Episode
How Generative AI Is Writing Doctor Visit Summaries
- Published
- Jul 10, 2026
- Duration seconds
- 459
- Processing state
not_requested
Actions
POST https://stenobird.com/v1/public/podcasts/healthtech-talks-with-fexingo-digital-health-telemedicine-and-medical-software-7871821/episodes/how-generative-ai-is-writing-doctor-visit-summaries/transcription-requests
Idempotently request low-priority transcript generation for this episode.GET https://stenobird.com/podcast/healthtech-talks-with-fexingo-digital-health-telemedicine-and-medical-software-7871821/how-generative-ai-is-writing-doctor-visit-summaries.md
Read the agent-friendly Markdown representation of this episode resource.
Summary
Lucas and Luna explore how generative AI is automating the clinical note, one of the most hated tasks in medicine. They focus on the specific case of NYU Langone Health, which piloted an ambient AI scribe from Nuance's Dragon Ambient eXperience (DAX) in 2024. By mid-2026, over 200 physicians are using it, saving an average of 7 minutes per visit. Lucas breaks down the workflow: a microphone captures the patient-doctor conversation, the AI generates a structured note in the electronic health record, and the physician reviews and signs it. They discuss the accuracy challenge — studies show about 85-90% of notes need zero edits, but the remaining 10-15% can contain subtle errors like hallucinated symptoms. Luna raises the risk of documentation bias, and Lucas cites a JAMA study finding the AI sometimes omits patient expressions of uncertainty. They also touch on reimbursement: Medicare now covers virtual scribes, which has accelerated adoption. The episode ends with a question about whether AI will change how doctors listen to patients. #GenerativeAI #ClinicalNotes #NYULangone #Nuance #DAX #AmbientScribe #MedicalDocumentation #PhysicianBurnout #ElectronicHealthRecord #AIinHealthcare #Medicare #VirtualScribe #JAMA #DocumentationBias #Healthtech #Business #FexingoBusiness #BusinessPodcast Keep every episode free: buymeacoffee.com/fexingo