JMIR Publications

ABSTRACT

Burnout among medical educators is common, driven largely by the administrative and assessment work built into teaching roles. Artificial intelligence (AI) is widely promoted as a way to lift that burden, and AI tools have been shown to reduce documentation load and burnout among clinicians. That evidence stops at clinical documentation and has not reached teaching, while the parallel literature on AI in medical education maps where these tools can be used without offering a rule for deciding which teaching tasks should be delegated and which should not. We propose a task-allocation framework for the human–AI division of labor in medical education. The framework places each teaching task on two axes, the intensity of the human factor it involves and the capability and safety of AI to perform it, yielding four modes of allocation (Automate, Augment, Reserve/Protect, and Monitor), with a stakes modifier that returns summative and safety-critical work to human control and a rubric for rating new tasks. The framework’s central aim is to redirect the capacity freed by automation into the tasks only a clinician-teacher can perform: bedside reasoning, professional identity formation, and mentorship. We present the framework as a governance tool that stands on pedagogical and ethical grounds independent of any wellbeing claim; the further proposition that reallocation protects educators and, through them, learning is advanced separately, as a hypothesis for testing. The framework is conceptual and not yet validated; we situate it against existing models, specify the ethical guardrails it requires, and propose a validation agenda.