An AI Competency Framework for Emergency Medicine: A Multiphase Consensus Process

Artificial intelligence (AI) is increasingly embedded in U.S. emergency department workflows, but no medical specialty has defined what competent use of these tools requires of its physicians. Existing accreditation requirements are silent on AI, and parallel national efforts span the learning continuum but are intentionally specialty-agnostic. As one output of a working group of the 2026 Society for Academic Emergency Medicine (SAEM) Artificial Intelligence Consensus Conference, we developed an emergency medicine-specific framework for AI competency. We employed a multiphase consensus design: a Nominal Group Technique session generated themes from a curated set of clinical AI scenarios, and a two-round modified Delphi process (panel n = 13 across 12 academic medical centers), separated by a consolidation videoconference, evaluated and extended the framework. A priori consensus required [≥]70% of panelists rating [≥]4 on a 5-point Likert scale ([≥]75% for themes) with interquartile range [≤]1; we followed established Delphi reporting guidance. The framework comprises 5 themes (Communicating about AI, Understanding appropriate use cases, Interacting with AI, AI risk management, Cognitive impacts of AI), 5 derived competencies (one-to-one theme-to-competency mapping endorsed by 12 of 13 panelists), 19 subthemes, and 10 retained clinical scenarios. All themes, derived competencies, and rated subthemes met consensus thresholds; 9 of 10 scenarios reached consensus, with 1 retained as a future-state operational model. Two cross-cutting conceptual frames emerged: pre-emptive versus post-hoc AI integration, and tiered competencies as an articulated need rather than a pre-specified answer. This is, to our knowledge, the first specialty-specific AI competency framework for a United States medical specialty with quantitative content validity evidence from expert consensus. It provides a structural target for emergency medicine curriculum development, assessment design, and faculty development, with downstream priorities including stage-specific competency assignment, assessment instrument development, and periodic re-evaluation as AI deployment evolves.