AI in the Hospital: Navigating Deployment and Adoption Challenges

Abstract

Healthcare systems worldwide are experiencing escalating clinician burnout driven in large part by administrative burden and documentation-heavy electronic health record (EHR) workflows. Large Language Model (LLM)-powered interfaces promise to reduce manual data entry through conversational documentation, summarisation, and generative clinical assistance. However, translating these systems from research prototypes to deployment within national hospital infrastructures introduces significant technical, regulatory, governance, and human–AI interaction challenges. This Birds of a Feather (BoF) session convenes researchers, clinicians, designers, and industry practitioners to critically examine the deployment realities of LLM-powered clinical interfaces at scale. We focus on administrative burden reduction, safety and automation bias, off-premise cloud-hosted model governance, limited fine-tuning and institutional control, prompt engineering expertise requirements, and evaluation methodologies for generative clinical systems. Our goal is to foster interdisciplinary dialogue and seed a sustained community shaping responsible, scalable AI-enabled clinical interface design.