It is not hard to see very practical applications of AI that can improve lots of things in healthcare for “us.”
But this is a fascinating piece of how this is working out in real time – and helping the presently inexorable climb in healthcare costs continue.
It is also relevant in almost every other industrial application of AI – no one is standing still. So today’s headline beneficiary of AI usage is likely to have a presently subterraneanly equal and opposing force designing process to negate the headline.
Good fun.
Why AI Will Accelerate Health Care Inflation
David Brailer
June 11, 2026
AI will not determine whether American health care becomes more or less affordable. The payment system will determine that, as it always has. What AI changes is the speed and scale at which the payment system’s incentives express themselves.
Artificial intelligence (AI) has arrived in American health care with extraordinary promise: earlier diagnosis, accelerated drug discovery, reduced administrative burden, expanded access to care that geography and income have long rationed. Some of those possibilities are already becoming real. The technology has the long-term potential to reduce unnecessary care, compress diagnostic timelines, and route patients to the right level of treatment. It is also being used today for extraction: optimizing billing, automating denials, and inflating the volume and intensity of coding and claims.
People who pay for care are trying to understand the net impact of AI on health care costs. They are attempting to balance potential AI-driven care improvements against payment system gaming and extraction to see which will win out. That is the wrong frame since both care improvement and extraction are inflationary. The right frame is to understand how fast AI will drive these forces and compound them, on top of an inflation baseline that was already unsustainable. That points toward remedies unique to AI, ones that can be applied before the compounding turns into runaway inflation and outruns policy.
What Electronic Health Records Taught Us
I spent several years in Washington leading the federal government’s strategy to digitize American medicine. The promise was substantial and sincere: longitudinal patient records, clinical decision support, population health management, reduction in medical error. Those benefits were real. Electronic records reduced transcription errors, improved medication safety, and made clinical information available at the point of care in ways that genuinely helped patients. Better information also increased utilization. Research found that electronic access to lab and imaging results increased diagnostic testing by 40 to 70 percent at some institutions, as clinicians acted on information that paper records had obscured. Care improvement and cost reduction are not the same thing.
What we did not anticipate was the speed and scale of the billing optimization that followed. Hospitals and physician practices learned quickly that electronic systems made it easier to document more diagnoses per visit, carry prior notes forward, and select higher-intensity billing codes. A Health Affairs study found that changes in coding behavior during the EHR adoption period were associated with $14.6 billion in excess hospital payments across payers in 2019 alone, not from delivering more care but from documenting it differently. The clinical infrastructure we expected to emerge (outcome measures, care process loops, longitudinal feedback on what worked) was not built at scale, because the payment system never required it and never rewarded it. Beyond long term care improvement, what the era produced was the digitization of billing at scale.
The extraction economy of American health care does not resist useful technology. It recruits it. AI is being recruited the same way, and AI’s scale and speed is different from anything electronic records achieved. Understanding why requires understanding how AI scales and where, structurally, it performs best.
Read the full article.
