It has always been a topic of great interest in air transport management to be able to estimate controller workload. So far, research has not had the opportunity to make use of real data on the controller’s actions. We have enough data to be able to use machine learning methods. The aim of this work is to predict the controller’s actions to know his workload. Several machine learning models were tested to try different combinations of features and the selected algorithms and two models were finally chosen. The predictions provided by the models are good enough to be used when a first approximation of the workload in a sector is to be obtained. Finally, explainability techniques were employed to discover the patterns found by the AI in the machine learning models. Thanks to these techniques, we can build a profile of the critical flights that increase the workload the most.