The paper considers the application of stochastic modelling to building energy management systems. The results presented relate to research conducted on a full-scale experimental test room with its own dedicated HVAC plant. It is shown that the digital models deduced from correlating the system input/output and climatic data can be used not only to accurately predict the thermal behaviour in the room, but also to allow for the effects of occupancy. In particular, the effects of altering futniture and fittings within the occupied spaces are discussed.