In an increasingly competitive environment, organizations need to continuously innovate (explorative learning) while making steady improvements to their existing operations (exploitative learning). The capacity to pursue both exploratory and exploitative learning simultaneously is called ambidexterity. Therefore, ambidexterity has become one of the important research topics in the field of organizational study. This paper focuses on ambidexterity learning at the team level. The objective is to propose a generic agent-based simulation model that can be used to examine how ambidextrous learning affect team performance under different levels of task complexity, communication intensity and communication cost. The experiment shows that the model can reproduce what have been reported in the team performance literature.