In the '90s, the Royal Meteorological Institute (RMI) of Belgium started to replace its conventional ''manual'' meteorological network by automated weather stations (AWSs). The meteorological measurement network is now fully automated. RMI counts 18 AWSs that made automated observations centrally available in our headquarters in Uccle, Brussels to internal as well as external users. Due to the large increase in the data amount associated with the automation, quality assurance (QA) procedures are being automated. However, human operators continue to play an essential role in the data validation processes. This contribution describes our newly developed semi-automatic quality control (QC) of 10-min air temperature data. After an existence test, the data are checked for limits consistency, temporal consistency and spatial consistency. At the end of these automated checks, a decision algorithm attributes a flag to each particular data. Each day the QC staff analyzes the preceding day observations in the light of the quality flags assigned by automated QA procedures during the night. It is the human decision whether or not a value is accepted.