Interesting framing. The distinction between "tool does one thing when asked" vs "tool keeps going until goal is met" is actually solid. Quite interesting on the MIT study: only 5% of AI investments returning positive ROI even though everyone claims they're already using it.