Abstract Building energy prediction is a key factor to assess the energy performance of commercial buildings, identify operation issues and propose better operating strategies based on the forecast information. Different models have been used to forecast energy demand in buildings, including whole building energy simulation, regression analysis, and black-box models (e.g., artificial neural networks). This paper presents a different approach to predict the energy demand of commercial buildings using case-based reasoning (CBR). The proposed approach is evaluated using monitored data in a real office building located in Varennes, Quebec. The energy demand is predicted at every hour for the following 3 h using weather forecasts. The results show that during occupancy, 7:00–18:00, the coefficient of variance of the root-mean-square-error (CV-RMSE) is below 13.2%, the normalized mean bias error (NMBE) is below 5.8% and the root-mean-square-error (RMSE) is below 14 kW. When the statistical criteria are calculated for all hours of the day, the CV-RMSE is 12.1%, the NMBE is 1.0% and the RMSE is 11 kW. The case study demonstrates that CBR can be used for energy demand prediction and could be implemented in building operation systems.