Healthcare AI is quietly entering a dangerous arms race.
We are building upcoding engines for providers and denial machines for payers—and calling it “innovation.”
Here is a classic example of where healthcare needs to move beyond payer vs. provider.
A provider meticulously documents “severe protein-calorie malnutrition.” On the surface, that may be a completely appropriate and accurate diagnosis.
But here is the problem: he same diagnosis can represent a very wide clinical spectrum.
One patient may require enteral tube feeding and intensive nutritional intervention. Another may receive a multivitamin, an appetite stimulant, and oral nutritional supplements such as Ensure—with little meaningful change in the overall intensity of care.
Now add a large reimbursement differential to the DRG.
The provider has every reason to defend the diagnosis and the associated reimbursement: “It is clinically documented, and it is the right diagnosis.”
The payer has every reason to challenge or downgrade it: “The clinical care and resource utilization don’t appear materially different.”
Both sides can make a legitimate argument.
This is exactly where AI should add value.
The objective shouldn’t be to build an AI product that is payer-centric or provider-centric. It should be clinically centric.
The product should rigorously evaluate the totality of the clinical evidence—diagnosis, severity, treatment intensity, resource utilization, trajectory, and outcomes—and arrive at the most defensible representation of the patient’s actual clinical state.
Not what maximizes reimbursement.
Not what minimizes reimbursement.
What is clinically true.
That is the standard both sides ultimately have to acknowledge.
Unfortunately, at the sales level, the conversation often starts with ROI. And when ROI becomes the primary product pitch, we lose the plot. We trade long-term administrative efficiency for short-term margin manipulation, creating an algorithmic arms race where no one actually wins.
The real opportunity in healthcare AI is harder—and far more valuable:
Build systems that neither side can easily bend toward its own financial incentive because the clinical reasoning is rigorous, transparent, and defensible.
That is where I believe the next generation of healthcare AI needs to go.
How do we break the payer-provider AI arms race?