Statistics and Machine Learning -- When to Use What?

We live in a world of data explosion where computers are like a commodity, that's why associating almost every problem with trending tech buzzwords like "artificial intelligence", "machine learning", and "deep learning" seems like an avant-garde thing to do. It is almost intuitive and convenient to do so given readily available software and programming libraries on the internet. The most daunting part is probably to pick the suitable one and feed it with your data, then voilà -- here are the results. A search on Google with "machine learning models" swiftly returns you more than 700 million results in less than a second, just to show you how easy it is to gather on the magnitude of availability for ML models but how difficult it is to actually decide on the one that suits you. And then here come the million-dollar questions-- Am I actually selecting the best ML for my use case?