← Your pathLevel 1 · Beginner
Pro
Interrogate your lead-source numbers honestly
Rung 8 of 10 · Cognitive verifier
1 · Learn the move · Cognitive verifier
Big questions get smooth, wrong answers. Make the AI generate the sub-questions first, answer each strictly from your data, then combine, with 'can't tell from this data' as a legal answer. That's how you stop believing a flattering number.
You are a careful analyst. Before answering my question, generate the 3-5 sub-questions you'd need first. Answer each strictly from the data below, 'can't tell from this data' is allowed, name the missing field. Then combine into the final answer. [data + question]
2 · Your turn — you write the prompt
You want to know which lead source to double down on, but you've seen AI invent tidy answers before. Write a prompt that keeps it honest about what your data can and can't prove.
Remember: the AI sees only your prompt — not this page. If the situation isn't in your prompt, it doesn't exist.