By Moyn Islam
I use AI tools constantly — for research, for stress-testing an idea before I commit resources to it, for drafting the first version of almost anything. I’ve also watched founders use the exact same tools to talk themselves into decisions the market would have talked them out of in five minutes. The difference isn’t the tool. It’s what people ask it to do.
Here’s how I think about where AI actually earns its place in a founder’s decision-making, and where it quietly starts doing damage.
Where AI Genuinely Helps
The strongest use of AI I’ve found isn’t generating answers — it’s generating better questions. Before I act on an idea, I’ll use it to argue the other side: what would make this fail, who’s already tried something similar, what assumption am I not stating out loud. That’s a genuinely different use than asking a model to validate a plan I’ve already decided I like.
This lines up with something I said in a recent interview: AI is useful when it helps a founder examine an idea from more than one angle. Not because the model has some special insight into your market — it doesn’t — but because it’s very good at surfacing angles a founder under deadline pressure won’t think to ask themselves.
We built this directly into BE’s platform rather than treating it as a side feature. The AI Co-Pilot exists specifically to widen the range of questions a learner is asking, not to hand them a finished answer and let them stop thinking. That distinction — tool for widening the aperture versus tool for shortcutting the thinking — is the one that matters.
Where It Quietly Starts Doing Damage
Here’s the part that gets skipped in most “AI for founders” advice: a well-formatted output feels like validated thinking, and it isn’t. I’ve watched founders spend an afternoon getting a model to produce a polished go-to-market plan, financial model, or competitive analysis — and treat the existence of that document as evidence the underlying idea is sound. It’s not. It’s evidence that a language model can produce fluent, structured text about your idea. Those are very different things.
The tell is usually the same: the plan looks more finished than the founder’s actual confidence in the market should allow. A model can help you stress-test an assumption. It can’t tell you whether a real customer will actually open their wallet — only the market can answer that, and only if someone goes and asks it directly instead of asking a model to simulate the answer.
I’d put it this way: treat AI the way you’d treat a sharp advisor who has read everything but has never run a business. Useful for widening what you’re considering. Dangerous the moment you let it replace the step where you go find out for real.
The Judgment Call AI Can’t Make for You
Every strategic decision eventually comes down to a judgment call that has to account for context no model has access to — your specific team’s capacity, the relationship history with a specific partner, what your gut is telling you about a hire that looks good on paper. AI can help you organize the inputs to that decision. It can’t sit in the room and make the call, because the call requires weighing things that were never written down anywhere a model could learn them.
This matters more as these tools get better, not less. The more fluent and confident an AI’s output sounds, the easier it becomes to mistake fluency for correctness. The founders I’ve seen navigate this well are the ones who use AI to expand the range of options they’re considering, then deliberately step away from the tool to make the actual decision — treating the output as one input among several, not the verdict.
A Simple Way to Use It Better
If there’s one habit I’d recommend adopting: before you act on anything AI helps you produce, ask what real-world evidence would prove it wrong — and then go get that evidence from an actual person, not another prompt. If the honest answer is “I haven’t checked,” that’s the signal to slow down, not the moment to trust the output more.
AI can accelerate your journey. It can help someone act boldly and make meaningful strides from day one, which is a real and valuable thing. What it can’t do is replace the specific, unglamorous work of testing an idea against reality — and the founders who forget that are the ones who end up with a beautifully formatted plan for a business nobody wanted.
For more on how I think about building and scaling with AI, visit https://moynislambiography.com/
Frequently Asked Questions
Should founders use AI to write their business plans? AI can help structure and stress-test a plan, but the underlying assumptions still need to be validated against real customer behavior — a polished AI-generated plan isn’t evidence that the idea itself is sound.
What’s the biggest risk of relying on AI for strategic decisions? Mistaking a fluent, well-formatted AI output for validated thinking. A model can organize information persuasively without having tested any of it against reality.
How should founders think about AI as a tool? As something that widens the range of questions and angles being considered — not something that replaces the judgment call a founder still has to make with full context of their specific situation.
Can AI predict whether customers will actually pay for a product? No — that requires going to the market directly. AI can help you frame the question, but only real customer behavior can answer it.