Modern call centers are highly complex queuing systems, in which there are several possible call types, and many agents groups. Due to this complexity, simulators are now preferred for their management than standard Erlang queuing models. Several studies have shown through data analysis that agents often have quite different speeds for processing a call type. However, this agent heterogeneity is often neglected, which is why most simulators have not been designed to use heterogeneous agents to process a service. In this work, we use simulation to study the impact of agent heterogeneity on the performance of call centers. We developed and integrated a module into the call center simulator, ContactCenter, which allows the use of heterogeneous agents to process a service. We have analyzed data collected from a real call center and have shown that agents have different processing speeds for handling a call type. We have modeled the distribution of arrivals, service times and patience times that will enable us to simulate this call center. Simulation result show that call center performance on a given day depends largely on the agents chosen to answer the calls.