Subtitle: Stop using AI to confirm what you already believe. Start using it to question what you think you know.
The Moment I Realized I Was Doing It All Wrong
I was sitting in front of my screen with ChatGPT open, about to make a hiring decision. I’d already decided in my gut who I wanted to hire. What I was actually doing with AI? Asking it to validate my choice. “Give me 5 reasons why hiring for cultural fit is important.” I wasn’t searching for truth. I was searching for confirmation.
It hit me hard. I’d spent months talking about AI as a thinking partner, a decision-making tool, and the whole time I was using it as a smarter search engine for answers I’d already decided on.
That morning changed how I approach AI entirely. And if you’re like most entrepreneurs I talk to, it’s about to change yours too.
Here’s what I learned: The smartest thing you can do with AI isn’t faster research. It’s pre-mortems on your own thinking. Before you commit to the big decision—the hire, the pivot, the launch, the partnership—you challenge your own assumptions with an AI tool that won’t let you hide from yourself.
This single habit has become one of the highest-leverage uses of AI available to anyone building a business today.
The Real Problem: You’re Not Asking Better Questions
Most entrepreneurs are using AI wrong, but not the way you might think.
They’re not using it wrong because they don’t know how to write good prompts. They’re using it wrong because they’re treating it like an answer machine instead of a thinking partner.
Here’s the pattern: You have a decision to make. You open your AI tool. You ask it questions designed to solve your version of the problem. And here’s the sneaky part—your version is almost always shaped by what you already believe. You’re not discovering truth. You’re building a case for a conclusion you’ve already reached.
The research on confirmation bias is damning. Psychologists have long understood that people “search for, interpret, favor and recall information in a way that confirms or supports one’s prior beliefs.” It’s not that you’re dumb. It’s that your brain is built to protect the story you’ve already decided to believe. And when you’re under pressure, racing to make a decision, your brain does this faster, not slower.
So you open ChatGPT or Claude. You ask it the question you’ve already framed. It gives you a smart answer. You feel validated. You move forward.
And three months later, you’re wondering why the hire didn’t work out. Why the product nobody asked for didn’t sell. Why the partnership became a nightmare.
The problem wasn’t the information. The problem was the assumption underneath the question.
Most of the biggest mistakes I’ve seen entrepreneurs make aren’t information problems. They’re assumption problems. The founder assumed customers wanted what they built. The hiring manager assumed cultural fit was the right criteria. The partnership seemed solid because both parties assumed they wanted the same outcome.
These aren’t knowledge gaps. These are blind spots.
And that’s where most entrepreneurs waste AI. They use it to fill knowledge gaps, when what they actually need is a mirror to see the blind spots.
Why Unchallenged Assumptions Cost You Everything
Let me show you what the research actually says.
CB Insights analyzed hundreds of startup failures and found that 42% of failed startups died because there was no market need for what they built. Let that sink in. Not because the product was badly made. Not because the team wasn’t talented. Because the foundational assumption—that people wanted this thing—was never tested.
The startup Webvan is a brutal case study. In 1999, at its peak, the company was valued at $1.2 billion. It looked like the future. Webvan was going to revolutionize grocery delivery. Every investor believed it. The team believed it. The assumption was solid.
Two years later, Webvan filed for bankruptcy. The company laid off 2,000 people. It closed completely.
What happened? The core assumption was wrong. Webvan believed that people wanted groceries delivered so badly they’d pay premium prices for it. They never truly tested whether the market would bear the cost of their model. They built. They scaled. They spent. And then they hit reality.
Kodak is another one. The company literally invented the digital camera. But Kodak’s core assumption—that film photography would remain the money-maker—was never challenged by leadership. That unchallenged assumption cost the company its entire relevance. By the time they woke up, it was too late.
Here’s what’s even more revealing: confirmation bias isn’t just a personal psychology problem. It’s an organizational one. Once a leader commits to a direction, the entire organization starts gathering evidence to support it. People don’t bring bad news. They polish the information that confirms the narrative. Dissent feels dangerous.
The deeper problem: we’re built to be overconfident. Research on pre-mortems—a technique where you imagine a project has already failed and work backward to find the reasons—shows that when teams use this approach, they reduce overconfidence significantly more than traditional risk-analysis methods do. In one study, the pre-mortem method improved ability to forecast risk by as much as 30%.
Think about that. A 30% improvement in foresight. Not from more data. Not from smarter people. From forcing yourself to think backwards from failure instead of forwards from hope.
The cost of an unchallenged assumption isn’t just money. It’s time you’ll never get back. It’s opportunity cost. It’s the team members you hired for a problem that didn’t exist. It’s the market share you gave to a competitor while you were building the wrong thing.
The Shift: From Getting Answers to Interrogating Your Thinking
So here’s what I want you to hear: the highest-leverage move with AI isn’t faster information retrieval. It’s using AI as a rigorous thinking partner who will challenge you before you commit.
