When humans collaborate with AI: critical reflection as a pathway to learning from innovation failures

Purpose Although learning from innovation failures (LFIF) is essential for sustaining innovation and competitive advantage, individuals often struggle to overcome the cognitive and motivational barriers that impede deep reflection on failure. Drawing on reflection theory, this study conceptualizes AI as a reflection partner that reshapes how individuals interpret and learn from failures. On this basis, the study examines how human–AI collaboration facilitates LFIF through critical reflection and under what organizational contextual conditions this effect holds. Design/methodology/approach This study comprised a three-wave longitudinal survey of 221 product managers across diverse regions and industries in China. This temporal design mitigated common method bias and captured the sequencing of the proposed mediation process. Findings The results reveal that human–AI collaboration fosters LFIF indirectly through critical reflection. Specifically, human–AI collaboration stimulates alternative interpretations and challenges assumptions, thereby deepening reflection and enabling LFIF. Moreover, an organizational proactive climate strengthens the effect of human–AI collaboration on LFIF via critical reflection, while an organizational competitive climate weakens this indirect effect. Originality/value This study extends research on human–AI collaboration from routine to non-routine learning, theorizing AI as a reflection partner that reshapes higher-order cognition. Moreover, it identifies critical reflection as a central cognitive mechanism linking human–AI collaboration to LFIF, providing a new theoretical perspective for understanding LFIF in the AI era. By bridging technological, cognitive and contextual perspectives, this study elucidates when and how human–AI collaboration enables effective LFIF.