The problem of the study is to increase the efficiency and necessity of implementing the KPI decision support system in the agro-industrial enterprise. In this regard, this article is aimed at describing the process of developing the KPI system. Bayesian networks are the main method of studying the KPI system. The quality control subsystem in the KPI system consists of a list of the work performed by the machine operators. For the evaluation of each work, the KPI indicators are selected. The choice of the indicator depends on the type of work performed and the type of culture. Assess the selected indicators of a separate work performed is not enough, it is important to consider the dependencies between the indicators. Many indicators have probabilistic characteristics, as a result of which the Bayesian network was chosen to take into account interrelated indicators. The article gives an example of using a network to account for the interrelations of indicators. This approach to the development of decision support systems allows us to obtain an adaptive Bayesian network setup procedure that is distinguished by the ability to change the vertexity of the network with the recalculation of the probabilities of influence on the final quality assessment. The structure of information display when issuing a waybill is obtained, differing in the representation to the dispatcher of the possibility of searching for an employee based on the Bayesian network of an average percentage of the machine operator's efficiency. The developed mathematical apparatus can be used in the agro-industrial enterprise, which requires the implementation of the KPI decision support subsystem.