The prior authorization arms race is accelerating. Health systems are deploying AI to pre-empt denials, auto-generate appeals, and predict which claims will be flagged. Payers, in turn, are deploying AI to issue denials faster, identify outlier billing patterns, and automate coverage determinations.
Both sides call this “innovation.” It isn’t. It’s an adversarial equilibrium where the only winner is the compute bill.
The Equilibrium Trap
When both sides of a transaction invest in AI that cancels out the other side’s AI, the net effect on patient care is zero. The net effect on cost is negative — both sides are now paying for infrastructure that produces no incremental value.
This is the classic red queen problem: you have to run faster just to stay in place. Except in healthcare, the patient is the one standing still while the algorithms sprint past.
What Actually Reduces Denials
The denial rate isn’t a technology problem. It’s a structural problem. The vast majority of denials stem from three root causes:
Documentation gaps — the clinical note doesn’t contain the specific language the payer’s criteria require. This is a workflow problem, not an appeal problem.
Criteria mismatch — the treatment is clinically appropriate but doesn’t meet the payer’s published coverage criteria. This is a policy problem.
Administrative errors — wrong codes, missing referrals, expired authorizations. This is a process problem.
AI that fights denials after they happen addresses none of these root causes. AI that prevents the conditions that create denials addresses all of them.
The Builder’s Opportunity
The real opportunity isn’t in the arms race. It’s in building the intelligence layer that sits between the clinical decision and the administrative submission — ensuring that every order, every referral, every treatment plan is documented, coded, and submitted in a way that makes denial structurally unlikely.
This is what an agentic patient advocate should do. Not fight the system. Rewire it.