Meet Clara. She is sitting in a sterile examination room, staring at a beige wall while holding a folded piece of paper that feels heavier than lead. It is an explanation of benefits, though there is nothing beneficial about it. A procedure her cardiologist ordered to check an irregular heartbeat has been denied. Not by her doctor. Not by a human being who looked into her chart, listened to her history of early-morning chest flutters, or considered the quiet terror that wakes her at three in the morning.
It was denied by an algorithm. For an alternative view, read: this related article.
Across town, inside the gleaming, glass-walled administrative hub of a major hospital system, another machine is humming. This one belongs to the provider. It is an artificial intelligence trained to review patient charts, scour medical codes, and ensure that every possible diagnostic justification is wrung from the electronic health record to maximize reimbursement. It is fighting for the hospital's bottom line with the cold, unblinking efficiency of a high-frequency trading desk.
Two silicon minds are now locked in perpetual combat over Clara's chest. And the most expensive part of modern American medicine is not the scalpel, the MRI machine, or the pharmacology. It is the war between them. Similar analysis on the subject has been provided by CDC.
The Cost of Silicon Friction
We talk about healthcare spending as if it were a weather pattern—an inevitable storm of aging populations, rising drug costs, and bloated administrative bureaucracy. We point fingers at pharmaceutical executives and insurance boardrooms. But we are missing the structural transformation happening beneath the floorboards.
Automation was supposed to save us. For years, the promise of software was simple: efficiency. Computers would take the friction out of billing, eliminate human error, and clear the administrative cholesterol clogging our clinics. Algorithms would process claims in milliseconds, freeing doctors to practice medicine instead of drowning in paperwork.
Instead, we built an arms race.
Consider how the cycle begins. A patient visits a physician with persistent abdominal pain. The doctor recommends an advanced diagnostic scan. In the past, this might have required a brief phone call or a faxed form. Today, the hospital’s revenue-cycle management software—powered by predictive models—scans the patient's electronic health record. It anticipates what the insurer’s automated denial engine is looking for. The software automatically injects specific buzzwords and clinical modifiers into the billing codes to preemptively satisfy the insurer's algorithmic filters.
It is a digital game of poker where both sides can see each other's cards, yet the stakes are paid in human suffering and billions of dollars in overhead.
The insurer’s algorithm, trained on millions of historical claims, scans the incoming request for patterns that justify denial or downgrade. It looks for minor discrepancies in symptom duration, alternative treatments not yet tried, or rigid policy thresholds. When it finds a match, it flags the claim for rejection—often without a human clinician ever laying eyes on the file.
Faced with this automated rejection, the hospital does not simply accept defeat. It deploys its own army of appeal-generating bots. These systems parse the insurer’s rejection codes, generate targeted rebuttals, and resubmit the paperwork with additional documentation.
Round and round it goes. Silicon generating friction. Silicon answering friction. And every cycle of this digital tug-of-war inflates the administrative overhead of the American healthcare system to staggering heights. Administrative costs now consume a quarter of every healthcare dollar spent in the United States, creating an expensive shadow industry that produces no healing, no comfort, and no health.
The Anatomy of an Algorithmic Denial
To understand why this is driving medical costs higher, you have to look inside the black box of decision-making.
Human judgment is messy. It relies on nuance, context, and empathy. A doctor looking at Clara might notice the subtle tremor in her hands, the strain in her voice, and the family history of cardiovascular disease that does not fit neatly into a standardized checkbox. A human reviewer can weigh the probability of a rare condition against the cost of a test and decide that caution is warranted.
Algorithms do not do nuance. They do probability based on past averages.
When an insurer deploys an automated decision-support tool to evaluate medical necessity, it evaluates Clara not as a living, breathing person, but as a vector of risk and cost. If historical data suggests that patients with her exact symptom profile only develop severe complications in two percent of cases, the algorithm may classify the diagnostic scan as "not medically necessary."
