Health systems are now processing tens of thousands of prior authorization decisions per day through AI-powered workflows. The technology works. Payers and providers alike are racing to automate one of the most expensive friction points in American healthcare.
That's exactly why PE-backed platforms should slow down before they automate themselves into a corner.
Revenue cycle automation isn't new. What's new is the scale. Healthcare Finance News has documented the prior-auth arms race accelerating, with health systems stacking AI decision layers to keep pace with payer-side automation. The vendor pitch: remove human touchpoints, lower denial rates, improve cash flow.
It's an incomplete pitch.
EY's agentic AI token-cost analysis found that interaction costs in complex AI workflows jumped from $0.04 per interaction in 2023 to $1.20 in 2026, a 30x increase as systems added reasoning loops, tool calls, and iterative workflows. EY's framing was direct: "Optimizing tokens without understanding total cost of ownership is like managing a factory by watching the electricity bill." Governance, change management, exception handling, and risk exposure tend to swamp the headline savings.
For PE healthcare portfolios, that's a margin risk that won't show up in the initial pro forma.
The deeper issue is clinical judgment. Prior auth isn't a routing decision. It's a clinical determination layered with payer policy, patient history, and real-time context. When systems optimize for throughput over accuracy, denial rates go up, not down.
Providers reporting that more than 10% of their claims are denied went from 30% in 2022 to 41% in 2025. That tracks with the acceleration of automated prior-auth engines on both sides of the payer-provider table. Cedar's 2026 RCM report frames it well: the hardest parts of the revenue cycle to automate are the messy, dynamic, human parts that rule-based systems treat as edge cases. Those edge cases are where the money is.
What holds up in PE healthcare portfolios: platforms that use AI to surface decisions and route exceptions to humans, not ones trying to get humans out of the loop entirely.
The regulatory exposure alone is worth paying attention to. CMS and state payer mandates are tightening around automated prior authorization. A platform that can document human oversight goes into an audit, or an exit, with a cleaner posture.
On top of that: physicians reviewing flagged requests bring context models don't have. Human review at decision boundaries cuts downstream denial recovery costs faster than any labor savings justify skipping it. And providers that automate aggressively without human controls tend to escalate payer tensions, which affects contracting leverage, which affects rates. Rates are the actual value driver in most healthcare PE plays.
When I talk to ops partners inside healthcare portcos, the question usually framed as "how do we automate more" is better framed as "where do we keep humans in the loop, and how do we make those touchpoints faster?"
That changes vendor selection. It changes how you measure ROI: not by headcount removed, but by denial rate reduction and clean claim rate.
The platforms with durable margin three years out aren't the ones that went furthest removing human judgment from revenue cycle. They figured out where human judgment pays for itself and built the AI around that.