Faculty AI programs must move from demos to documented trust and workflow change

Key Takeaways

• The strongest signal was narrow but useful: AI education is being framed less as tool exposure and more as faculty development for workflow change.
• Trust is becoming an instructional design problem, not just a governance or disclosure problem.
• Blended, longitudinal formats fit the need better than isolated AI sessions, but outcomes have to measure adoption, confidence, and role clarity—not just completion.

Oncology educators this week described AI as already embedded in clinical workflow, while the education ask is moving from tool literacy to trust-building faculty development. The signal is emerging and largely comes from oncology and provider-owned education sources, so it should be treated as directional rather than broad clinician consensus.

AI faculty development has to teach implementation, not fascination

The useful tension this week was not whether clinicians are interested in AI. It was whether education is preparing faculty and teams to use AI without eroding trust.

In a BackTable Tumor Board discussion, oncology speakers framed near-term AI value around workflow—scribes, chart overlays, repetitive tasks, and information retrieval—while keeping clinical decision-making claims restrained (BackTable). That matters for CME providers because many AI activities still treat the tool as the learning object. The clinician-facing problem is different: how to test it, explain it, assign responsibility, and embed it into the day without creating fear or workarounds.

The Total Health Oncology session made the trust problem explicit. One speaker cited public concern that “50% of Americans are more concerned than excited about the use of AI in daily life” and then connected clinician adoption to transparency, governance, culture, and experimentation (Total Health Oncology). The provider implication is straightforward: an AI faculty program should include QA rehearsal, failure scenarios, disclosure language, escalation paths, and role boundaries. A demo of a scribe or search tool is not enough.

Provider-owned education signals pointed in the same direction. AMA Ed Hub promoted faculty development that includes “using AI in medical training” (AMA Ed Hub), while AOE Consulting’s note on “other/blended learning activity” formats is a reminder that some AI education may not fit neatly into a single live or enduring format (AOE Consulting). That is a format clue, not proof of broad demand. But it supports the same operating question: if AI adoption depends on trust over time, why are so many programs still built as one-time orientation?

We saw a related pattern in an earlier brief on AI assistants and clinical judgment: the learning need was shifting from knowing that AI exists to knowing how to work with it responsibly. This week’s wrinkle is that faculty development becomes the delivery system. CME teams should ask whether their AI offerings create repeatable behaviors inside workflow—or merely improve vocabulary about AI.

What CME Providers Should Do Now

• Audit AI activities for longitudinal touchpoints: pre-work, live rehearsal, post-implementation check-ins, and faculty feedback loops.
• Add trust measures to outcomes plans, such as confidence in when to use AI, when to escalate, and how to explain AI-supported work to colleagues or patients.
• Build cases around workflow roles: clinician, faculty preceptor, trainee, administrator, and support staff should not all receive the same AI instruction.
• Test simulations in which the AI output is plausible but wrong, incomplete, or operationally disruptive.

What to reconsider

Conference-linked oncology content can supply credible cases, especially where ASCO or NCCN material shapes local workflow. But the sharper move is not to convert conference updates into more AI sessions. It is to use those cases to test whether faculty can teach transparency, QA, and role accountability when AI touches real work.

For CME teams, the audit question is simple: if a learner completes your AI program, what will they do differently on Monday—and who else in the workflow has been prepared for that change?