Never mind that Clara belongs to that two percent. Never mind that the cost of catching a condition early is a fraction of the cost of treating an acute cardiac event. The algorithm is optimized for short-term financial containment, not long-term health outcomes.
And this is where the economic feedback loop turns vicious. Because insurers use these automated filters, hospitals cannot afford to take chances. They must invest millions in proprietary software designed to outsmart the insurers. They hire armies of coders, compliance officers, and revenue integrity specialists whose entire job is to feed the hospital's AI with the right inputs to bypass the insurer's AI.
We have created an entire economy dedicated to administrative warfare. The cost of running these competing systems—licensing fees, cloud computing infrastructure, cybersecurity, specialized consulting firms, and internal compliance teams—gets baked directly into the cost of care. Clara’s MRI doesn’t cost thousands of dollars because the magnets are made of gold. It costs thousands of dollars because the bill has to subsidize the digital bureaucracy required to get the insurer to pay for it.
The Human Toll of the Machine Standoff
Numbers on a macroeconomic balance sheet rarely capture the visceral reality of how this system feels to the people caught inside it.
When a claim is denied by an algorithm, the burden of proof shifts entirely to the patient and their doctor. Clara does not receive a polite explanation. She receives a letter written in dense, bureaucratic legalese informing her that her coverage has been withheld. If she wants the procedure, she can pay out of pocket—an impossible sum—or she can wait while her doctor's office appeals.
Waiting, in medicine, is rarely neutral. Waiting is a verb that carries physical consequences.
During the weeks it takes for the hospital's appeals bot to cycle through the insurer's review queue, Clara's anxiety spikes. The chest flutters grow more frequent. She stops sleeping altogether, lying awake in the dark wondering if the tightness in her chest is stress or something failing inside her. Her blood pressure rises, ironically worsening the exact cardiovascular markers the denied test was meant to evaluate.
This is the hidden tax of the hospital-insurer AI standoff. It commodifies peace of mind. It replaces the traditional fiduciary trust between patient and physician with an adversarial digital courtroom where the patient is presumed guilty of wanting unnecessary care until proven otherwise by computational force.
Worse still, these automated systems are plagued by biases inherited from their training data. Studies have repeatedly shown that predictive algorithms used in healthcare often under-allocate resources to minority populations because historical spending data reflects systemic inequities in access to care. If past patients of color received fewer diagnostic tests because of socioeconomic barriers, the algorithm learns that they need fewer tests today, codifying historical discrimination into a veneer of mathematical objectivity.
Breaking the Cycle
We are standing at a dangerous inflection point. If we continue down this path, the administrative apparatus of medicine will soon consume the clinical core entirely. We will have the most sophisticated, computationally advanced billing machinery in human history attached to a collapsing delivery of actual healing arts.
Reversing this trend requires stripping away the digital armor that both sides have erected.
It starts with transparency and regulatory alignment. Insurers should not be allowed to use opaque, automated algorithms to issue wholesale denials without mandatory human clinical review. The burden of proof for denying care should be substantially higher than the burden of requesting it. If a machine can be used to streamline approvals, it should be used to expand access, not restrict it.
At the same time, hospital systems must step back from the arms race of aggressive revenue capture. When hospitals view every patient encounter through the lens of maximum code extraction, they invite the very algorithmic pushback from insurers that they complain about. Simplifying billing structures, reducing administrative bloat, and returning clinical decision-making to the examination room are the only real antidotes to the software-driven inflation strangling our health system.
Clara eventually got her scan. After three rounds of appeals, a human medical director at the insurance company finally looked at her file, overrode the initial algorithmic denial, and approved the procedure.
The scan revealed a minor arrhythmia—easily manageable with medication and lifestyle adjustments. She is going to be fine.
But as she walks out of the clinic into the late afternoon air, holding the revised billing statement, the silence of the street feels different than it used9 to. Somewhere out there, across thousands of miles of fiber-optic cable, two servers are still humming, calculating, and anticipating the next encounter. And the bill for their war is already waiting in the mail.
