The paper presents a hybrid method of risk analysis in the complex systems predicting the possible accident development associated with the social systems, as well as recommendations in prevention of such accidents. The proposed method in order to determine operational state of a complex system and endow it with additional ability to withstand failures combines system dynamics models (to help in identifying interactions of the elements of the system under study in dynamics), event and failure tree models (used to simulate the risk scenario evolution) and artificial neural networks. The hybrid risk management methodology in sociotechnical systems is based on combining capabilities of different artificial intelligence technologies and makes it possible to introduce advantages of several technologies by integrating them. Six stages of research carried out within the framework of hybrid technique are presented, as well as mathematical description of the neural network model. Effectiveness of the proposed methodology was tested using three implemented software products. On the example of a construction company and using the developed original software package, accident scenarios were simulated, and a neural net-work was built to predict risks and determine the company operation status. Simulation results are provided