(A)I Did That: Toward a Model of Leaders’ AI use and Disclosure Effects on Employees

Although leaders’ appropriate self-disclosure has long been viewed as an important aspect of leader communication, recent evidence suggests that when people disclose AI use in workplace communication, transparency backfires. As AI-supported leadership is likely here to stay, leaders face a dilemma: how can AI be disclosed while ensuring it is interpreted as appropriate and relationally attuned? Drawing on theorizing about leadership meta-talk, we shift focus from whether leaders should disclose AI use to reconceptualizing disclosure as a sensegiving act that can be framed in different ways to shape employees’ interpretations of leaders’ AI use. Specifically, we distinguish simple AI use disclosure from two meta-talk framings of disclosure: a task-focused explanation emphasizing productivity and a relation-focused explanation emphasizing fairness. We tested our theorizing in two studies. In an experiment ( N  = 207), task-focused meta-talk reduced employee trust relative to simple disclosure, whereas relation-focused meta-talk did not differ significantly from simple disclosure. A critical incident study ( N  = 335) offered complementary insights into how employees make sense of leaders’ AI use in practice, yielding a model showing how employees’ attributions shape relational interpretations and subsequent evaluations. Together, these findings indicate that meta-talk does not inherently lead to favorable interpretations, and some forms may even be detrimental.