Less than a month remains for physicians, health systems, technology companies, and other stakeholders to comment on several questions buried deep in a Request for Information (“RFI”) within the Centers for Medicare & Medicaid Services’ (“CMS”) Calendar Year 2027 Physician Fee Schedule Proposed Rule.
Among them is a deceptively simple question: “What are the payment implications of including technology in primary care?”[1]
Most of us probably do not yet appreciate how consequential that question could become.
On its face, this is a question about Medicare payment policy. The answer, however, could influence much more than reimbursement. It could affect how medical services are defined and delivered, how physicians and other clinicians are trained, how health care organizations structure their workforces, and how much of the work of caring for patients remains with people as technology becomes increasingly capable of performing tasks that clinicians perform today.
It reminded me of a question from a health policy class I took in the early 2000s, should medical students be taught how to use electronic health records as part of their medical school curriculum?
At the time, it was a reasonable question. Electronic health records were still becoming part of clinical practice, and it was not yet clear how central they would become to the physician's work. Today, the question seems almost quaint. Medical students do not simply learn how to use an EHR. They enter a health care system in which the electronic record is fundamental to how clinical information is documented, communicated, accessed, measured, and analyzed. The questions we are asking today about artificial intelligence may eventually look much the same.
AI is already moving into clinical workflows. The more important question is what happens as its role expands.
- Will AI primarily help physicians manage information and make better decisions?
- Will it change how clinicians are trained and the skills they are expected to develop?
- Will it allow health care organizations to redesign clinical workflows?
- Will it change the number and mix of clinicians required to provide care?
- And as we continue to emphasize the importance of keeping a “human in the loop,” what happens when the economic incentives associated with AI make greater automation more attractive than maintaining current staffing models?
These are not simply technology questions. They are questions about the structure and economics of health care.
Why CMS Request for Information Matters
CMS is approaching one part of this much larger issue through payment policy. If technology becomes part of the way primary care is delivered, Medicare eventually will have to determine what it is paying for.
- Is the technology simply an input into an existing physician service?
- Does its use create additional clinical work that should be recognized?
- Should different types of AI-enabled services be valued differently?
- What evidence should demonstrate that technology is actually improving care?
Those questions become more important as AI moves from administrative functions toward clinical decision-making.
The American Medical Association has already developed a taxonomy distinguishing assistive, augmentative, and autonomous AI applications in medical services.[2] Organizations such as the Coalition for Health AI are developing governance frameworks. Health care technology standards are evolving to support the movement of increasingly sophisticated clinical data.
CMS is now asking what happens on the payment side. The proposed rule also contains a separate request for information about CMS’s reliance on the AMA’s CPT coding system and the AMA/Specialty Society Relative Value Scale Update Committee (“RUC”) in defining and valuing physician services.[3]
Taken together, these questions suggest that CMS is beginning to confront a broader issue: What happens to the systems used to define and pay for medical services when technology changes the work involved in providing those services?
But before we get to the question about how Medicare should pay for AI-enabled care, there is a much more fundamental question: In what ways and where does AI actually make medicine better?
That answer will not be the same in every clinical setting. Consider a total hip replacement, a procedure that for 2027 is facing significantly lower reimbursements. Once the surgeon and patient have determined that surgery is appropriate, there are numerous decisions involving implant selection, sizing, positioning, anatomy, patient characteristics, and anticipated outcomes. An AI system could potentially analyze thousands of previous procedures and identify patterns that would be difficult for any individual surgeon to retain from experience alone.
That is a potentially valuable use of technology. The physician remains responsible for the decision, but the physician has access to a much larger body of information when making it.
Other applications may prove less valuable. An algorithm that simply reproduces existing clinical protocols or decisions may make a health care system more technologically sophisticated without making its decisions more effective, accurate, or meaningful.
This matters because the goal should be to identify those applications of AI that demonstrably improve clinical decision-making, patient outcomes, efficiency, or access to expertise, and then determine what payment and regulatory structures are necessary to support their responsible use.
What Should Physicians/Clinicians Be Asking?
The September 14 comment deadline provides an opportunity for physicians and other stakeholders to begin asking these questions directly. For example:
- What clinical value should be demonstrated before Medicare recognizes an AI-enabled service?
- How should physician judgment and responsibility be reflected when AI becomes part of a clinical service?
- How should payment policy account for the costs of validating, monitoring, and governing AI?
- How will the use of AI affect clinical education and workforce planning?
- And perhaps most importantly, what incentives will Medicare create as technology becomes capable of performing an increasing share of the work currently performed by clinicians?
These Questions Deserve the Attention and Perspective of the Physicians’ and Clinicians’ Voices.
CMS has opened the door by asking what technology means for payment in primary care. The health care community should use this opportunity to consider the much larger implications.
The deadline for comments is September 14, 2026. There is still time to influence the questions that will shape the next phase of technology-enabled medicine. Please contact the authors or the EBG attorney with whom you ordinarily work.
This blog post is the first in a series exploring the implications of AI as it becomes increasingly integrated into medicine. It also serves to continue the conversation from our inaugural Epstein Becker Green’s Digital Health Roundtable, and sets the stage for upcoming roundtable discussions, where we will continue to bring together health care leaders to consider what this transformation means for clinical practice, payment, workforce, governance, and accountability. There is much more to come as we explore what happens when AI becomes part of medicine.
* * * *
If you have questions, please reach out to the author(s).
The Health Law Advisor blog is currently edited by Emily Chi Fogler.
ENDNOTES
[1] See 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, 91 Fed. Reg. 43842, 43935 (proposed July 16, 2026).
[2] American Medical Association, CPT Appendix S: Taxonomy for Artificial Intelligence in Medical Services & Procedures (updated June 8, 2026). Appendix S classifies AI-enabled medical services and procedures as assistive, augmentative, or autonomous.
[3] See 91 Fed. Reg. 43842, 43952–56 (proposed July 16, 2026).
Authors
Senior Counsel