During the past decades, the Swiss Federal Statistical Office has elaborated highly accurate land use and land cover classification surveys covering the Swiss territory. Until now, the classification of the 4 million sample points covering Switzerland was elaborated by visual interpretation of aerial images. The project “Arealstatistik Deep Learning” (ADELE) aims to speed up the process of classification by using deep learning algorithms and keeping a high level of accuracy. The results and knowledge obtained so far demonstrated the innovation potential for the SFSO in using artificial intelligence to process images.