Organizational unlearning, defined as intentionally discarding established routines, benefits further learning and improves R&D performance in organizations. However, compared to its beneficial consequence, our understanding of ways to facilitate organizational unlearning remains unclear. Noting that individuals are central not only to the emergence of routines but also to their endogenous evolution over time, we investigate how the replacement of a member may affect organizational unlearning. We simulate an R&D organization in which members closely collaborate with each other and take interdependent actions. We examine the properties of the replacement by disentangling two variations—new ideas and new beliefs—that may be introduced into the organization, and using organizational memory and constancy of action as two key metrics to uncover the temporal evolution dynamics of routines. Our results suggest that both variations can facilitate organizational unlearning but through different means: while idea-related variations prompt individual deviations in actions to alter the perceived payoffs shaped by interdependence and then modify members’ formed beliefs regarding old routines (i.e., action?belief), belief-related variations induce direct changes in members’ formed beliefs, enabling them to reduce their commitment to existing actions (i.e., belief?action). When the alternative action is believed to be more effective than the one which it replaces, organizations may be inclined to repeatedly take this deviate action and allow the emergence of the new routine; otherwise, unlearning may be followed by relearning (i.e., ending the trial and reverting to the old routine). Therefore, organizational unlearning does not necessarily lead to better performance outcomes. Furthermore, idea- and belief-related variations may jointly facilitate organizational unlearning more and reduce the reversion to the old routines more. This effect becomes even stronger when the replacement involves a new hire with higher influence.