Nothing about the call. Everything about what you get back from it.

You already capture the audio, whether that's a scheduled AI check-in, a coaching call, or an intake conversation. Sona-2 sits underneath that same recording and reads the acoustic layer, pitch, pause structure, loudness, spectral measures, for signals the transcript and the sentiment score never touch. No second recording step, no new consent flow beyond the one you already run, no change to how the visit feels on the patient's end. The output is a structured signal your platform can route, trend, and flag before someone's next scheduled interaction, which for most check-in cadences is days or weeks away.