I’m not talking about asking your AI tool to give you pros and cons on a decision. You can do that in your own head, usually while driving to work. I’m talking about something much more specific.
I’m talking about explicitly naming your assumptions to an AI tool and asking it to tell you what you might be missing.
Before any major decision—and I mean before I’ve decided—I open my AI tool and I do something very specific. I write out the decision I’m considering, the assumptions I’m making, and I ask the AI to challenge those assumptions like a skeptical partner who wants me to succeed.
Here’s the shape of it: I’m about to [make this decision]. Here’s what I believe to be true [name your assumptions]. What am I missing? What could be wrong with my thinking?
That’s it. That simple move changes everything.
Because now the AI isn’t reinforcing what you believe. It’s stress-testing it. It’s poking holes in your logic before you’ve committed capital, time, and credibility to a direction that might be wrong.
Here’s why this works: When you ask an AI tool to validate you, it will. That’s what it’s trained to do—to be helpful, to be supportive, to move the conversation forward. But when you explicitly ask it to challenge you, to find flaws, to play the skeptic—suddenly you’re not seeking confirmation. You’re seeking clarity.
This is the move that separates entrepreneurs who build empires from entrepreneurs who repeat the same mistakes at higher volumes.
The companies I know that are making the smartest decisions aren’t the ones with the most data. They’re the ones brave enough to interrogate their own thinking before they’ve committed to it. They use AI the way a fighter uses a sparring partner—not to validate technique, but to expose holes in it.
How to Make This a Habit: 7 Steps You Can Do This Week
Here’s the practical version. You can start this today.
1. Name your decision clearly.
Before you do anything, write this down: “I am about to [be specific]. The timeline is [when]. What I’m trying to solve is [the real problem].” This forces you to be concrete instead of vague. Vague decisions hide vague assumptions.
2. Write out your assumptions explicitly.
Don’t just think them. Write them. “I’m assuming that [X]. I believe [Y] is true. My reasoning is [Z].” Write at least 5-7 assumptions. Most people stop at 2-3. The ones you discover on number 6 and 7 are usually the dangerous ones.
3. Open your AI tool and paste your decision + assumptions.
Don’t edit it. Don’t clean it up. Paste the raw version. AI doesn’t care if your writing is perfect, and raw thinking is often more honest thinking.
4. Ask the hard questions.
Write this exact prompt: “I’ve written out a decision I’m about to make and the assumptions underneath it. I need you to think like a skeptic. What could be wrong with my thinking? What am I not seeing? What assumptions might be untested? Where might I be overconfident?” You’re explicitly asking the AI to challenge you.
5. Read the response carefully. Sit with discomfort.
This is the part most people skip. When the AI points out a flaw in your logic, your first instinct will be to dismiss it or explain it away. Don’t. Read it twice. Let it be uncomfortable. That discomfort is often the signal that you found something important.
6. Ask one follow-up question.
“If this assumption is wrong, what would be the impact? How would I test it? What would I need to see to change my mind?” This moves from abstract critique to actionable reality-testing.
7. Before you decide, take 24 hours.
Don’t commit immediately after this conversation. Sleep on it. The best insights from interrogating your thinking come the next morning, when your brain has had time to process. You’ll often find yourself testing the AI’s challenge or even designing an experiment to validate or invalidate your assumption.
The Real Payoff: Decisions That Actually Stick
I can’t promise that this habit will prevent you from ever making a bad decision. But I can tell you what I’ve seen in my own life and in the lives of the 500,000+ entrepreneurs I work with.
When you do this before you commit, three things happen:
First, you catch the truly dangerous assumptions early. The ones that will sink the whole thing. And you either validate them through real testing, or you pivot before you’ve spent $100,000 and six months on the wrong path.
Second, you move forward with more conviction. Sounds weird, right? But when you’ve actually interrogated your thinking and come out the other side, you’re not carrying doubt. You’ve tested your logic. You’ve imagined failure. You’re harder to shake.
Third, you build a culture where this is normal. When your team sees you interrogating your own assumptions before major decisions, they start doing the same. The culture stops rewarding confidence and starts rewarding clarity. And that changes everything about the quality of decisions you make as an organization.
Here’s what I know: the decision you make next week—the hire, the product pivot, the market shift—will be significantly better if you’ve spent one hour interrogating your own thinking with AI before you commit to it.
Not because AI has all the answers. Because you finally asked the right questions.
The smartest entrepreneurs aren’t the ones who use AI to move faster. They’re the ones who use AI to think deeper before they move at all.
About the Author
Jonathan Mast is the founder of White Beard Strategies, where he helps 500,000+ entrepreneurs use AI to build faster, smarter businesses. He’s the creator of the Perfect Prompt Framework and a sought-after AI speaker and coach. When he’s not helping entrepreneurs rethink how they use AI, he’s usually proving that you can build something meaningful without sacrificing what matters most.
