Artificial intelligence (AI) is rapidly reshaping hospital-based care and medical education, yet frontline clinicians often lack practical frameworks to evaluate and implement these tools safely and effectively. This presentation describes a multidisciplinary, hospital-based approach to adopting AI applications in both clinical workflows and teaching. The use of AI scribes, automated discharge summaries, and educational chatbots are used as representative case studies. First, we outline key medico-legal, privacy, bias, and workflow considerations derived from national guidance and institutional experience at a large community teaching hospital in Toronto with a multidisciplinary AI working group. We then examine AI scribes for patient care, focusing on informed consent, data handling, specialty-specific prompting, and their impact on cognitive burden and documentation quality in history-heavy fields such as geriatric medicine. Next, we present early work on large language model–generated discharge summaries, highlighting feasibility, workload reduction potential, and current barriers related to data privacy, hallucinations, and integration with electronic medical records. Finally, we report results from an AI chatbot named Medbot designed to teach medical syndromes to learners, including completion rates and agreement between AI and human raters on performance assessment. Collectively, these examples illustrate how thoughtfully governed AI solutions can decrease clinician cognitive load, enhance patient care and enrich medical education while underscoring ongoing ethical, legal, and sociotechnical challenges that require clinician leadership and interdisciplinary collaboration. We will also demonstrate the AI scribe, automated discharge summary solution and educational AI chatbot at the conference.
