I went to the doctor recently. Nothing dramatic, a routine visit, the kind that should be a non-event. At the front desk, before I had even sat down, I was given a choice I did not expect. Pay out of pocket, and the visit would cost fifty dollars. Use my insurance, and the same visit, the same doctor, the same fifteen minutes, would cost two hundred and ninety-seven dollars.
Nobody offered an explanation. Nobody seemed to think one was necessary. I stood at that desk and understood, in a way that no policy paper had ever quite clarified for me, exactly what is wrong with this system. Not in the abstract. In the specific. The price of healthcare in America is not determined by what care costs. It is determined by who is paying for it, a calculation that has almost nothing to do with medicine and almost everything to do with the machinery surrounding it.
I have spent nearly two decades in that machinery. And the longer I have spent inside it, the more convinced I have become that the dysfunction in healthcare delivery and the dysfunction in group benefits distribution are not two separate problems. They are the same problem, running on parallel tracks, wearing different professional clothes. Same sickness. Different chart.
The Supplier Engine
Every few years, a major hospital system and a major insurance carrier go to war in public. The most recent iteration of this drama that most New Yorkers remember involves a major hospital network and its prolonged contract dispute with a major carrier over reimbursement rates. The pattern is always the same. The hospital system announces that the carrier’s proposed rates are inadequate to cover its costs. The carrier announces that the hospital’s demands are unsustainable and will drive up premiums for consumers. Both sides issue statements. The media covers the standoff. Patients are caught in the middle, uncertain whether their doctors will still be in network when their next appointment arrives.
What nobody asks, at least not loudly enough to receive an answer, is where the hospital’s costs are actually coming from. The supply side of the healthcare equation is a black box that makes the carrier underwriting process look transparent by comparison. Why does a bandage cost a hundred dollars to acquire and a thousand dollars to bill? Is it the supplier raising prices on medical equipment? Rising property taxes on the hospital campus? Administrative overhead ballooning year over year? The honest answer is that the public does not know, the carrier does not know, and in many cases the hospital’s own billing department could not produce a coherent explanation if asked.
What we do know is that hospital systems with strong brand recognition have learned to leverage that brand as a negotiating instrument. When a health system that carries genuine name equity in a community threatens to leave a network, it is not issuing a business ultimatum. It is issuing a social one. And the carrier, facing the prospect of explaining to hundreds of thousands of members why their preferred hospital is suddenly out of network, almost always blinks first. The brand is the moat. The patients are the hostages. The premium increase is the ransom, passed quietly to the employer, and then to the employee, and then forgotten until the next renewal cycle.
The Asymmetry
There is a law, or more precisely a regulatory framework, that governs how much of every premium dollar an insurance carrier is permitted to keep. The Medical Loss Ratio provision of the Affordable Care Act requires that carriers in the large group market spend at least eighty-five cents of every premium dollar on actual medical care. The remaining fifteen cents covers administration, overhead, and profit. Spend less than that on care, and the carrier owes its members a rebate.
It is, in isolation, a reasonable consumer protection. The problem is what sits on the other side of that equation and faces no equivalent constraint. Hospital systems, pharmaceutical manufacturers, and medical device suppliers operate with almost no regulatory ceiling on what they can charge. The government intervened precisely where the margin was most visible and most politically convenient, and left the underlying cost engine running without a governor. The result is a system where carriers are legally capped in what they can earn while the entities driving the actual cost increases are free to raise prices without explanation, year after year, negotiation after negotiation.
And then, when the carrier has no choice but to pass those costs to the employer in the form of a rate increase, the hospital issues a press release expressing concern for the patients who will be hurt by rising premiums. The audacity of it is almost architectural. They build the fire and then mourn the smoke.
Empathy, deployed strategically, is one of the most effective weapons in a negotiation. The hospital systems have understood this for a long time. Statements warning that a failure to agree on reimbursement rates will result in patients losing access to critical care are not public health announcements. They are leverage, calculated and deliberate, designed to make the carrier the villain in a story the hospital is writing. It works because it is difficult to argue against without appearing callous, and because the media, understandably, covers the human dimension rather than the actuarial one.
The Dead Ends
The standard political response to a broken system is to propose a bigger system to replace it. In healthcare, that proposal is single payer, a centralized government mechanism that eliminates the carrier as intermediary and pays providers directly from a public fund. The appeal is intuitive. Remove the profit motive, remove the administrative complexity, remove the negotiating theater, and what remains should be cleaner and cheaper.
