The Prompt Lab — Socratic Reversal

Socratic Reversal

The Technique

Instead of asking the model to answer your question, you ask it to generate the questions you should be asking before an answer is even attempted. This works because LLMs trained on expert reasoning know that most problems are mis-specified — the real value is often in reframing the problem, not solving the stated one.

The Naive Prompt

I'm a solo founder about to have my first call with a potential enterprise customer 
(a 400-person logistics company). What should I say to close the deal?

Why It Falls Short

The model jumps straight to tactics — pitch structure, objection handling, pricing anchors — without knowing whether the founder has validated the pain, understood the buying committee, or confirmed the company is even closeable. The output is confident but generic, and may actively point the founder in the wrong direction. You get an answer to the wrong question.

The Improved Prompt

I'm a solo founder about to have my first call with a potential enterprise customer 
(a 400-person logistics company). Before you give me any advice on what to say, 
generate the 6 most important questions I should be able to answer about this 
situation before getting on that call. For each question, explain in one sentence 
why getting it wrong would hurt me.

Why It Works

The model now acts as a diagnostician rather than an advisor, surfacing assumptions the founder hasn’t examined — Who is the economic buyer vs. the champion? Is there an active budget cycle? What’s the cost of inaction for them? This forces the founder to either answer those questions (producing a much richer follow-up prompt) or realize they’re going into the call underprepared. The technique converts a “give me answers” prompt into a structured self-audit.

When to Use This

• High-stakes decisions with hidden dependencies — enterprise sales calls, architecture choices, hiring decisions — where acting on surface-level advice is risky and the framing matters as much as the answer.
• When you’re early in a problem and don’t yet have enough context to know what a good answer even looks like; Socratic Reversal is especially powerful with reasoning-heavy models like GPT-5.5 Thinking or Claude Opus 4.8, which will generate genuinely probing questions rather than shallow checklists.
• Breaking out of tunnel vision — if you’ve been staring at a problem long enough that you’ve stopped questioning your premises, asking the model “what should I be asking?” is a reliable way to get a perspective you haven’t considered.