The first time I heard a doctor describe an AI scribe, she made it sound almost magical. The tool sat quietly on her desk, she said, listening, taking notes, freeing her up to actually look her patients in the eye. Well, that was the pitch. What nobody mentioned at the time was that the same little tool, humming away in the corner of an exam room in suburban Pennsylvania, was also quietly nudging the cost of that visit upward.
That’s the strange contradiction sitting at the center of America’s AI health boom right now. Tech executives and insurance spokespeople have spent years promising that artificial intelligence would finally bend the cost curve in healthcare. Mario Schlosser of Oscar Health called AI the only realistic way to make doctor visits cheaper. McKinsey, back in 2024, floated a number so big — three hundred sixty billion in annual savings — that it almost felt designed to end the argument before it started. Two years on, the argument hasn’t ended. It’s just gotten messier.
| Key Information | Details |
|---|---|
| Topic | U.S. AI Health Boom and Payment Policy |
| Primary Source | Penn LDI (Leonard Davis Institute of Health Economics) |
| Featured Expert | Amol Navathe, MD, PhD — Senior Fellow, Penn LDI; Former Vice Chair, MedPAC |
| Date of Analysis | April 8, 2026 |
| Published In | Health Affairs perspective piece |
| Key Concern | AI scribes and ambient documentation inflating billing complexity |
| Industry Voice | Caroline Pearson, Executive Director, Peterson Health Technology Institute |
| McKinsey 2024 Forecast | Up to $360 billion in projected annual savings — largely unmet |
| Counter-Finding | AI tools currently driving costs upward, per hospital admins and insurers |
| Policy Body Referenced | Medicare Payment Advisory Commission (MedPAC) |
| Pricing Comparison Model | Prescription drug pricing framework |
Amol Navathe, a senior fellow at Penn LDI and former vice chair of MedPAC, has been raising the alarm in a recent Health Affairs piece. His worry isn’t AI itself. It’s that the American payment system, built around inputs like a clinician’s time and skill, simply wasn’t designed for software that scales infinitely at almost no marginal cost. Once you build the model, serving the millionth patient costs nearly nothing. That’s a beautiful thing economically, until it collides with a billing structure that still thinks in fifteen-minute increments.
More than any policy paper, the AI scribe story demonstrates the issue. Physicians who were burned out used to scribble the bare minimum: brief notes, straightforward codes, and modest billing. Every word, every casual reference to a UTI or a borderline blood pressure reading, is now recorded by an algorithm, which kindly recommends adding each one to the list of diagnoses. On paper, visits that were previously coded as simple appear complex. The reimbursement increases. Eventually, the patient makes the payment.
It seems like no one truly anticipated this. Ambient scribes are inflationary, according to Caroline Pearson of the Peterson Health Technology Institute. It’s an eye-catching statement from someone whose job it is to assess health technology. And when you recall the initial sales pitch, it hits harder.
According to Navathe, value-based payment, clinical benefit standards, and cost-sharing incentives may provide a partial blueprint for prescription drug pricing. It’s not a neat solution. In contrast to pills, AI updates and adapts. It seems almost philosophically challenging to set a fixed price for a tool that silently retrains itself every quarter.
It’s difficult to ignore the story’s recognizable structure as you watch this develop. A disruptive technology emerges. By taking more money from it, the incumbents adjust. The cost is borne by the patients. It is still genuinely unclear if lawmakers will act quickly enough to change the regulations or if they will allow the market to improvise its way into another ten years of rising premiums. The boom is genuine. So far, the savings are not.
