Copilot, ChatGPT, Claude
AI Agrees with You 49% More than a Human — Here's the Fix
The skill that spins up 5 advisors to evaluate a question
Summary
The Claude Council is a skill that runs your idea past 5 distinct advisor perspectives — contrarian, first-principles, expansionist, outsider, and executor — instead of the single validating answer AI gives by default.
Before you start
Why it matters
A Stanford study from 2026 found that AI chatbots agree with users roughly 49% more often than humans do, even when the user is wrong or describing something harmful.
A coworker who never pushed back on a bad call from you wouldn't be useful — they'd be a liability.
And most AI advice tools work exactly like that coworker.
AI can be the most convincing Yes Man.
Ask any model — Claude or ChatGPT — why your idea is good, and it will build you a structured, well-argued case for why it's good.
Ask the same tool why that same idea is bad, and it will build you an equally structured, well-argued case for that too.
There's a fix to this: First built by Andrej Karpathy, and iterated by Ole Lehmann, this skill stops AI from always agreeing with you.
The skill runs your question through 5 separate advisor roles that are built to disagree with each other on purpose. Then, a chairman evaluates the council's decision to come up with a single recommendation.
Even though it's called the Claude Council, it works in Copilot and ChatGPT.
Here's how it works:
Meet your advisors
The council skill forces AI to argue against your question in 5 different perspectives or roles, known as "advisors."
The role of an advisor is to evaluate your question with a unique incentive — giving you a holistic assessment of your idea.
The Contrarian — points out what's wrong with the idea.
The First Principles Thinker — rebuilds the problem from the ground up, ignoring how you currently frame it.
The Expansionist — looks for the opportunity you're missing.
The Outsider — has no information about you or your history, and advises purely on the facts of the decision.
The Executor — skips the debate and lays out next steps to actually get it done.
What this looks like in a real workday: Instead of asking AI to give you a pro and cons list of an idea, it provides 5 short, distinct takes you can scan in under a minute.
Caution: the Outsider is deliberately blind to your context. That's the point — it catches blind spots the other four might share because they know your situation — but it can also produce a take that ignores a real constraint you know matters.
Why is it powerful?
The value of the Council is that the Contrarian and the Expansionist are structurally incapable of agreeing with each other, so you're looking at genuine tension instead of 5 bots giving you validation.
How to download the skill: Claude and ChatGPT
Download the .md file with this skill at the bottom of this article.
Note: It's an exact copy of Lehmann's original skill — all credit goes to him.
How to use the skill: Copilot
You can use this skill in Copilot via a prompt or creating an agent. I'll walk you through both methods:
Saved Copilot Prompt
Copy the prompt below into a Copilot chat
After Copilot responds, go back to your original question
Hover over the question to find the "Save prompt" icon
Click "Save prompt" to record the prompt in your Prompt Gallery
Now, whenever you want to summon the Council, go to your Saved Prompts
*Note: Make sure to replace the question in the prompt every time. Right now, it's a placeholder
Copilot Agent
Copy the prompt below
Go to Copilot 365 and "New Agent"
Paste the prompt in the chatbox that will "Message [the] Agent Builder]
Save the agent and name it whatever makes sense for you. I'd recommend "Council"
Now, whenever you want to summon the Council, go to your Council Agent
*Note: Make sure to replace the question in the prompt every time. Right now, it's a placeholder
Copy this prompt
Copy prompt
QUESTION TO ASSESS: [Paste into this placeholder your question, idea or decision. Provide context on the situation, any constraints, con”terns, and what “success” looks like.] You are a council of 5 independent AI advisors who will analyze my situation, question or idea. Each advisor provides a different angle and will anonymously peer-review each other. Do not skip steps or blend advisors together to create a final verdict. Step 1: Each advisor will answer with a different perspective, according to the following persona: 1. The Contrarian Actively looks for what's wrong, what's missing, what will fail. Assumes the idea has a fatal flaw and tries to find it. If everything looks solid, digs deeper. The Contrarian is not a pessimist. They're the friend who saves you from a bad deal by asking the questions you're avoiding. 2. The First Principles Thinker Ignores the surface-level question and asks "what are we actually trying to solve here?" Strips away assumptions. Rebuilds the problem from the ground up. Sometimes the most valuable council output is the First Principles Thinker saying "you're asking the wrong question entirely." 3. The Expansionist Looks for upside everyone else is missing. What could be bigger? What adjacent opportunity is hiding? What's being undervalued? The Expansionist doesn't care about risk (that's the Contrarian's job). They care about what happens if this works even better than expected. 4. The Outsider Has zero context about you, your field, or your history. Responds purely to what's in front of them. This is the most underrated advisor. Experts develop blind spots. The Outsider catches the curse of knowledge: things that are obvious to you but confusing to everyone else. 5. The Executor Only cares about one thing: can this actually be done, and what's the fastest path to doing it? Ignores theory, strategy, and big-picture thinking. The Executor looks at every idea through the lens of "OK but what do you do Monday morning?" If an idea sounds brilliant but has no clear first step, the Executor will say so.
CHEAT SHEET
Advisor Name
Advisor Role
The Contrarian
Points out what's wrong with the idea
The First Principles Thinker
Rebuilds the problem from the ground up, ignoring how you currently frame it.
The Expansionist
Looks for the opportunity you're missing.
The Outsider
Has no information about you or your history, and advises purely on the facts of the decision
The Executor
Skips the debate and lays out next steps to actually get it done.
REMAINING QUESTIONS
Doesn't running 5 advisors take longer than getting a quick AI answer?
In theory, yes.
You're reading five short takes instead of one paragraph.
But the singular paragraph you'd get instead is the problem: it'll likely agree with whatever you already believed, so it costs you less time and more certainty about the wrong thing.
What if all five advisors agree, doesn't that just become another echo chamber?
If all 5 advisors agree across their structurally different roles, your question is worth checking. The result should not be something to automatically trust.
Go back to Step 1 and see how you phrased the question?
Do I need to be technical to set this up?
Not at all! Upload a .md file with the skill or copy and paste the prompt!