Indoor thermal comfort has been the subject of several studies in the last years, aiming to create a satisfying work environment for the occupants inside a building. However, thermal comfort is strongly related to the occupants. Accordingly, it is necessary to take into consideration the user’s thermal preferences when adjusting the thermal conditions in buildings. This work aims to develop an innovative and smart solution, based on BIM, IoT, Machine Learning, and the user experience, to build a thermal comfort model. A BIM model was designed to ingest data streams from the IoT devices together with the user’s thermal preferences specified via a mobile application. The study area was divided into a three-dimensional grid where, for each voxel, the Predicted Mean Vote (PMV) was calculated. The occupant indicates his thermal sensation, noted as Personal Vote (PV), associated with his profile and his position. These measures were used to train a logistic regression model to find the correlation between the PMV and PV. Ultimately, an LSTM Deep Learning model was trained to forecast the optimal temperature of the study area in the next working hours. The results were visualized both on the mobile application and in the BIM environment.