Answers are grounded in this paper and cite their sources.
This Account of Practice reflects on a Chinese initiative in which generative AI was embedded in the problem-clarification stage of action learning. In 2025, within a chief product manager development programme at H3C (New H3C Technologies), the group first conducted one round of open human questioning around the real problem of a wireless product-line head (Jeff, a pseudonym), and then introduced an AI ‘Question Master’ to run three structured rounds—Pain, Constraint and Blind-spot—together termed Pain–Constraint–Blind-spot (PCB) questioning. The paper presents the precise questions, the owner's in-the-moment reactions and the gains of each round. It traces how each round progressively redefined the problem—and then the problem owner's mindset: Pain and Constraint clarify the facts and root causes, while the Blind-spot round illuminates the owner's role and mental models, shifting learning from the task to the person. From this we derive principles of ‘people before machine, questions not answers, and safeguarding the learning space’. We also outline a suite of Action Learning AI agents spanning topic selection, interviewing, reframing, solutions and feedback, offering transferable experience for human–AI collaboration.
No takes yet. Share an insight, caveat, or question.
Changjun et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: