Where Revenue Cycle AI Earns Its Keep and Where Hype Runs Out

A veteran revenue cycle leader sees real AI gains in verification and coding, but warns that autonomy is overstated, governance is improvised, and prompt skill, not tools, decides results.

One revenue cycle director with 18 years in healthcare billing offers a grounded view of where AI helps and where it does not. He sees measurable value in insurance verification and coding, but insists autonomous coding tops out near 90 percent and demands a permanent human exception team. Denial follow-up and appeals, he says, are not ready because they need detective work machines cannot yet do. His organization has no direct AI governance. Instead it relies on HIPAA compliance, keeping patient data out of public tools and using a build-template-then-load-data workaround that depends on individual discipline. He names prompt-writing skill as the true constraint, willing to spend hours on one prompt if productivity rises. Weekly adoption is only 20 to 25 percent, split sharply by age, with some peers still on paper even as the CFO pushes AI in every weekly call. From consulting, he warns that vendors oversell, having sold organizations three or four unnecessary tools on promises of full automation. His wish: an honest, published account of AI's real, bounded benefits. For executives, the lesson is to buy for specific tasks, train people to ask well, codify data handling into policy, and ignore claims of do-everything automation.