It makes no sense: it depends on the product, really. Prof. Patrick Winston used to say, ten years ago: "Machine Learning does what we do deterministically, but badly". Now, scale and new techniques at this stage have enabled ML to achieve amazing outputs, in areas where we could not get similar results deterministically. That does not mean that it should be used for everything. You are not supposed to use hammers instead of screwdrivers just because hammers become fashionable.
You should probably clarify your concern.