> How are note quality improvements measured?

Every provider is under an Assistant Clinical Director, and they report to the Clinical Directors, who report to the CMO. ACDs see fewer patients than regular providers because they have more admin time. That admin time is used to check charts. We don't review every chart, but a pretty good sampling. I meet with them monthly to talk about tech issues, and that's where I helped them create templates for notes that we can have the system output in that same format. We'll tweak the formats as needed, or the ACDs will talk with a provider about changes in how they handle the patient.

Also, we look at denial reasons. Any time a claim is rejected by a payor for note related reasons it gets a full review from clinical staff other than the original provider.

> Vibe-notes might be more verbose and better sounding (which would explain the NPS and satisfaction metrics), but still not actually match the doctor's actual words or intent.

That's the great thing about these, they listen to the entire visit, they hear everything that happens, make a full transcript, then create a summary. It's not a situation where the doc talks for 30 seconds into a mic then the AI fleshes it out, it's the exact opposite. We're using AI to distill the visit into the note, not expand a small note into a larger one. We're not generating data, we're condensing it. Doctors must read each note, and they are legally liable for the note quality. Doctors are highly competitive and image conscious, so they're actually a great backstop for accuracy. If they notice inaccuracies in their summaries, I ASSUME you I personally hear about each and every one. I'm ok with that, though, the buck stops at my desk.

> Are the AI-generated notes actually compared with ground truth to prove they are accurate?

Yes. A doctor could lose their license, so every provider checks their notes, and our CMO and clinical oversight staff take that extremely seriously.