Predictive Maintenance Question (binary classification)

I have a question regarding Predictive Maintenance: in this tutorial here: https://docs.microsoft.com/en-us/learn/modules/predictive-maintenance-model-builder/3-choose-scenario-data

It says:

Choosing a scenario for predictive maintenance

Depending on what your data looks like, the predictive maintenance problem can be modeled through different tasks. For your use case, because the label is a binary value (0 or 1) that describes whether a machine is broken or not, the data classification scenario is appropriate

Now, how can this be used to predict machine failure BEFORE it gets broken? The tutorial approach is to monitor a set of parameters, and at some point the machine will break .. and I'll only then know that.

Am I missing something?