AI Scribes Fail Complex Medical Documentation

Current AI scribes struggle with complex medical visits, generating inaccurate and unhelpful notes, forcing clinicians to still manually document and undermining the technology's purpose.

FM outpatient, DAX only, haven't branched out to anything else so take this for what it is!

Setup is honestly fine. You drop the DAX sections into your Epic templates and it works mechanically, which is nice.

Actual performance though? All over the place, lol. Simple one or two issue visits it does a decent job. AWV and physicals it just completely falls apart on me. Wrong details in wrong sections, hallucinated statements, dramatic urgent sounding language about the most basic anticipatory guidance, and then like one sentence about something we genuinely spent ten minutes discussing in depth. I finish my day and have nothing useful to work from.

I still take my own shorthand because I just don't trust it, which kind of defeats the whole purpose if I'm being honest with myself.

Other docs in our system are not making this look better either. I see notes that are paragraphs of flowery garbage that flat out contradict their own typed plan. It's pretty clear a lot of people were voluntold to use it and never really engaged with it seriously, and nobody is proofreading. It's a little alarming, to be totally frank.

I keep using it as a detail catcher for the small things I might miss. That's genuinely about all it's worth to me right now in its current state.

If someone tells me DAX is the best thing that ever happened to their practice, I'm probably going to assume their notes are a disaster and they've stopped caring about documentation, lol.

For those of you who actually have something working well for complex visits, AWV, physicals, high volume days, what are you using and what actually made the difference? DAX, Nabla, Freed, something else? Genuinely open to hearing what's working because this isn't cutting it for me.