What is prior authorization, and can AI actually fix it?

In US healthcare, "prior authorization" is the process that requires an insurer to approve a treatment or medication before a doctor can proceed with it. It's officially meant to prevent unnecessary or excessively costly treatment, but the process has long been one of the most complained-about bureaucratic bottlenecks for patients and physicians alike — sometimes delaying the start of essential treatment by weeks.

Traditionally, prior authorization meant a doctor's office staff filling out forms and faxing or uploading them through a portal to the insurer, who would then have an employee manually review the request. The process was slow and inconsistent — the same treatment request could be judged differently depending on which employee reviewed it.

Now both insurers and health systems are deploying AI on opposite ends of the same process. On the insurer's side, AI quickly screens incoming requests and approves or denies them against predefined criteria. On the hospital's side, AI helps doctors phrase their requests in the way most likely to match what the insurer's criteria are looking for.

That sets up an odd dynamic: both sides are now using AI against each other. Critics say this could create an "arms race" loop — as an insurer's AI grows stricter, hospitals' AI grows more aggressive at tailoring requests to fit those exact criteria.

Supporters argue automation can genuinely speed things up, at least for simple, routine cases. Rather than a human employee spending hours on a clear, standard treatment request, AI can process it in seconds — lowering the insurer's operating cost while potentially shortening the patient's wait.

But concerns concentrate on complex or borderline cases in particular. Whether an AI model can make the right call for a rare condition or an unusual combination of treatments remains uncertain. Patient advocacy groups worry that automatically denied requests, without human review, could delay patients' access to care.

Regulators are trying to close that gap: several US states now require that any denial issued by an AI system also be confirmed by a human reviewer. The goal is to capture the speed of automation while ensuring critical decisions aren't left entirely to an algorithm.

Health economists stress that the root of the problem isn't technology at all, but incentive structure: insurers are motivated to lower costs, hospitals to raise revenue. AI doesn't remove that underlying conflict of interest — it just lets both sides run it faster.

Some experts argue the real fix is political, not technological: narrowing the list of treatments that require prior authorization in the first place, or fully standardizing the process, could offer a far more durable solution than AI automation ever will.

For now, the question of whether AI will actually "fix" prior authorization remains genuinely open. The technology has the potential to make the process both faster and more adversarial at the same time — and which outcome ultimately wins out depends largely on how regulators choose to balance those two forces.