Abstract: The method and apparatus teach agents in a communication system, such as an automatic call distribution system. In a first step of one embodiment of the method, a previously stored call transaction having a plurality of dialog portions and decision points is provided to the agent. The call transaction is replayed, and at least one decision of the respective decision point in the call transaction is simulated. This simulation is different than an original decision at the respective decision point in the call transaction. The results are evaluated and presented to the agent. In one embodiment at least one decision is a best practice decision. Also, the method further may overlay preferred paths of a call transaction based on stored best practice call transactions. The apparatus implements the method for replaying call transactions.