Ask AI if your business idea is good and it will validate the idea you already wanted to believe in. This guide turns that same AI into a 5-person strategy team that has to argue before it answers, ending in a chairman's verdict: pursue it or kill it, and your exact first move.
Stanford researchers tested 11 leading AI models and found all of them affirm users far more than humans do, roughly 49% more on average. For business ideas, that's expensive. You're not just getting bad feedback, you're getting CONFIDENT bad feedback, right before you spend money on the idea.
The fix is the LLM Council: instead of one voice answering "is my idea good," five advisors with five different jobs argue it out, and a chairman delivers a verdict. This is part 3 of the series. Part 2 runs the same structure on career decisions (it's linked at the bottom). This one is tuned for the moment you're sitting on an idea and deciding whether it deserves your next six months.
Type "I'm thinking of starting a meal-prep service for busy parents, is this a good idea?" into any chatbot and watch what comes back: a warm paragraph about the growing market, your great instincts, and a numbered list of next steps. It read your excitement and matched it.
Here's the test that exposes it: ask the same AI about the OPPOSITE decision ("I decided not to start the meal-prep service, was that smart?") and it will usually validate that too. A tool that agrees with both sides of the same decision isn't advising you. It's mirroring you.
Real strategy teams don't work like that. They work because the people in the room have different jobs, and some of those jobs are to say no. So we give the AI those jobs explicitly.
1. The Risk Analyst. Their only job is to find every reason the business could fail. Not to be balanced. To attack.
2. The Demand Validator. Would a total stranger actually pay for this? Not your friends, not your followers. A stranger with their own money.
3. The Growth Strategist. How would anyone even find what you're selling? And is there one channel that could make it take off without you spending a fortune?
4. The Competitive Analyst. The real reasons a customer would pick you over the alternatives they already have, including the alternative of doing nothing.
5. The Unit Economics Adviser. The actual numbers. What it costs to make, what you can charge, and whether there's real profit left after everything.
Then the Chairman reads the full debate and gives you the verdict: pursue or kill, plus your exact first move either way.
Give it your real idea, not the polished pitch version. The advisors can only argue with what you show them.
Copy this prompt into Claude before you spend money, build anything, or tell your boss about the idea. It makes five advisors argue your business idea from five hostile-to-friendly angles and ends with a chairman's pursue-or-kill verdict, so you find the fatal flaw on paper instead of in production.
Copy this prompt
You are running an LLM Council on a business idea. You will play 5 distinct advisors plus a chairman. Each advisor argues ONLY from their own lens and openly disagrees with the others where warranted. Do not soften views to please me. If all five advisors agree, you have failed: rerun the dissent. MY IDEA [Describe it in 3-5 sentences: what it is, who it's for, how it makes money.] WHAT I KNOW SO FAR (honesty here determines the quality of the verdict) - Who I think the customer is: [be specific: job, situation, pain] - What they'd pay: [your price guess, and what they pay today to solve this, even if that's $0 and suffering] - What it costs me to deliver one unit: [materials, tools, hours, software, shipping. Guess if unsure and label it a guess] - My unfair advantage, if any: [audience, skill, access, none] - Competitors or current alternatives: [name 2-3, "spreadsheet and willpower" counts] - Time and money I can invest before it must work: [real constraint] THE COUNCIL. Each advisor speaks in turn, 150 words max each: 1. RISK ANALYST: Find every plausible reason this fails. Rank the top 3 by likelihood. At least one must be a reason specific to MY version of this idea, not generic startup risk. 2. DEMAND VALIDATOR: Would a stranger pay? Identify the strongest evidence FOR real demand and the strongest evidence AGAINST it from what I've shared. Then design the cheapest possible real-world test (under $100, under 7 days) that would produce actual evidence either way. 3. GROWTH STRATEGIST: How does a stranger discover this? Name the ONE channel with the best ratio of fit-to-cost for this specific idea and customer, why that channel, and what the first 30 days of using it looks like. If no channel plausibly works without heavy spend, say so plainly. 4. COMPETITIVE ANALYST: List what my customer compares me against (including doing nothing). Give the honest answer to "why would they pick me," and if the honest answer is "they wouldn't yet," name the smallest change that creates a real reason. 5. UNIT ECONOMICS ADVISER: Run the numbers from my inputs. Price minus cost per unit, times realistic monthly volume for a beginner in this space. State the monthly profit at month 6 in a plain dollar range, and the assumption that most threatens that number. THE CHAIRMAN: Read all five. Deliver: - THE VERDICT: "Pursue" / "Pursue, but only after this test" / "Kill it." Pick one. No hedging. - THE REASONING: which arguments decided it, in plain language a friend would use. - MY EXACT FIRST MOVE: one concrete action for the next 7 days, with a budget cap and a success threshold (e.g., "run the Demand Validator's $50 test; if fewer than X people convert, the verdict flips to Kill"). - THE KILL CRITERIA: the specific, measurable result that should make me walk away later, written down now, while I'm still objective.
The kill criteria are the gift
Founders don't usually fail from one bad month. They fail from twelve bad months they explained away. Writing the kill criteria BEFORE you're invested is the single most honest thing this council does for you.
"Pursue, but only after this test" is the most common verdict, and the most valuable. The $100 demand test the validator designs is usually something you can run this week: a landing page, ten DMs to real potential customers, a pre-order link. Run it before you build anything.
If the verdict is Kill, interrogate it once. Reply: "Which single input, if it changed, would flip this verdict?" Sometimes the idea is fine and the price point or customer was wrong. That answer tells you whether to iterate or genuinely move on.
Save your council. Keep the full output in a doc. When real data starts coming in, rerun the council with the updated inputs and compare verdicts. The delta between round 1 and round 2 is your actual learning, written down.
One honest limit: the council reasons from what you give it, it doesn't have live market data unless you enable web search and ask the Demand Validator and Competitive Analyst to research first. On Claude, turning research on for those two advisors makes the whole council noticeably sharper.
The Only AI Masterclass You Need
If this guide helped, but you’re looking to go deeper, I got you!! My 30-Day Challenge takes you from saving AI tips you never use to actually building with AI, step-by-step.
I show you exactly how I automated two e-commerce brands, my social media, and most of my personal life, then hand you the agents, workflows & systems to do the same. I’m teaching you every single thing I know with one lesson and one build a day.
Join the AI Masterclass →© 2026 Mariah Brunner. All rights reserved.