In the era of generative AI, doctors are becoming context engineers, providing specific context for AI models to work better.[1] The main problem with medical AI models is contextual errors, where the answers are not accurate for a specific context, such as medical specialty, geographic location, or socioeconomic factors.[1] These flaws are not unique, but represent a broad limitation of all types of medical AI models.[1] AI trained predominantly in a single specialty may give wrong answers or miss multisystem diseases.[1] It is proposed to develop models trained on multiple specialties that can switch contexts in real time and focus on relevant information.[1] Such models should incorporate geographic data for locally accurate responses with global health implications.[1] AI is already helping clinicians write patient notes and find relevant scientific articles, while context-switching models would improve treatment by adapting to different phases, such as symptom analysis or therapy suggestions.[1] The authors see the greatest benefit of AI in supporting doctors as a team partner, not a replacement.[1][2]