Physicians as context engineers: redefining clinical competency for artificial intelligence
10.5124/jkma.26.0022
- Author:
Sanghyun AHN
- Publication Type:Special contribution
- From:Journal of the Korean Medical Association
2026;69(4):358-363
- CountryRepublic of Korea
- Language:Korean
-
Abstract:
The adoption of artificial intelligence (AI) in clinical practice is accelerating. Many physicians perceive effective AI use as requiring unfamiliar technical skills. This paper reexamines that assumption and proposes a reinterpretation of physicians' existing competencies for AI collaboration.Current concepts: Recent AI discourse has moved from prompt engineering toward the broader practice of context engineering. These concepts are not mutually exclusive: prompt engineering may be understood as one component of context engineering, which involves the deliberate collection, structuring, and integration of contextual information to improve AI output quality. In 2025, Gartner identified context engineering as a core AI competency. A 2026 Nature Medicine paper proposed context engineering as an emerging physician role across four axes: data, task, tool, and norm contexts. Whereas that paper framed this as a competency to develop, this paper argues that physicians have already been trained in all four axes. History taking structures subjective context; physical examination generates primary data inaccessible to AI; clinical reasoning defines task context; and guideline adherence with patient preferences constitutes norm context. Furthermore, patients increasingly bring AI-generated health conversation logs—a novel semi-structured context—to clinical encounters, expanding physicians’ context engineering scope with clinical and medicolegal implications.Discussion and conclusion: The challenge is less acquiring a novel technical skill than reframing existing competencies and adopting tools capable of handling rich clinical context. Future research should examine which forms of physician-supplied context most improve AI-assisted decisions and how patient-generated AI logs should be incorporated into clinical documentation.
