A practical guide to establishing an Action Learning Set focused on ethical AI:

Recruit 4–8 members
Sets work best with enough diversity to generate genuinely different perspectives, but small enough to allow deep individual focus. Aim for diversity of role, background, and experience with AI ethics. Consider including people from outside your organisation: external perspectives often ask questions that insiders cannot.

Establish a voluntary, confidential contract
Members join voluntarily and commit to confidentiality about what is shared in the Set. This psychological safety is essential for genuine inquiry.

Meet regularly over a sustained period
Monthly meetings over 6–12 months allow genuine learning to develop. One-off workshops are not Action Learning.

Commit to the questioning discipline
The hardest and most important norm to establish is the commitment to questioning rather than advising during individual focus time. Name it explicitly. Practice it. Reflect on it.

Bring real problems
Action Learning only works if the problems are real: genuinely important, currently unresolved, involving genuine uncertainty. Theoretical case studies are useful for other purposes. Action Learning requires the discomfort of real stakes.

Act between meetings
Learning in Action Learning comes from acting on what has been explored in the Set, observing what happens, and bringing that experience back to the next meeting. Without action, there is no learning: only conversation.

Reflect on the Set’s own practice
Regularly devote time to reflecting on how the Set is working. What kinds of questions are most generative? What patterns are emerging in members’ challenges? What is the Set not yet talking about?