Preserving bedside teaching: An AI-supported system for capturing spontaneous clinical teaching

Bedside rounds generate highly valuable spontaneous teaching, yet these moments are rarely preserved. Learners often struggle to document teaching points while participating in patient care, and teaching is easily forgotten once rounds move on. Existing tools-manual notes, post-rounds summaries, or static teaching cards do not capture the dynamic, real-time nature of bedside teaching.

We developed Rounds Copilot, an AI-supported system that automatically captures spoken bedside teaching and converts it into structured learning materials. The tool integrates automated speech transcription, extraction of discrete teaching points, classification of guideline-backed evidence versus clinical pearls, and generation of formative assessment questions.

A simulated 15 min pericarditis teaching encounter was processed using the tool. Within three minutes, the system produced seven teaching kernels, evidence classification, and five multiple-choice questions. Informal feedback from four residents indicated the evidence-versus-pearl distinction clarified learning priorities and immediate assessment questions supported knowledge consolidation.

The system costs less than USD $1 per typical session, requires minimal technical skills, and includes open-source components. Offline transcription supports use in low-bandwidth settings.

Planned work includes real-world testing, usability studies, and development of privacy-compliant clinical deployment pathways.