"These AI Scribes Are Crap!"

But they make us faster and we love them.

The headline and sub-heading pretty much sums up medical life with a digital scribe, and is encapsulated in the quantitative and qualitative feedback captured by this article.

Of this little pilot group – most of whom went into the pilot believing AI scribes were going to be beneficial – the DAX Copilot integrated into Epic did, indeed, make them faster:

It also, as you can see above, made them wordier – a “feature” of LLM output with which we are all familiar. And, that imaginative wordiness has its drawbacks:

The diversity in feedback is interesting, of course, and does make me wonder about some of the factors influencing clinicians’ experiences:

• Are certain clinicians having poor output due the manner in which they conduct patient interviews?

• Are certain clinicians holding on to a certain bespoke formatting and style for their notes?

• Are some clinicians just gloriously slow at typing and anything resembling a note will make them happy?

There are all interesting questions relevant to either the refinement of digital scribes or the roll-out of these tools to wider clinician audiences.