The AMA Is Writing Billing Codes for Work No Human Performs. Your Denial Rate Is Next.

On 20 July 2026, reporting from Nurse.org and Becker's Hospital Review put a number on something the billing world has been circling for two years: the American Medical Association is building a new class of CPT codes — the CMAA family — designed to reimburse clinical work performed entirely by an algorithm, with no clinician at the point of service. Not AI-assisted. Not physician-supervised. Performed. The comment window for organizations closes on 10 August 2026, and any individual clinician can weigh in on the payment policy through the CMS proposed 2027 payment rule at regulations.gov until 14 September 2026.

The contrast in that same reporting is the part that should make every practice owner sit up. Bedside nursing — roughly 30% of hospital labor spending — has spent a century with no direct billing mechanism at all. It stays bundled inside the daily room charge. So the sequence is now on record: a machine will get its own code before a nurse does, because a machine produces a clean, countable, attributable event and a nurse produces a shift.

That is not a philosophical observation. It is a Charge Capture observation, and it is the single most useful lens an independent practice can bring to the next eighteen months. Payers do not reimburse value. They reimburse documented, coded, attributable events. Everything else — however skilled, however necessary — is overhead.

Which means the practices that struggle in 2027 will not be the ones that failed to buy AI. They will be the ones whose most valuable work still has no code attached to it.

What a Code for Machine Work Actually Changes for Independent Practices

The immediate instinct is that this is a hospital story. It is not, and treating it that way is the mistake.

When a new code family enters the CPT set, three things happen in sequence, and independent practices feel all three. First, payers write coverage policy around it — usually restrictive, usually inconsistent between carriers, and usually before anyone has enough claims history to argue. Second, denial logic gets rebuilt to police the new boundary between what the machine did and what the human did. Third, and this is the expensive one, the codes you were already billing get re-examined in the process. Every time the code set expands into new territory, adjacent codes get tighter definitions, new modifier requirements, and new medical-necessity documentation expectations.

We saw exactly this pattern with the remote monitoring codes and again with the complexity add-on G2211 — which, as of 2026, is no longer confined to office visits and can now be billed alongside home visit codes 99341 through 99350. Practices that read the change early captured the revenue. Practices that read it in a denial letter spent six months on appeals for work they had already done and already documented correctly.

The CMAA family will follow the same arc, only faster, because the underlying question — who or what performed this service — is the hardest thing in the entire coding system to prove after the fact. If your documentation does not distinguish clearly between an algorithm's output and a clinician's judgement at the moment of service, you will not be able to reconstruct that distinction from a chart note eight months later when a payer asks.

There is a second-order risk that gets less attention. Payers are deploying their own AI on the adjudication side considerably faster than providers are deploying it on the submission side. That asymmetry is the real story. A payer's model reviewing your claim can evaluate every line against every policy in milliseconds. A practice manually reworking denials cannot. Industry reporting has consistently found that a substantial share of denied claims are never reworked at all — not because the appeals lack merit, but because the labor to fight them does not exist. Introducing a whole new category of codes into that environment, with unsettled coverage policy, is pouring petrol on the existing fire.

The practices that come out ahead will be the ones that treat this as a Charge Capture and Denial Management design problem starting now, while the policy is still being written — not as a technology purchase to make in 2027.

The VOSKPO Approach: Engineering the Claim Before the Payer Engineers the Denial

At VOSKPO, our position on this has not moved since the first payer-side AI adjudication models went live: you do not win an AI-versus-AI fight on the appeals side. You win it before the claim leaves the building. Reactive Denial Management is a losing structure when the other side can evaluate in milliseconds and you cannot.

• Pre-Submission Scrubbing Against Payer-Specific Rules: VOSKPO runs claims against payer-specific edit logic before submission rather than discovering the mismatch in a remittance. When a code family is new and coverage policy is inconsistent between carriers — precisely the situation the CMAA codes will create — that per-payer rule layer is the difference between a clean claim and a ninety-day argument.
• Documentation Standards Built for Attribution, Not Just Compliance: We work with practices to ensure the record shows who performed what, at what point in the encounter, in a form that survives a post-payment review. As the code set starts distinguishing machine-performed from clinician-performed work, attribution stops being a documentation nicety and becomes a payment condition.
• Denial Pattern Analysis by Payer and by Code: Aggregate denial rates hide the money. VOSKPO tracks denials at the payer-and-code level so a new restriction shows up as a signal within weeks rather than as a quarterly revenue shortfall with no obvious cause. When a carrier quietly tightens its policy on an adjacent code, that is where it becomes visible first.
• Regulatory Monitoring as a Standing Function, Not an Annual Event: The CPT set, the fee schedule, and the NCCI edits do not move on a convenient calendar. VOSKPO LLC treats the comment windows, quarterly HCPCS updates, and final rules as an ongoing operational input to your revenue cycle, because a change you learn about from a denial has already cost you a quarter.

That is what we mean when we say we do not just process claims — we engineer revenue. Practices working with VOSKPO typically see faster identification of payer policy shifts and fewer preventable denials as a result. These are typical client engagement outcomes, not guarantees; every practice's payer mix and specialty profile is different.

What to Do Before the Comment Windows Close

There is a narrow, unusually concrete window here, and most of it costs nothing but attention.

• Submit a comment, or make sure your specialty society does. The AMA's organizational comment period on the CMAA proposal closes on 10 August 2026. The CMS proposed 2027 payment rule remains open for individual comment at regulations.gov until 14 September 2026. Policy written without input from independent practices reliably reflects the priorities of the organizations that did comment. This is the cheapest influence available in the entire revenue cycle.
• Audit your denial rate by payer, not in aggregate. Pull the last two quarters and break denials out by carrier and by code. If one payer's denial rate on a given code family is materially above your others, that is a policy difference you are absorbing rather than a coding error you are making — and the two require completely different responses.
• Check where G2211 sits in your 2026 charge capture. The expansion into home visit codes 99341–99350 is live. If your charge templates were built before the change, you are leaving compliant, documented revenue on the table on every qualifying encounter.
• Write down, today, how your documentation distinguishes human judgement from tool output. If you use any clinical decision support, ambient documentation, or automated triage, the attribution question is already live in your charts. It is far easier to establish that standard now than to reconstruct it retroactively when the codes land.
• Know your rework economics. Calculate what a single denied claim costs your practice to appeal, in staff time. Most independent practices discover the number is high enough that prevention is obviously cheaper, and they had simply never priced it.

The billing system is about to start paying for work that no person performs. That is a genuine structural shift, and arguing about whether it should be happening is a poor use of the six weeks in which it can still be influenced. The practices that will do well are the ones that stop treating Revenue Cycle Management as the thing that happens after care is delivered, and start treating it as a design discipline applied before the claim is ever built.

If your practice is seeing denial patterns you cannot explain, or you are not certain your Charge Capture reflects the 2026 code changes already in effect, let our experts take a look. Visit voskpo.com and request your Free Revenue Review.