AI systems that capture and transcribe clinical encounters, generating structured notes from ambient audio.
What does the peer-reviewed literature say about how these systems fail — and where they succeed?