Training for Jobs That May Not Exist: Artificial Intelligence and the Future of Medical Residency

Artificial intelligence (AI) is rapidly changing the tasks performed by physicians, raising questions about how graduate medical education should prepare trainees for future practice and whether residency workforce planning should change as AI adoption expands. This brief review examines the implications of AI for residency education and physician workforce planning, using radiology as an early model of technology-driven task transformation while considering broader effects across diagnostic, procedural, and relationship-based specialties.

AI is already being used for clinical documentation, image interpretation, predictive analytics, administrative workflows, and large-scale data analysis. AI-assisted scribing has particular relevance to chronic pain and other documentation-intensive specialties, where it may reduce clerical burden and preserve clinician attention for patient care. However, current evidence more strongly supports task redistribution and physician augmentation than complete replacement. Specialties built around standardized digital information may experience earlier automation, whereas procedural skill, complex communication, longitudinal judgment, and responsibility for final clinical decisions remain difficult to substitute. At present, broad reductions in residency positions based only on anticipated AI capabilities seem premature.

Residency programs should adapt curricula now by incorporating AI literacy, verification skills, data interpretation, and safeguards against deskilling. Changes in residency position numbers should occur more cautiously and should be based on durable, specialty-specific workforce evidence, including hiring patterns, patient access, workload, productivity, and regional need. The goal should be to train physicians who can safely supervise and integrate AI while preserving the human expertise that remains central to high-quality care.