This pilot study tested Whisper and WhisperX on standardized medical recordings from native and non-native English speakers and found much higher error rates for non-native speakers. Adding a GPT-4o correction step cut word error rates by up to 6.9% and nearly eliminated accent-related disparities. The chained ASR–LLM pipeline boosts transcription accuracy across diverse clinical speakers.

Shiv Rao, cardiologist-turned-CEO of Abridge, explains why the company focuses on doctor-patient conversations rather than just notes, how it builds 70–80% of its own AI models, and why it retains full control despite raising $800 million. He also outlines Abridge’s edge over dozens of rival scribes, its new real-time prior-auth tool, and how it navigates liability and regulation.