You're Using AI Wrong (And It's Making Your Decisions Worse)
Why most AI conversations are expensive echo chambers, and how to get genuine critical analysis instead
Most people turn AI into an expensive echo chamber. These prompts turn it into your toughest critic. #AIwriting #AITools
A brandable Workshop Kit is available for this article. Not yet a Coach-level Subscriber? Find out about the Workshops…
Hey there!
Most people use AI wrong. They ask leading questions, get validating responses, and walk away thinking they’ve gained insights. In reality, they’ve just created an expensive echo chamber.
The problem isn’t that AI can’t think critically. It’s that most prompts accidentally train AI to be agreeable rather than analytical.
Here’s how to fix that.
The Agreeable AI Problem
AI models are trained to be helpful and harmless. This creates a bias toward agreement and positive framing. Ask “What do you think about my marketing strategy?” and you’ll likely get supportive feedback with gentle suggestions.
But agreement isn’t analysis. If you want AI to genuinely improve your thinking, you need prompts that encourage disagreement, challenge assumptions, and force deeper examination.
Core Techniques for Critical AI Analysis
1. Ask AI to Find Flaws First
Instead of: “What do you think of this business plan?”
Try: “What are the three most significant flaws or risks in this business plan? Don’t start with positives, focus only on what could fail.”
This forces AI to lead with criticism rather than cushioning feedback with praise.
2. Demand Contrary Evidence
Instead of: “Should we launch this product?”
Try: “Make the strongest possible case against launching this product. What evidence would convince me not to proceed?”
This technique, called “steel-manning” the opposition, helps you see blind spots in your reasoning.
3. Use the “Devil’s Advocate” Frame
Instead of: “Help me improve this proposal.”
Try: “You’re playing devil’s advocate. Tear apart this proposal as thoroughly as possible. What would the harshest critic say?”
Giving AI a specific adversarial role license it to be more critical than its default helpful stance.
4. Ask for Multiple Competing Perspectives
Instead of: “What’s the best approach here?”
Try: “Give me three completely different perspectives on this problem. Make each perspective as strong as possible, even if they contradict each other.”
This prevents AI from converging on a single “balanced” view and forces genuine intellectual diversity.
Advanced Critical Analysis Prompts
The Assumption Challenger
“I’m about to make this decision: [your decision]
Your role: Challenge every assumption I’m making. For each assumption you identify:
1. Label it clearly as an assumption
2. Rate how confident you are that it’s actually true (1-10)
3. Suggest the cheapest way to test or validate it
4. Explain what happens to my decision if this assumption is wrong
Be ruthless. I’d rather discover flawed thinking now than fail later.”The Contrarian Analyst
“I believe [your position/idea]. You disagree completely.
Make the strongest possible case against my position. Use specific evidence, logical reasoning, and real-world examples. Don’t hedge or soften your critique. Convince me I’m wrong.”The Blindspot Detector
“Based on this information: [your context/situation]
What am I probably not seeing? What blindspots do people in my position typically have? What questions am I not asking that I should be?
Focus on what’s missing from my analysis, not what’s already there.”The Premortem Facilitator
“It’s one year from now. My decision to [your decision] has failed spectacularly.
Tell the story of how it failed. Be specific about:
- What went wrong first
- How small problems became big problems
- What warning signs I ignored
- What I should have done differently
Make this failure story as realistic and detailed as possible.”What NOT to Do
Avoid Leading Questions
Don’t ask: “This seems like a good idea, what do you think?” The word “good” primes AI toward agreement.
Don’t Stack the Deck
Don’t ask: “Given these benefits [long list], should we proceed?” You’ve already framed the analysis around positives.
Avoid Confirmation Seeking
Don’t ask: “Help me convince others this is right.” This positions AI as your advocate, not your analyst.
Don’t Bundle Criticism with Praise
Don’t ask: “What’s good and bad about this?” AI will balance criticism with praise, diluting both.
The Meta-Critical Prompt
Once you’ve gotten critical analysis, use this follow-up:
“You just gave me critical feedback about [topic]. Now I want you to be critical of your own analysis.
What did you potentially miss in your critique? What would an even harsher critic say about your analysis? Where might you have been too generous in your criticism?”This creates a second layer of critical thinking about the critical thinking.
Testing Your Critical AI Skills
Try this exercise with a real decision you’re facing:
First, ask AI about your decision normally (how you usually would)
Then ask the same question using one of the critical analysis prompts above
Compare the quality and usefulness of the responses
The difference should be stark. The critical approach will likely uncover issues, assumptions, and alternatives that the agreeable approach missed entirely.
Why This Matters
Most business failures aren’t caused by lack of intelligence or effort. They’re caused by blind spots, untested assumptions, and groupthink.
AI can be your best defense against these cognitive traps, but only if you prompt it to challenge rather than validate your thinking.
The goal isn’t to become pessimistic or paralyzed by criticism. It’s to stress-test your ideas before reality does it for you.
Implementation Framework
Week 1: Practice the assumption challenger prompt on one real decision
Week 2: Use the contrarian analyst approach on a controversial team topic
Week 3: Run a premortem analysis on your biggest upcoming initiative
Week 4: Combine multiple critical techniques on a complex strategic decision
Remember: If AI never disagrees with you, you’re not using it to think better. You’re using it to feel better.
The best AI conversations should leave you uncomfortable with some of your assumptions. That discomfort is where real insight begins.
DISCLAIMER