What that argument does not address is the supply side. A single payer system changes who writes the check. It does not change what the hospital charges to cash it. The supplier inflation engine, the brand leverage, the opaque cost structure that nobody inside or outside the system can fully explain, all of it remains intact. The government becomes the largest insurance carrier in history, deploying taxpayer dollars like fingers in a dam, absorbing cost increases it has no more power to control than the carriers it replaced, and raising taxes on the same consumers it promised to protect. The problem does not disappear. It simply acquires a new administrator.
The fee-for-service model sits underneath all of it, undisturbed. Providers are still paid for volume. Hospitals are still incentivized to see more patients, order more tests, bill more codes. The transaction remains the unit of value, regardless of whether the transaction produces a better outcome for the person at the center of it. Changing the payer does not change the incentive. And incentives, in the end, are what the system runs on.
Climate, Not Weather
Here is something worth sitting with. Right now, if you go to see a doctor, you are expected to remember your own medical history. The medications you have taken, the specialists you have seen, the scans you have had, the dates of procedures that happened years ago in a different state with a different insurance card in your wallet. The system’s understanding of your health is whatever you can recall in the few minutes before the doctor walks in.
Records exist, technically. They are fragmented, disjointed, often handwritten, frequently inaccessible, and distributed across a lifetime of providers who have no shared infrastructure and no particular obligation to communicate with one another. The electronic health record was supposed to solve this. What it produced instead was a digital version of the same fragmentation, incompatible systems siloed by institution, generating structured data that nobody outside the institution can easily read.
I am thirty-six years old. I am generally healthy. I rarely interact with the healthcare system as a patient. And I cannot reliably reconstruct my own medical history without significant effort. Now consider a family physician with five hundred active patients, twenty years of practice, and a filing system that spans paper records, two different EHR platforms, and a memory that is doing its best. The idea that this physician can synthesize a patient’s full health narrative from the information available to her, and make genuinely expert recommendations based on it, is not a reasonable expectation. It is a fantasy we have collectively agreed to maintain because the alternative is too uncomfortable to name.
Health is climate, not weather. It is not the single visit, the isolated diagnosis, the test result reviewed in a fifteen-minute appointment. It is the accumulated pattern of decades of choices, exposures, conditions, and interventions that have led to a person’s present state. A system built around individual transactions cannot see that pattern. It can only see the moment it is billing for.
What AI Could Actually Do
There is no fix-everything button here. Anyone who tells you artificial intelligence will solve the fee-for-service model is either selling something or has not thought carefully about the incentive structures that keep the model in place. The outcomes-based compensation question, who gets credit when five physicians treated the same patient, how much is appropriate when the patient does not recover, what happens when a doctor does everything correctly and the biology does not cooperate, these are genuinely unsolved problems that no model can adjudicate. The ethical and contractual architecture required to answer them does not yet exist.
But the data infrastructure problem is not unsolved. Not anymore. The technology capable of ingesting a lifetime of fragmented health records, reconciling them into a coherent longitudinal narrative, and surfacing genuinely relevant clinical context at the moment a physician needs it, that technology exists in its early forms right now. AI is, at a fundamental level, better at processing large volumes of heterogeneous data than any human who has ever lived. the physician with forty years of patients and chicken-scratch notes from three states ago is not a failure of medicine. She is a failure of information architecture. And information architecture is exactly the problem that machine learning was built to address.
What becomes possible when a physician can see the climate instead of just the weather is a different kind of medicine. Not a perfect one. Not one that resolves every ethical question about how outcomes should be measured or compensated. But one where the script factory and the revolving patient door become visibly, measurably counterproductive, where the data makes the cost of volume-over-outcomes impossible to ignore for everyone in the room.
That visibility will not fix the supplier inflation engine on its own. It will not rewrite the MLR asymmetry or end the practice of using patient welfare as a bargaining chip in contract negotiations. But it will change what is knowable, and in a system that has survived this long on strategic opacity, that is not a small thing.
The Diagnosis
The fee-for-service model and the group benefits distribution model share a diagnosis. Both were built to process transactions rather than produce outcomes. Both use complexity as insulation against accountability. Both reward the wrong behavior so consistently, and for so long, that the wrong behavior has come to feel like the natural order of things. And both are now facing a technology that does not particularly care how long they have been doing it this way.
The chart has different labels. The sickness is the same.
AI will not cure it automatically. But it will, for the first time, make the true cost of the wrong incentive structure legible in real time, to physicians, to carriers, to employers, and eventually to the patients absorbing the bill for a system that was never designed with their outcomes in mind. Legibility is where accountability begins. And accountability, however uncomfortable, is the only thing that has ever actually changed a broken system.
Overnight Intelligence | Group Benefits in the Age of Machine | jtminogue.substack.com