This article examines the possibility of using the mathematical apparatus of generalized networks to model and study the behavior of an intelligent agent - broker in a virtual environment. In the proposed model the agent-broker is presented with a generalized net model. The transitions cover the broker's knowledge of the sequence of stages of buying a property. Conditions describe its knowledge of the opportunities and risks that exist at every step of this process. The tokens represent the consumers who are looking for the most suitable way for them to buy a property. They are implemented with a knowledge base for their: financial status; for the risks they consider acceptable or unacceptable; for the compromises they may or may not make; for their preferences and goals. The transition of the tokens through the generalized net model allows to observe the process of interaction between the model of the expert broker and the models of the client agents.