Although the author has fair results with his given case, the author is mistaken the use of AI/ML in such scenario. In the example, they make the decision of "We should send emails to people who did 'case a'.". This is a pure 'instinct' by the decision maker. But in AI/ML case, this would be learnt from the feedback of the click rates etc. Naturally, decision maker becomes the AI, which actually can find interesting scenarios and exploit these behaviours to increase the desired outcomes.