The evolving physician-AI relationship: a five-tier framework for integrating intelligent systems into clinical practice and medical education.

Artificial intelligence (AI) is increasingly entering oncology, with systems demonstrating physician-comparable performance in selected tasks such as imaging interpretation, digital pathology analysis, and clinical documentation. However, limitations including factual errors, algorithmic bias, and lack of accountability mean that oncologist oversight remains essential. The key question is therefore not whether AI will enter oncology, but how the interaction between oncologists and AI systems will evolve. We propose a five-tier framework describing progressively different modes of physician-AI interaction. Tier 1 represents traditional oncology practice without AI. Tier 2 involves oncologists using existing AI tools for defined clinical tasks. Tier 3 describes physicians adapting or developing AI-enabled solutions tailored to clinical needs. Tier 4 reflects AI-dominated workflows supervised by experienced oncologists. Tier 5 represents autonomous AI systems performing selected clinical tasks with minimal physician involvement. The framework illustrates how oncologists' roles may evolve from direct decision makers to users, innovators, supervisors, and designers of AI-supported care, and outlines the competencies required across these tiers. It provides a practical structure for understanding AI integration into oncology and guiding the training of future oncologists.