The AI Sessions Clinicians Will Stop Believing

Key Takeaways

• AI education faces a credibility problem when it still promises broad time savings instead of showing where benefit is real, limited, and governed.
• The sharper AI distinction now is between lower-risk information retrieval and higher-risk clinical-management use that needs curation, auditability, safe environments, and review.
• A separate but narrower signal suggests CME teams may need to build interaction into faculty planning and workshop structure rather than relying on speaker style alone.

The strongest signal this week is that AI education may lose credibility if it keeps overselling efficiency. The evidence is narrow and partly duplicated across formats, led more by an informatics perspective than by a broad clinician chorus, but it points to a concrete implication for CME programming.

AI claims are being judged against cleanup work

In a recent informatics discussion, a practicing clinician drew a blunt line: AI can be useful for broad search and information gathering, but claims about major time savings weaken when clinicians still have to review, edit, and stand behind the output (podcast, video). The same conversation also emphasized curated models, audit trails, and HIPAA-safe environments for higher-risk use.

For CME providers, this is a narrower continuation of the AI thread already visible in our earlier brief on clinicians asking harder questions about AI than accuracy. This week’s added pressure is on the educational promise itself: if an activity still treats AI as a general productivity upgrade, it may sound less like guidance and more like marketing.

The implication is straightforward. AI sessions should stop treating benefit as self-evident and start showing the tradeoffs: which tasks are relatively safe, where review burden cancels out saved time, and what governance has to be in place before a use case is ready to teach.

Interaction looks less like a teaching preference and more like a production choice

A second, lighter signal this week came from faculty-development commentary and a specialty-specific QI workshop discussion. In one source, CPD leaders described the familiar problem of overpacked sessions and pointed to polling, Q&A, role play, reflection, and chair-led planning as ways to build interaction into the session (podcast). In another, radiation oncology educators argued that QI learning worked better when learners tackled a local problem through workshop methods such as stakeholder mapping and PDSA-style problem solving rather than passive modules (video).

This is not broad consensus, and one source is specialty-specific. But together they support a useful operator point for CME teams: interaction does not reliably appear just because a faculty member is engaging. It often has to be specified in planning templates, moderation plans, and time allocation.

The portability beyond these examples is plausible rather than proven. Still, for practice-change, systems, or QI education, CME teams should ask a harder planning question before launch: where exactly will learners apply, discuss, or test the idea during the session rather than only hear it?

What CME Providers Should Do Now

• Review current AI activities for vague efficiency claims and replace them with task-specific statements about benefit, review burden, and limits.
• Separate low-risk AI use cases from higher-risk clinical-management scenarios in session design, and explain the governance conditions required for each.
• Update faculty planning templates so speakers must build in defined interaction moments, especially for QI, systems, and practice-oriented education.

Watchlist

• Live discussion and near-peer teaching still merits watching in narrow specialist settings. A fellowship conversation argued that post-lecture discussion improves retention and patient-level application, while peer procedural teaching fills gaps when supervision is limited, but the current evidence is too fellowship-specific to support a full section (podcast).