From Prompt Engineering to Context Engineering: Why Precision Oncology Needs Context

Abstract

The integration of artificial intelligence (AI) into precision oncology has generated both excitement and debate. Early applications—exemplified by systems that simply summarize standard guidelines—risk homogenizing cancer care and overlook the nuanced personalization that experienced oncologists provide. In oncology, multiple evidence-based treatment paths often exist for the same patient, and clinicians’ choices are influenced by local practices, resource constraints, and individual patient factors. If AI merely regurgitates one-size-fits-all recommendations, it replaces personalization with standardization, offering little more than an automated guideline. This commentary argues that the next leap forward for AI in oncology is “context engineering”: embedding models with the why behind institutional practices, clinician experience, and patient-specific nuances. By incorporating domain-specific data, real-world evidence, and clinician feedback, AI tools can adapt recommendations to reflect a given clinical context—academic or community, trial-ready or resource-limited—while still upholding national guidelines. Such context-aware AI could evolve from a generic “copilot” into a true partner in decision making, enhancing individualized patient care rather than diluting it. We discuss the need for this paradigm shift, its potential benefits, and challenges to its implementation, emphasizing that the future of oncology AI lies in a better context, not just better prompts.