The right questions to ask are "what is the incremental benefit I can get from ML over a simple rules-based system? How much does that added complexity cost?".

The right questions to ask are "what is the incremental benefit I can get from ML over a simple rules-based system? How much does that added complexity cost?". Costs include technical debt, increased maintenance, general opacity, and the risk that the complex model runs amok and does something stupid (which is more common than you might think). Sometimes those costs are justified, sometimes they aren't.