The paper introduces an Agent-Based Model (ABM) of the Anti-Money Laundering (AML) system. The model replicates rule-based AML system, including details of transactions of financial institution’s customers, specification of AML systems rules, heterogeneous structure of internal customers and external parties. The latent behavior of individuals involved in money laundering is represented with a utility-based agent capable of adjusting transaction patterns to maximize the throughput of illegal transfers passing through the financial institution’s accounts. We present results of two experiments to demonstrate the use cases of the model in the context of operational excellence of the AML system deployed within a financial institution. The first simulation experiment shows the changing behavior of the laundering individuals in response to increased investigation capacity. The second experiment reveals a zonal decrease in the quality of the AML system due to the emergence of a new financial market.