Medical education hasn’t caught up with AI. We’re training doctors for a healthcare system that’s rapidly changing, using curricula developed for a pre-AI era.

I’ve been involved in discussions about medical education reform, and there’s growing recognition that something needs to change. Here’s my perspective on what that looks like.

The Current Gap

Today’s medical students learn to:

• Take histories and perform examinations
• Interpret investigations (labs, imaging) with minimal AI involvement
• Make clinical decisions based on their own reasoning
• Document care manually

Tomorrow’s doctors will work in environments where:

• AI assists with diagnosis and decision support
• Imaging and pathology include AI interpretation
• Documentation is AI-assisted
• Algorithms influence care pathways

We’re not preparing students for this reality.

What Future Doctors Need to Know

AI literacy fundamentals. Not data science—that’s specialist training. But basic understanding of:

• How AI/ML systems work conceptually
• What AI can and can’t do reliably
• How to interpret AI recommendations critically
• Common AI limitations and failure modes

Critical evaluation of AI tools. Like evaluating any clinical tool:

• What evidence supports this AI’s use?
• In what populations was it validated?
• What are its sensitivity, specificity, and predictive values?
• Where might it perform poorly?

Working alongside AI. Practical skills for AI-augmented practice:

• How to integrate AI recommendations with clinical judgment
• When to trust AI and when to override
• How to explain AI involvement to patients
• How to maintain clinical skills when AI is available

Ethical frameworks for AI. Understanding:

• Consent and disclosure for AI-involved care
• Algorithmic bias and its clinical implications
• Responsibility when AI contributes to errors
• Equity considerations in AI access

Data skills. Basic competency in:

• Understanding clinical data quality issues
• Interpreting quantitative AI outputs
• Recognising when data limitations affect AI reliability

How Medical Schools Could Adapt

Integrate AI into existing teaching. Rather than adding AI as a separate subject, integrate it into clinical teaching. When learning about chest X-ray interpretation, include discussion of AI detection tools. When learning about prescribing, discuss AI medication checking.

Update clinical placements. Students should encounter AI during clinical rotations. Where AI is deployed, students should learn how clinicians use it.

Critical appraisal skills. Extend evidence-based medicine training to include AI evaluation. How do you critically appraise an AI study? What questions should you ask about AI evidence?

Simulation and scenarios. Simulation centres could include AI-augmented scenarios. How does a student respond when AI and clinical assessment disagree? What happens when AI flags something unexpected?

Ethics curriculum expansion. Medical ethics teaching should include AI-specific content—algorithmic justice, responsibility for AI-influenced decisions, patient autonomy in the context of AI.

The Deskilling Concern

There’s legitimate concern that AI creates deskilling risks. If students learn to practice with AI assistance from the start, will they develop underlying clinical skills?

This is a real tension. Some approaches to managing it:

Foundation skills first. Ensure core clinical skills are established before introducing AI augmentation. Students should learn to read X-rays before learning about AI interpretation.

Practice without AI. Include assessment and practice situations where AI isn’t available. Doctors need to function when AI systems fail.

Understand AI limitations. If students understand where AI fails, they’ll recognise when their own skills are essential.

Maintain clinical reasoning teaching. Diagnostic reasoning and clinical judgment should remain central to medical education, not replaced by “follow the algorithm.”

The goal isn’t doctors who can’t function without AI—it’s doctors who can use AI effectively while maintaining independent clinical capability.

What About Current Practitioners?

Medical education reform addresses future doctors. What about the current workforce?

CPD requirements. Continuing professional development should include AI literacy. Medical colleges could develop AI modules for specialty training.

Institutional training. Health services implementing AI should train their medical staff—not just technical training, but critical appraisal of the specific AI tools deployed.

Specialty-specific guidance. Medical colleges could develop position statements and guidance on AI in their specialties.

Peer learning. Creating spaces for clinicians to share AI experiences—what works, what doesn’t, how to manage challenges—would accelerate learning.

The Research Training Gap

Medical research training also needs attention. Future clinician-researchers need to:

• Understand AI research methodology
• Critically appraise AI studies
• Conduct research involving AI systems
• Consider ethical implications of AI research

PhD programs, research fellowships, and postgraduate research methods training should include AI-specific content.

Implementation Challenges

Several barriers to reforming medical education:

Curriculum crowding. Medical curricula are already packed. Adding AI content means removing something else.

Faculty expertise. Many medical educators lack AI expertise themselves. Faculty development is needed.

Assessment challenges. How do you assess AI literacy and AI-augmented practice? New assessment methods are required.

Rapid change. AI capabilities evolve quickly. Curricula developed today might be outdated in three years.

Accreditation lag. Medical school accreditation standards take time to update. Standards may not yet require AI content.

These challenges are manageable but require coordinated effort.

Who Should Lead This?

Reform needs multiple stakeholders:

• Medical Deans Australia and New Zealand for coordinating curriculum change
• Australian Medical Council for accreditation standards
• Medical colleges for specialty training content
• Health informatics bodies for expertise and input
• Students for advocacy and feedback

Individual medical schools can innovate, but sector-wide change requires coordination.

The Urgency

Some might argue we have time—AI deployment is still early. I disagree. Medical education has long lead times. Changes made now affect graduates in 5-10 years. By then, AI in healthcare will be far more prevalent.

The students entering medical school today will practice in AI-augmented healthcare systems for their entire careers. We should prepare them accordingly.

---

Dr. Rebecca Liu is a health informatics specialist and former Chief Clinical Information Officer. She advises healthcare organisations on clinical AI strategy and implementation.