Ambient AI is coding for payers, not patients

Ambient AI documentation tools are increasingly sold on a single selling point: listen to the visit, settle on a diagnosis, and produce a ready-to-submit ICD-10 code. Jay Anders, MD, Chief Medical Officer of Medicomp Systems, argues that's the wrong finish line — and that the rapid rollout of these tools is outpacing hard questions about what they actually capture.

A code that satisfies a payer isn't the same as the clinical picture a physician needs to treat the patient. Anders says that gap is becoming a real patient safety issue as care moves toward precision medicine and genomics, where a vague ICD-10 code can quietly block a patient from a targeted therapy that only exists for a more specific diagnosis. It's a dynamic that echoes a broader trend already showing up in the data: as ambient AI drives higher-acuity coding, billing intensity is rising even when the underlying clinical detail hasn't necessarily improved.

In this conversation, Anders explains why most ambient AI gets the order backwards — reverse-engineering a diagnosis to fit a billing code instead of building the code from a complete clinical picture. He makes the case for why these tools capture what's said out loud but never prompt for the one missing symptom or family-history detail that would sharpen the diagnosis. For physicians relying on ambient AI every day, it's a conversation about what documentation is really for.