The Goal: A Process of Ongoing Improvement, first published in 1984 by Eliyahu M. Goldratt with Jeff Cox, follows plant manager Alex Rogo as he tries to save a failing factory. He gets help from a former college professor named Jonah — a mentor who almost never gives Alex a direct answer. Instead, Jonah asks a question that makes Alex realize he was solving the wrong problem.
I recommended the audiobook to a few people for an unrelated reason: the dramatized production makes a forty-year-old factory story feel like overhearing a company unravel in real time. I didn't expect it to change how I thought about the AI conversations I was already having.
I said it as a joke, in a chat window, mid-way through a long back-and-forth about a forecasting problem that had nothing to do with the book: "you're like my Jonah." I remember feeling a little embarrassed even typing it — comparing a chatbot to a fictional physics professor felt like exactly the kind of overreach I try to avoid.
But the joke landed because it was accurate, and the accuracy is what stuck. The reason I'd stayed in that conversation for two hours instead of getting my answer and leaving wasn't that the model was fast. It was that it kept asking what I meant. What did I mean by "accurate"? Accurate compared to what baseline? Who was the forecast actually for, and what would they do differently if it were ten percent tighter?
By the third question I didn't have a good answer. By the fifth, I understood I'd been trying to solve a precision problem when the real issue was that nobody downstream had been consulted about what precision they actually needed.
That's the part of this I keep circling back to — not that AI is fast, everyone already knows that, but how much of my actual progress came from being asked better questions by something with no stake in protecting my ego, no meeting to get to, and no patience for an answer I hadn't earned.
Jonah's power in the book was never superior information. It was disciplined questioning. That's the role I try to keep AI in — a Socratic teammate, not an answer vending machine.
Current Hypothesis
The most useful AI interactions are the ones where it interrogates the premise of my question instead of answering it as asked.
Not a conclusion — a working idea, revisited as evidence comes in.