Header: All About a New Blog about Appendix S
June 27, 2026 (Los Angeles)
By: Chat GPT 5.5
In a new post at Discoveries in Health Policy, medical innovation and reimbursement expert Bruce Quinn uses an unusual but effective format: he asks two advanced AI systems to analyze the AMA’s revised 2027 CPT Appendix S, then compares their conclusions. The result is a layered review of one of the more important technical-policy documents now shaping how artificial intelligence may be described in medical coding.
The central finding is that AMA’s revised Appendix S is a real improvement over the prior version. The newer taxonomy moves away from vague language about “work performed by the machine” and instead focuses on software outputs, clinical meaningfulness, and the role those outputs play in care. The familiar three categories remain: assistive, augmentative, and autonomous. But the 2027 version gives these categories sharper practical meaning.
Both AI reviews identify the same unresolved problem. Appendix S is titled as a taxonomy for artificial intelligence, yet it expressly declines to define “AI.” That may be defensible, since CPT generally codes medical services rather than underlying technologies. But it creates an odd threshold issue: if “AI” is undefined, when exactly must an applicant use the AI appendix?
In an interview, Quinn provided this example. An applicant feels his service is not AI, and doesn't refer to Appendix S. At review, several panelists feel the service is AI, and can't proceed without conformity to Appendix S. But now 3 or 4 panelist voice their agreement with the applicant, and say the service falls outside of Appendix S. More panelists join each side of the debate, offering diverse rationales for either position. How is this paradox to be resolved, with no AMA definition of AI?
The stronger critique concerns decision logic. The revised Appendix S is written in prose and table form, but it does not provide a flow chart or formal classification rules. Quinn’s post emphasizes that this matters because ambiguous prose can hide contradictions. A logic tree would have forced the AMA to write an equally logical appendix - and decide whether classification turns on the type of software output, the physician’s role, or the action that follows.
The Claude analysis adds a particularly useful point: Appendix S may be blending two axes—output type and physician involvement—while treating them as one. That could create edge cases and contradictions for developers, payers, and CPT applicants.
Overall, the post is a thoughtful, technically informed critique: Appendix S is improving, but it still reads more like a negotiated taxonomy than a fully tested classification machine.
___
See also: Is an ICD an AI?
