Avoid Doubling Your Workload by Using Coding-Aware AI
Question: What’s the difference between coding-naive and coding-aware artificial intelligence (AI)? Revenue Cycle Insider Subscriber Answer: The primary difference between coding-naive and coding-aware AI is how it interacts with your electronic health record (EHR) software and coding software. An AI that is unsophisticated or untrained in medical coding will look at a medical record and choose a code related to what it initially sees while also producing insufficient documentation. Coding-naive example: A diabetic patient is seen for a bandaged wound on their left foot. The physician performs wound care multiple times a week and continues to monitor the healing process. A coding-naive AI might suggest E11.621 (Type 2 diabetes mellitus with foot ulcer) as the code to report and create a generic encounter note that is short on details. “However, through the lens of a coder, you would say there are actually two separate conditions that we really should be talking about,” said William Morris, MD, MBA, during the “AI in the Exam Room: Transforming Physician Documentation with Generative Tools” panel session at AAPC’s DOCUCON 2025. A coding-aware AI will factor in the guidelines and coding instructions, and generate an accurate encounter note that tells an accurate story of the patient’s condition. Coding-aware example: A patient with a history of type 2 diabetes mellitus is experiencing issues with foot ulcer healing. The patient presents with a bandaged ulcer on their left heel, which isn’t causing any pain. The provider will continue to monitor the healing of the ulcer and encourages the patient to avoid putting pressure on the affected foot to promote healing. In this example, the coding-aware AI would recommend E11.621 and L97.42- (Non-pressure chronic ulcer of left heel and midfoot), the latter of which would need to be further specified following review of the documentation. “If your AI doesn’t understand how to think like a professional coder, it’ll just crank out a note,” Morris explained. “We know from coding and all of the work you spend educating clinicians on how to appropriately document, or perhaps standardizing templates, that if you turn on AI that’s dumb to the realities of coding, you actually start creating more work for yourself,” he added. Mike Shaughnessy, BA, CPC, Production Editor, AAPC
