After every blast, rock mass classification is performed. However, no person can see beyond the face. There are methods commonly used for this purpose: core drilling, measurement while drilling and probe holes. In this project, machine learning techniques (Artificial Intelligence), were applied to geotechnical information from probe hole drilling and face mappings, in order to find patterns and infer functions based on data used for training. Information had to be organized to be accessed by the machine learning method. The model learns from training data; it is understood that non-experienced situations cannot be predicted. Because of this, it was assumed that every project would need its own training. Data from a testing tunnel was considered. Once the model was trained, it was used as a forecasting tool during the performance of new Probe Holes. Results show that the model has an accuracy of +85% forecasting rock mass classification.