goal of external policy experts.
The wider context it gives is the HHS Secretarial priority described as Make America Healthy Again, and an interest in whether Original Medicare might incorporate more robust incentives to invest in high-value care that reduces long-term costs, building on recent HHS rulemaking that established such incentives for commercial health plans.
So the question on the table is not whether AI should be paid for. It is whether the unit of payment for primary care should change, with AI cited among the reasons the current unit no longer describes the work.
Why the framing direction matters commercially
There are two ways an agency can reason from AI to physician payment, and they point in opposite directions.
One is efficiency substitution: if software absorbs part of the work, the time and intensity behind a code fall, and the valuation should fall with them. That argument has been made in the health policy literature, and the rule's own footnotes cite work asking whether artificial intelligence could affect physician payment for nonprocedural services.
The other is the one CMS actually advances here: that primary care is under-valued relative to its importance, that the shift toward preventive care requires stronger incentives, and that technological change is among the reasons the existing measurement basis needs rethinking.
For anyone modelling revenue, the practical point is that the comment period is where those two readings compete, and the record CMS has created cites AI in support of the second. A comment that engages only with the first will be arguing against a frame the agency has not adopted.
The other AI provision in the same rule
Separately, the proposed rule contains a section on Software as a Medical Service laboratory analyses. CMS observes there have been rapid developments in the use of software-based technologies with novel functionalities, including artificial intelligence, to support clinical decision-making in outpatient and physician office settings, and that new clinical software including clinical decision support software, clinical risk modelling and computer aided detection is becoming increasingly available to practitioners.
That is a coding and payment question rather than a valuation question, and it runs alongside the same concept as it appears in the outpatient payment system.
Read together, the rule contains both a near-term mechanical question about how AI-based software services are coded and paid, and a longer-term structural question about whether the relative value framework survives contact with agentic tools. They are on different timescales and should not be conflated.
What to file, and by when
Comments must be received by 14 September 2026 to be assured consideration, referring to file code CMS-1848-P, submitted by one of the three routes the rule sets out.
The most useful evidence will be empirical: what generative or agentic tools actually do to the time and intensity of an office visit today, measured rather than asserted. CMS is soliciting comment on a valuation question, and valuation arguments without data tend to be recorded and not adopted.
A date note, and it is the second instance of this pattern in the same channel. The Federal Register text sets 14 September 2026; the regulations.gov docket record for this rule displays 15 September, as it does for the CLIA request for information published the same day. The published rule governs.
The rule is very large, running from page 43841 to 44557. The primary care valuation discussion is a small part of it, and a comment on this point should identify the section rather than the rule at large.
Two things to take from this. First, a factual one worth quoting accurately in any internal paper: CMS has stated in a proposed rule that generative and agentic AI were not anticipated when the relative value units and the time and intensity of physician services were defined. That is now on the record and can be cited. Second, a strategic one: the solicitation is framed around revaluing primary care upward in support of preventive care, not around AI reducing physician work, and those two framings imply opposite payment outcomes. If your organisation has a view, 14 September 2026 under file code CMS-1848-P is where it counts, and empirical evidence about what these tools do to visit time and intensity will carry further than argument.
Source File
https://www.govinfo.gov/content/pkg/FR-2026-07-16/html/2026-14327.htm
Open the GPO text and confirm four things: the header showing 42 CFR Parts 400, 405, 406 et al. with the CY 2027 Physician Fee Schedule title, published 16 July 2026; the DATES line setting 14 September 2026 and file code CMS-1848-P; the primary care introduction stating that advances in technology including generative and agentic AI were not anticipated when defining the relative value units; and the Software as a Medical Service laboratory analyses section.
Advances in technology, including generative and agentic artificial intelligence (AI), are poised to transform both beneficiary experience and the role of the primary care clinicians. These were not anticipated when defining the relative value units and time and intensity of services delivered by physicians, which are essential inputs to the PFS. CMS, CY 2027 Physician Fee Schedule proposed rule, 91 FR 43842, 16 July 2026
FAQ
Is CMS proposing to change physician payment because of AI?
No. This is a solicitation of comment on how to re-imagine and improve the relative valuation of primary care services. No payment change is proposed on this basis in the rule.
What exactly did CMS say about AI?
That advances in technology, including generative and agentic artificial intelligence, are poised to transform both beneficiary experience and the role of primary care clinicians, and that these were not anticipated when defining the relative value units and the time and intensity of services delivered by physicians, which are essential inputs to the Physician Fee Schedule.
Does this mean payment will go down because AI does the work?
That is not the framing CMS uses. The discussion sits inside an argument for revaluing primary care in support of preventive care, with AI cited as a reason the existing measurement basis no longer fits. The opposite reading exists in the policy literature and the rule's footnotes cite it, but the agency's own framing is the former.
What alternatives is CMS considering?
It is focused on how revaluation might occur within the current construct of office and outpatient evaluation and management services, and via alternatives to fee-for-service payment including outcomes-based payment and expansion of prospective primary care payment.
When are comments due?
By 14 September 2026, referring to file code CMS-1848-P. The regulations.gov docket record displays 15 September; the Federal Register text governs.
Is there anything else about AI in this rule?
Yes. A section on Software as a Medical Service laboratory analyses notes rapid developments in software-based technologies with novel functionalities, including AI, supporting clinical decision-making, and names clinical decision support software, clinical risk modelling and computer aided detection.
Related briefings
Sponsored Training
Practical AI training for regulated professionals, built around verification, documentation and a defensible process. See the courses.