One idea I run across frequently when hearing about natural language processing (NLP) and the electronic health record (EHR) is how real-time transcription of the conversations between patient and physician will free up the physician’s time so she can “concentrate on the patient.”

There is also the ever-present promise of real-time AI being able to offer insights into the patient’s health issues (as if the problems facing healthcare driven by a long drought of insights).

Maybe—and a very doubtful maybe—this is all true. But even if it is, the AI programmers and EHR developers are forgetting one big thing.

Often, the conversation between a patient and physician is very delicate and nuanced. It goes well beyond the spoken word and may include evaluation of body language, facial expression, tone, and reading in-between the lines.

Effectively, when both parties are aware their every word is being recorded and processed in real-time, I can guarantee this changes both the nature of the conversation and the direction and extent and the evaluation.

Whether this change will be for better or worse for the patient has yet to be seen, but let’s be very clear—making the EHR (and AI software behind it) the driving force and focus of patient-physician engagement will come with a cost.