Bayesian belief nets (bbns) are becoming increasingly popular in modelling complex systems. They provide a high level graphical representation of the system in which the problem owner easily recognizes his/her problem. They incorporate the effects of proposed decisions in a natural and transparent way. Their application is however limited by the excessive assessment burden, which often leads to informal or unstructured quantification. In this paper we introduce various bbn models and contemplate their advantages and disadvantages in computing risks of complex systems. Moreover, we stress their usefulness as a decision support system, and in particular updating on the basis of possible observations. We illustrate different bbn models with recent applications in areas of occupational and aviation safety.