I would suspect that even AI programs for all of their calculating intelligence are left stumped if all of the irrational humans decide to all jump off the cliff of selling like lemmings all at once!
Indeed, there's an Android app called "Replika" that is an AI-driven chatbot. Purportedly it "learns" more about you and can give you insights that are pretty amazing, if you feed it enough info. I did run the query by the developers as to what would happen if one Replika bot started chatting with another one, with no outside input? Could they bootstrap each other into something at least interesting, if not profound? Nope. They started talking around each other and as Montgomery Scott told Captain Kirk, "we're goin' around in circles an' getting nowhere pretty fast!" So in short, AI may have some potential for some of the things some of the time, but clearly not all of the things all of the time. And some things are probably best for never of the time. The old sci-fi movie "Colossus: The Forbin Project" warns what happens if you give a big AI system control of all the nukes to remove human error in causing a potential WW-III. A cold war piece, it still plays pretty well, especially in light of our increasing interest in, if not increasing reliance on AI.