Haroon Chaudhary, CEO of MedHeave, is cautioning medical practices against AI-driven claims processing after his own company's year-long AI development project failed during internal testing.

AI-powered billing is becoming a standard selling point across the revenue cycle management industry. Vendors market automated claim routing, predictive denial management, and AI-driven coding review as upgrades that reduce cost and speed up reimbursement.

Haroon Chaudhary built one of those systems. Then he watched it fail.

A project built with conviction

In late 2025, MedHeave began developing an internal AI system designed to route claims more efficiently, catch coding errors before submission, and improve reimbursement speed. The company committed senior staff, ran demonstrations for test practices, and treated the project as a strategic investment.

"I was personally invested in this," Chaudhary said. "We pulled in our strongest people. The demos performed well. We genuinely believed this would change how we operated."

The system was tested inside MedHeave's own operations using dummy practice data — a controlled environment with no live revenue at risk.

Where the system broke

Claims routed to incorrect payers. The error-detection module generated billing rules that did not appear in any payer's published guidelines and applied corrections based on those invented rules. New errors emerged from the system's attempt to fix old ones.
The most concerning pattern was not the errors themselves but their presentation. Every incorrect output looked carefully reasoned. The AI did not flag uncertainty or present alternative interpretations. It delivered fabricated rules with the same confidence as accurate ones.

"My first reaction was to blame the timeline or the dummy data," Chaudhary said. "Both were factors, but they were not the reason. The system was doing exactly what it was designed to do — generating confident answers. The problem was that confidence and accuracy are not the same thing."

A warning grounded in experience, not theory

Chaudhary's caution is directed at the broader industry, not just his own company. Practices evaluating medical billing solutions are increasingly being told that AI reduces denials, speeds payment, and lowers costs. Those claims may prove true in some applications.

But in claims processing — where an incorrect payer ID, a fabricated modifier rule, or a misapplied coding edit carries direct financial consequences — the cost of a confident error is higher than the cost of a human one.

"A human billing specialist who makes a coding mistake creates a pattern your QA team recognizes," Chaudhary said. "An AI that fabricates a rule creates something that looks like expertise. Those two failures require completely different detection systems, and most billing operations are only built for the first."

MedHeave continues to evaluate AI for internal analytics and payer behavior research but has removed it from any role that touches claim preparation or submission.

About MedHeave

MedHeave is a medical billing company that operates complete revenue departments for medical practices across the United States. The company delivers 14 integrated service areas — from insurance verification and coding through claim submission, denial management, and patient collections — inside each practice's own EHR and billing systems. MedHeave charges 4 to 7% of collections with no lock-in contracts.

For more information, visit https://medheave.com/.