The AI Billing Arms Race: Dr. Alex Sheppert on Denial Inflation

A family physician finishes a 15-minute visit with a patient managing diabetes. An AI scribe generates the note. The note goes to the payer for billing. Seconds later, a different AI reads that same note and denies the claim.

That sequence is now happening millions of times a day across US healthcare. In a commentary published in the Journal of Healthcare Finance, Dr. Alex Sheppert, Founder and CTO of Matic, argues that most organizations are still looking at it the wrong way.

A Co-Evolutionary Cycle

AI scribes and coding assistants on the provider side and automated adjudication on the payer side are usually discussed as separate developments. Dr. Sheppert’s central point is that they are better understood as a single, interdependent system in which every advance on one side prompts a response from the other.

The cycle begins when a payer deploys a model to enforce a specific documentation requirement, for example the criteria for a level-4 office visit. As that rule is applied at scale, some claims are paid and others are denied. Provider-side AI then learns from those denials, and subsequent notes arrive with the missing elements already in place, including clearer time documentation and explicit medical necessity language. As notes become harder to deny, payers refine their algorithms to enforce stricter interpretations, and the cycle begins again.

Denial Inflation

The piece introduces a term worth adding to your vocabulary: denial inflation. When the marginal cost of denying a claim drops to near zero, payers can afford to challenge claims that were never economically worth reviewing before.

The commentary points to documented examples. A 2023 ProPublica investigation reported that Cigna physicians used an automated system to deny more than 300,000 claims in two months, spending roughly 1.2 seconds on each. A later ProPublica report described a prior-authorization vendor whose adjustable algorithm could reportedly raise denial rates on demand.

Even modest movement matters at this scale. A rise in denial rates from roughly 9% to somewhere between 12% and 15% carries serious financial consequences in a sector that makes up nearly a fifth of US GDP.

The Automation Divide

The burden of this arms race is not evenly shared. Large health systems can invest in AI scribes, coding tools, and denial prediction. Smaller practices, rural clinics, and safety-net providers often cannot, and they are the groups least able to absorb the revenue loss.

Dr. Sheppert frames the consequence sharply: documentation variability starts to act as a social determinant of payment. Higher first-pass denials, more staff hours spent on appeals, slower revenue cycles, and added pressure to consolidate all follow.

Where This Ends, and What Helps

The commentary lays out two possible endpoints. In one, provider AI gets so good at anticipating payer rules that claims become effectively non-deniable (a “perfect note” system). In the other, payers and providers license the same small set of platforms, which could cut friction but also concentrate market power and create conflicts of interest.

An efficient endpoint would save billions in administrative cost. The risk is the transition, and the practices that don’t survive it.

His proposed response is practical: require automated denials to state which documentation element failed and which rule was applied. For an AI adjudication system, generating that explanation is technically trivial. It would speed provider adaptation, reduce unnecessary appeals, and narrow the information gap built into automated decisions today.

The Takeaway for Physician Leaders

If your organization is evaluating AI documentation tools, the question is no longer only whether they save clinician time. It’s how your documentation will perform against payer algorithms that are adapting just as fast, and whether your group has the visibility to see why claims are being denied.

Read the full commentary: Sheppert A. “Bidirectional AI in Claims and Documentation: Co-evolution, Denial Inflation, and the Case for Transparency.” Journal of Healthcare Finance, Fall 2026, Vol. 50, No. 1 (Commentary & Perspective). https://journalofhealthcarefinance.com/index.php/jhf/article/view/301