[D] resources for avoiding common mistakes in machine learning projects?

There are often managers or entrepreneur types that think of "AI" as a magical solution and then millions of dollars get wasted because they don't understand how fragile the solutions can be, how do much more data is needed to create solutions that generalize, susceptibility to bias, training data needed, etc. What accessible information is there that would help those people understand what is involved for practicality of project and successful and ethical outcomes?