Healthcare has a “convenience truth” problem.
The response to my last article “Healthcare AI is quietly entering a dangerous arms race” reinforced a question:
If payers and providers are both using AI, how do we make sure they pursue the same clinical truth—not simply get better at defending their positions?
There is another problem:
What happens when both sides have the evidence, but selectively emphasize what supports their position?
I call this “convenience truth.”
In my experience, I have seen individual clinical data points emphasized while the broader clinical context is overlooked.
Consider a patient with a UTI and an elevated WBC count.
The WBC may be cited as evidence of a more severe infection. But what if the patient is also receiving corticosteroids for an asthma exacerbation—something that can independently contribute to leukocytosis?
The laboratory value is real.
But is the interpretation complete?
The data point can be true while the conclusion drawn from it is wrong.
This is where healthcare AI has an enormous opportunity.
AI should not simply search for evidence that confirms a conclusion.
It should actively search for evidence that could disprove it.
What else could explain this finding? Is it actually attributable to the diagnosis? What evidence supports the conclusion—and what challenges it? What happened to the patient afterward?
Correlation is not causation.
Clinical guidelines have an important role, but clinical reality rarely fits neatly into a series of checkboxes.
The opportunity for AI is to move beyond evidence matching toward clinical reasoning.
Imagine payer AI and provider AI evaluating the same comprehensive clinical evidence, surfacing both supporting and contradictory information, making their reasoning transparent, and arriving at the most defensible interpretation of the patient’s clinical state.
The question changes from:
“How do I prove my side?”
to:
“What does the entire clinical record tell us?”
If we can get there, perhaps we need fewer disputes, fewer administrative layers, and fewer appeals—not because we eliminate disagreement, but because we make the underlying clinical reasoning better.
That, to me, is the real promise of AI in healthcare.
Not automating the argument.
Not making each side better at winning it.
Making the argument less necessary in the first place.
That is how we move from convenience truth to whole clinical truth.