In probabilistic models, there needs to be a way to decide how to make the final decision. For example, when using classification models that return a probability for each class, a decision rule needs to be set to choose which class to assign an instance to. This may have implications on the use case.
Typically, the simplest approach is to select the class with the highest predicted probability. However, in some cases, a decision rule needs to be more sophisticated. E.g. in a medical setting, it is more costly to make false negatives (e.g. predicting “not cancer” when it actually is cancer) than false positives.