Healthcare reimbursement has become so complicated that efficiency can no longer be measured only by how quickly a claim moves through a system. In my view, the more important question is whether the process is understandable, accountable, and capable of handling the many exceptions that arise in real-world care. That matters because administrative friction can reach both providers and patients.
A 2026 KFF Health Tracking Poll found that 47 percent of insured adults said they had experienced a delay or denial involving a prescribed service, treatment, or medication during the previous two years. For people with chronic conditions, the share was even higher. Those numbers reinforce something I have seen repeatedly in revenue-cycle work: reimbursement is not simply a back-office issue.
When the process becomes difficult to navigate, the effects can extend into the delivery of care itself. The burden on medical practices is also significant. The American Medical Association reports that physicians and their staff complete about 40 prior authorizations each week and spend an average of 13 hours on them.
To me, that illustrates why the conversation about efficiency cannot stop at automation. Technology can reduce repetitive work, but it cannot automatically make a complicated system more transparent or accountable. I believe artificial intelligence has an important role in medical billing, particularly where tasks are predictable and rules are clearly defined.
Coding support, routine claim processing, eligibility checks, and other structured activities may all benefit from automation. But that’s the easy 80 percent. The remaining 20 percent is where billing stops being a data problem and becomes a human one (aka human-in-the-loop): appealing a denial that hinges on clinical nuance, negotiating with a payer with logic, or deciding how to handle a patient’s bill.
That 20 percent doesn't shrink as AI improves. It was never really about processing claims, but about exercising judgment on people’s behalf. The best billing operations don’t ask AI to replace that judgment; they use it to free up the time to exercise it well.
Extract — continue reading at the source.
