[Daishiman]

The promise of visual designers has existed for the past 30 years and it has always blown up spectacularly because everything that can be made to conform to canned design principles is already there, and everything else actually requires non-trivial knowledge that is not easily automated.

Design guidelines change too, and accommodating existing software to new designs is not automatable either.

And for data science? Simple correlations are trivial, but doing useful data science requires domain understanding of the business and products and factors such as seasonality, markets and so on.