CME and Case Presentations with AI
How doctors can use AI to build CME decks, academic talks, case discussions, and teaching presentations without losing evidence discipline.
CME and case presentations need more rigor than patient education decks.
The audience expects clinical reasoning, evidence, and practical takeaways. AI can help you organize the talk, but it can also invent references, exaggerate recommendations, or simplify nuance too aggressively.
For academic presentations, use AI as an editor and structure assistant. Do not use it as an unsupervised literature reviewer.
Good CME Use Cases
AI can help with:
- Turning rough notes into a teaching structure
- Creating slide titles
- Drafting speaker notes
- Simplifying dense guideline text
- Creating case discussion questions
- Building pre-test and post-test MCQs
- Creating summary tables
- Checking whether the talk has a logical flow
AI should not be trusted to:
- Invent references
- Decide final treatment recommendations
- Summarize a guideline you have not provided
- Create patient-specific advice from real identifiers
- Generate medico-legal conclusions
CME Deck Structure
A practical CME deck often follows this pattern:
- Title and learning objectives
- Why this topic matters
- Clinical scenario or de-identified case
- Initial approach
- Key decision points
- Evidence or guideline summary
- Common pitfalls
- Practical workflow
- Audience discussion question
- Take-home points
- References
For a 20-minute talk, 12-16 slides is usually enough.
For a 45-minute talk, 20-30 slides may work, but only if many slides are visual or discussion-based.
Prompt: CME Deck from Notes
You are helping a doctor create a CME presentation.
Audience: [general physicians / residents / specialists / nurses]
Topic: [TOPIC]
Duration: [MINUTES]
Purpose: practical teaching, not exhaustive review
Use only the source notes below.
Do not invent references.
Do not add guideline recommendations unless they are in the notes.
If a claim needs verification, label it VERIFY.
Create:
1. Learning objectives
2. Slide-by-slide outline
3. Speaker notes
4. Suggested case discussion points
5. A references slide placeholder
6. A verification checklist
Source notes:
[PASTE NOTES]
Prompt: De-Identify a Case for Teaching
Convert this clinical case into a de-identified teaching case.
Remove or generalize:
- Name
- Exact age if not essential
- Dates
- Phone numbers
- Address
- Hospital number
- Rare identifying details
- Occupation if identifying
Keep:
- Clinically relevant sequence
- Presenting complaint
- Key findings
- Management decision points
- Learning points
Do not add new clinical facts.
Mark missing information as "not provided".
Case:
[PASTE CASE]
Prompt: Add Teaching Questions
Create teaching questions for this CME deck.
Audience: [AUDIENCE]
Level: [beginner / intermediate / advanced]
Create:
- 3 opening questions
- 3 case discussion questions
- 5 MCQs with answer explanations
- 3 reflection questions for practice change
Rules:
- Use only the content in the deck
- Do not introduce new treatment claims
- Mark any uncertain answer as VERIFY
Deck outline:
[PASTE OUTLINE]
Reference Discipline
For CME decks:
- Keep a references slide
- Do not accept AI-generated citations without checking them
- Prefer guideline names and URLs you personally verify
- Mark local protocol differences clearly
- Date the presentation
- Mention when content is opinion, local practice, or guideline-based
Example wording:
“This presentation is for education. Local protocols and patient-specific decisions may differ.”
Review Checklist
- Case is de-identified
- Learning objectives are clear
- Every major recommendation is source-backed
- References are real and checked
- Drug doses are verified if included
- Local protocol differences are noted
- No unsupported statistics remain
- Speaker notes do not add unreviewed advice
- Take-home points are practical
1-Minute Takeaway
Use AI to structure CME talks, not to invent medical authority.
For doctors, the value is faster organization, better speaker notes, cleaner teaching questions, and a stronger verification process.