Fabio Ynoe de Moraes, Radiation Oncologist and Associate Professor at Queen’s University, shared a post on LinkedIn:
“The most important Viewpoint clinicians should read this month just dropped.
And it delivers a message our field can no longer ignore:
Generative AI will change clinical medicine more in the next 3-5 years than the electronic health record did in the last 20.
A new JAMA IM Viewpoint by Morgan, Rodman, and Goodman outlines – in a remarkably clear way – what every physician must know to stay clinically relevant in the age of agentic AI.
Here are the three takeaways health systems should be discussing today:
1. We are moving from “scribes and summaries” to AI systems that perform core cognitive tasks
First-generation tools were easy: draft a prior auth letter, summarize a chart, clean up documentation.
What’s coming next is fundamentally different.
According to the authors, second-generation GenAI will:
• Take patient histories
• Generate differential diagnoses
• Interpret diagnostic tests
• Synthesize full guideline-level evidence
• Provide management recommendations
Some tasks will be embedded in EHR workflows. Others will reach patients directly, bypassing us entirely.
This shifts physicians from “primary producers of documentation” to expert adjudicators of AI-generated clinical reasoning.
2. Physicians will need a new skillset – not coding, but cognitive AI literacy
The Viewpoint makes a critical distinction: physicians don’t need to become engineers; they need to master clinical prompting and critical appraisal of AI outputs.
This includes:
• Knowing how to structure patient data for optimal AI reasoning
• Interrogating the model (“Which note documents the fever?”)
• Detecting hallucinations and sycophancy
• Understanding bias propagation
• Reviewing and editing AI-generated notes with legal accountability
This is clinical reasoning – but with a new layer of complexity.
3. The physician – patient relationship will evolve faster than most anticipate
Patients will increasingly arrive with:
• AI-derived symptom analyses
• Self-generated differential diagnoses
• Personalized recommendations
• Second opinions from consumer-facing LLM tools
Our role will shift from “source of truth” to trusted interpreter and contextualizer.
From information provider to clinical sense-maker.
This requires humility, communication skills, and a deep understanding of how these systems think and fail.
Why this matters now
Physicians who fail to engage with AI will quickly face “archaic workflows” and increased liability.
Those who embrace it will offload administrative burden and elevate the human parts of medicine:
judgment, empathy, shared decision-making, and complex clinical nuance.
AI handles information overload; clinicians handle meaning.
My take
This piece should be required reading for all clinicians, chief medical officers, academic leaders, and health-system executives.
AI is no longer a future technology -it is the new substrate of clinical practice.”
Title: How Physicians Can Prepare for Generative AI
Authors: Daniel J. Morgan, Adam Rodman, Katherine E. Goodman,
More posts featuring Fabio Ynoe de Moraes on OncoDaily.