One of the efficient methods of business process simulation is the use of a multi-agent system. As on other cases, such simulation requires sufficient input data. However, in the case of business systems, real business data is not always available, therefore multi-agent systems often operate with randomly (or pseudo randomly) generated parameters. This method can also allow for unpredictable phenomena both inside the simulated system and in the environment. The motivation of this paper is to introduce novel business process simulation method based on multi-agent system and to validate the simulation results with certain randomly generated agent parameters. Normal distribution was used to model two agent parameters in the simulation. These parameters are the amount of goods to buy and the ability of the seller agents. Certain number of negotiating agents representing various important roles within the company business system is described. The obtained results demonstrate that using random number generation function from normal distribution can lead to proper simulation results in comparison with real data. This partial research is a last step to validate the innovative implementation of a decision support tool.