Abstract
As AI is expected to take on the role of providing workplace feedback to employees in the future, understanding how AI and humans can complement effectively in this context is crucial. This study explores this through a randomized controlled experiment testing the feedback effectiveness of two different Chat-GPT-based AI chatbots-one that delivers positive feedback focused on individual strengths (n = 54) and the other that provides negative feedback aimed at areas for improvement (n = 48). Japanese office workers, including freelancers, participated in this study. We conducted mixed-effects model analysis to assess whether the effects of these AI chatbots on occupational self-efficacy are moderated by the level of social support from supervisors and colleagues. The results showed that positive feedback from the AI chatbots significantly enhanced occupational self-efficacy. Furthermore, negative feedback also boosted self-efficacy, but only when employees perceived a high level of emotional support from their supervisors and colleagues. These findings point to effective ways of integrating AI into the workplace, where humans and AI can assume complementary roles in employee development.