Every health tech conference this year will feature a panel titled something like “AI and the Future of Prior Authorization: Toward a Collaborative Care Ecosystem.” The speakers will use the words streamline, align, and partnership. A payer executive will sit next to a provider executive and agree, on stage, that we are entering a new era of cooperation.
It is theater. And the cost of the theater is that we keep building the wrong things, regulating the wrong things, and reassuring patients with the wrong story.
The American Medical Association’s latest physician survey found that 60% of doctors expect AI to increase prior authorization denials. They are right to worry. But the conventional explanation that AI will simply automate today’s denial machine understates the problem.
Prior authorization is a fight, and it’s always been a fight. A payer is trying to control spending; a provider is trying to deliver care and get paid for it; a patient is trying not to die or go bankrupt while the first two argue. These interests do not align, and no amount of pleasant language about shared goals makes them align. Anyone telling you otherwise is selling something, usually software.
The honest question is not whether AI will end the fight. It will not. The honest question is what AI does to a fight that has been structurally imbalanced for 30 years.
Here is what is actually about to happen.
Payers are deploying AI faster than providers, with more data, more capital, and more legal infrastructure. The largest national plans process more prior auth decisions in a week than most health systems see in a year. That asymmetry will only compound in the AI era. Recent litigation against major insurers over algorithmic denial practices is a preview of what scaled, automated utilization management looks like when the speed of denial outpaces the speed of appeal. When a payer can generate a defensible-looking denial in milliseconds and the provider’s appeal takes thirty days, AI has industrialized the imbalance.
Meanwhile, the provider side is being sold AI-powered prior authorization automation that promises to fight back. Some of it will work, and most of it will produce cleaner submissions that get denied more efficiently. The arms race begins, and like every adversarial AI arms race before it, in high-frequency trading, in programmatic advertising, in search optimization, it will not converge on truth. It will escalate. The cost of the escalation will be passed to premiums and to providers, which means to patients.
And the patients, let us be clear, have no AI. They have a phone tree, a deductible, and a disease.
The AMA’s data on what this system already costs is worth pausing on. More than one in four physicians report a prior authorization process has led to a serious adverse event for a patient in their care. One in three say PA criteria are rarely or never evidence-based. Ninety-five percent report care delays. This is the system AI is about to be layered onto.
So why is anyone optimistic? Because there is a version of this future that is genuinely better than the present, and it depends on getting one thing right.
The thing that has to be right is the evidence.
Today, prior authorization is an argument conducted between two parties who do not share a record. The payer has claims data and policy. The provider has notes, labs, imaging, and the patient in front of them. Nobody has a clean, complete, structured view of what actually happened to the patient. So both sides argue from partial information, and the side with more resources and more leverage wins more often. The patient loses in the gap.
If AI’s contribution to prior authorization is faster denials and faster appeals on top of the same broken evidence base, we will have built a more efficient version of a system physicians describe as moral injury. That is not rhetoric; it is the clinical literature. If AI’s contribution is finally producing a structured, complete, defensible clinical record that both sides have to reckon with, the calculus changes. Not because the fight ends. Because the fight finally happens on the merits.
This is a smaller claim than the one being marketed. It is also the only one that survives contact with reality.
The structural conflict between payers and providers is permanent. The asymmetry of resources is real and growing. The patient’s position in the middle is structurally weak. None of these facts can be solved by software, and any vendor who tells you otherwise is overselling.
What can be solved is the evidence problem. When the payer says the patient did not fail prior therapy and the provider says they did, the record should settle it. When the payer’s policy requires documented progression and the provider has it buried in three PDFs and a scanned fax, the record should produce it. When the appeal goes to an independent reviewer, the reviewer should not be making a judgment call on missing data. They should be reading a complete clinical picture and applying policy to it.
That is the actual prize. Not the elimination of prior authorization. Not the conversion of payers and providers into partners holding hands at HIMSS. The prize is a system in which the fights that remain are fights worth having, conducted on shared evidence, with a clock and a referee.
So here is what should happen instead of another collaboration panel. Regulators should stop being seduced by promises of AI-driven harmony and start demanding AI-driven transparency: every prior authorization decision carrying its reasoning, its evidence, and its policy citation in a form that can be audited at scale. Health plans whose AI generates denials that systematically disappear under independent review should face consequences proportional to the volume. And providers should stop buying tools that promise to beat payer AI and start investing in the clinical data infrastructure that makes their case undeniable in the first place. The winning move is not a faster fax. It is a record so complete and so structured that the argument is over before it starts on the cases that should never have been arguments.
Patients deserve the truth, which is that the system is adversarial, that AI is not going to make it loving, and that the only durable protection is a clinical record that follows them and tells their story accurately to whoever is deciding what gets paid for.
The future of prior authorization is not collaboration. It is contested, evidence-based, auditable adjudication, conducted at the speed of software, with the patient’s actual clinical reality finally in the room.
That is worth fighting for. The kumbaya version is not.
Photo: Piotrekswat, Getty Images
Mika Newton is the CEO of xCures, an AI-assisted platform that automatically retrieves and structures medical records from any US care site. He holds over twenty-five years of leadership experience in the life sciences.
This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.
