Most office workers work in open-plan offices. Distraction caused by irrelevant speech and lack of speech privacy are very typical problems in this work environment. Proper room acoustic design reduces these problems. However, the effectiveness of a specified room acoustic solution is difficult to predict exactly, because the most important room acoustic variables, i.e. ceiling absorption, furnishing absorption, screen height, masking sound level, speech effort and room dimensions, interact in a very complex way. The aim of this study is to present a model that can be used to predict the spatial decay of speech. The model represents a situation in which speech is produced at a single workstation and decay of speech is evaluated at different distances. The regression model is based on the measurement data of 16 different open-plan offices [9]. It takes into account the above-mentioned room acoustic variables. The predicted results are spatial decay curves of the A-weighted sound pressure level of speech and the speech transmission index, STI. The curves are used to determine the single-number quantities: the spatial decay rate of the A-weighted sound pressure level of speech, D2,S, the A-weighted sound pressure level of speech at 4m from the speaker, LA,S,4m, and the distraction distance, rD. The prediction accuracy of the single-number quantities was determined in 10 new offices. The prediction errors of D2,S, LA,S,4m and rD were less than 1.5dB, 3.0dB, and 2.5m in most cases, respectively. This accuracy is sufficient for most practical purposes. The model was programmed into a prediction tool, which is freely available on the internet and delivered with the electronic version of this journal.