Classic AI works on a statistical principle and the correlation between factors, considering that what happened will happen again exactly the same. It provides general probabilities on populations of events. Conversely, causal AI identifies the causal links between factors and builds models that can answer specific questions in the context of very specific factors or indicators. It makes it possible to estimate probabilities, to quantify confidence levels. Above all, it allows, on the basis of Bayesian logic, to verify and quantify intuitions and to compare them in their correct dimensions. It thus makes it possible to play scenarios by changing hypotheses, for the past (counterfactuals) or for the future (predictive).