With the widespread adoption of artificial intelligence (AI) in organizations, employee-AI collaboration has become a crucial mode of task accomplishment. Existing research has primarily focused on its impacts on task-oriented outcomes such as performance, innovation, and procrastination, while paying insufficient attention to how such collaboration reshapes employees’ interpersonal behaviors in the workplace, particularly competition. To address this gap, our study focused on the core question of how human-AI relationships shape interpersonal relationships. Drawing on the dual strategies theory of social hierarchy, we developed a serial moderated mediation model, aiming to explore how and when employee-AI collaboration affects employees’ covert competitive behaviors toward colleagues. Using a three-wave survey of 302 Chinese employees, the results reveal that, on the one hand, employee-AI collaboration enhances individuals’ role breadth self-efficacy, which subsequently strengthens their tendency to adopt prestige-based strategies, thereby promoting covert competition. On the other hand, employee-AI collaboration also elicits a sense of identity threat, which suppresses employees’ use of dominance-based strategies and consequently leads to an increase in covert competition. Furthermore, the study demonstrates that leaders’ AI symbolization not only positively moderates the effects of employee-AI collaboration on role breadth self-efficacy and identity threat, but also amplifies the serial mediation effects by reinforcing these underlying mechanisms. Our study shifts the focus of employee-AI collaboration research from task-related outcomes to interpersonal relationship-oriented outcomes. We enrich the application of the dual strategies theory of social hierarchy in the AI era and provides practical guidance to manage AI-driven changes in workplace interpersonal relationships.