Machine learning (which can be classified as AI), doesn’t tell the story as well as other models can.

So if you were trying to understand something from the part, it would be better to use a model like linear regression.

Which people do use.

But to train a prediction, you use historical data. So every prediction is based on what was in the past, but for the most part, it’s not terribly decipherable. Simpler models are better for describing patterns for understanding.