Most of the AI debate splits into two camps. One says it will replace half the workforce. The other says it’s overhyped and will settle into helping with administrative tasks.. Josh Cote at EAW Consulting thinks both camps are asking the wrong question, because neither one is about how the tool actually gets used day-to-day, inside an actual business, by an actual founder trying to make a decision.
The useful question isn’t what AI can do. It’s what you’re asking it to do for you. That distinction has turned out to matter more than almost anything else founders get right or wrong about the technology.
The Wrong Question Most People Ask AI
Watch how most people actually use these tools and a pattern shows up fast. They ask AI to write the email, summarise the report, or produce the answer. Used this way, AI functions as a shortcut, a way to skip the part of the work that used to require thinking.
That’s useful. It saves time. But in Josh’s experience as a consultant, it doesn’t always do much for the person using it beyond that. If AI is doing the thinking every time, there’s less reason for you to work through the problem yourself.
The result might be better, but that doesn’t necessarily mean the person producing it is getting better too. Harvard Business Review’s research on AI and reasoning makes a similar point. When people accept AI-generated answers without questioning them, they can start relying on the technology to do more of the mental work. One of the simplest ways around that is to use AI to challenge your thinking and ask better questions, rather than just hand over the answer.
The founders getting genuinely sharper because of AI are doing something different. They’re not asking it to produce the answer. They’re asking it to test the answer they already have.
Argue With Yourself, Not Just Answer Yourself
One of the best things AI can do isn’t replace expertise. It can help people learn faster, especially if they’re already curious and willing to question what they know.
That can be as simple as asking it to explain something you don’t quite understand, challenge an assumption you’ve been making for years, or make the strongest possible case against your own idea. Used like that, AI becomes less of a tool for giving you answers and more of something you can use to test your thinking.
You can often tell a lot about a founder by asking how they use AI in their business. Someone who says, “It drafts my emails,” is mainly using it to save time. Someone who says, “I use it to test my pricing model before I take it to the team,” is using it very differently. They’re using it to question their own thinking before making a decision, and that usually says something about how they approach the rest of the business too.
Why This Compounds Faster Than Almost Any Other Tool
A good mentor, a sharp co-founder, a demanding board member: these have always been the fastest way to get better, because they push back. The problem is they’re not always available, and even the best of them get tired of hearing the same question twice.
AI doesn’t get tired. It doesn’t soften the pushback to protect the relationship. It doesn’t need to be scheduled two weeks out. Ask it to poke holes in a plan at eleven at night and it will, without the social cost of asking a real person to do the same thing at that hour.
That availability is what makes the compounding effect real. A founder who runs one idea past a sharp critic once a quarter gets four rounds of sharpening a year. A founder who does something similar with AI every time a real decision comes up gets that same sharpening dozens of times a month. Same mechanism, wildly different rate of improvement.
Where This Shows Up at EAW Consulting
Josh Cote has started building this directly into how EAW Consulting works with clients. Before a strategy session, the team increasingly asks founders to bring the version of their plan that’s already been stress-tested against a set of hard questions, not just the polished version they’d present to a room.
It changes the sessions completely. Instead of spending the first hour finding the obvious holes, the conversation starts somewhere further along, because the founder has already had the easy objections raised and answered before anyone in the room said a word. As a consultant, that’s the difference between a session that surfaces new thinking and one that just repeats the thinking everyone already had.
It’s also changed how EAW Consulting evaluates opportunities as an investor. A business plan that’s clearly never been challenged, not even by its own founder using the tools sitting right in front of them, reads differently than one that has. It’s not about whether the plan is perfect. It’s about whether the person building it has a habit of testing their own thinking before someone else has to.
The Founders Who Get This Wrong
The failure mode isn’t ignoring AI. Almost nobody does that anymore. The failure mode is treating it as a faster way to be told you’re right.
It’s easy to ask a tool to help you write a pitch and hard to ask it to argue against the pitch you already believe in. The first request feels productive. The second one feels uncomfortable, because it’s designed to. That discomfort is the entire value. Founders who only ever use AI to confirm their existing thinking get the confidence of a second opinion without any of the benefit of an actual second opinion.
The people getting the most out of AI, as Josh puts it, aren’t the ones asking it to do their thinking. They’re the ones asking it to sharpen their own.
Three Prompts Worth Stealing
None of this requires anything technical. It just requires asking a different kind of question. Before the next big decision ask this:
“What’s the strongest argument against this plan, from someone who’s seen it fail before?” Not the obvious risks. The ones a genuinely sceptical outsider would raise first.
“What am I assuming here that I’ve never actually tested?” This one tends to surface the quiet assumptions that have been sitting unchallenged for years, the kind that only get noticed once someone, or something, finally asks.
“If I’m wrong about this, what’s the first sign I’d see, and would I actually notice it?” This forces a founder to define what failure would look like in advance, rather than recognising it only in hindsight.
None of these questions are complicated. What’s different is that most founders never ask them of anyone, because asking a real person to argue against your own plan carries a social cost. Asking a tool costs nothing but the willingness to sit with an uncomfortable answer.
The Real Edge Isn’t the Tool
Every founder now has access to roughly the same AI tools. That access stopped being the advantage a while ago. What still separates one entrepreneur from another is what they’re willing to ask the tool to do to their own ideas.
There’s no doubt: businesses that use AI to get comfortable will get comfortable faster than they used to, and businesses that use it to get challenged will get sharper faster than they used to. Same technology. Opposite outcomes. The tool was never going to make that decision for anyone.
FAQs
What’s the difference between using AI to answer a question and using it to challenge your thinking?
If you ask AI for an answer, you’re mainly using it to get somewhere faster. If you ask it to pick apart a plan, question an assumption or point out what you might have missed, you’re using it very differently.
Josh’s point is that the second approach can actually improve the way you think over time. You’re still making the decisions, but you’re using AI to test them first rather than simply asking it what to do.
Why does Josh Cote say most founders are using AI wrong?
Not because they’re not using it, but because they’re mostly using it to confirm or produce, rather than to question. The tools reward the harder habit of asking AI to argue against your own idea, not just answer it.
What questions should you ask AI to get more value from it?
Josh suggests starting with the strongest argument against your plan, the assumptions you’ve never actually tested, and the first sign you’d see if you were wrong. None require anything technical, just a willingness to sit with an uncomfortable answer.
Does relying on AI make people worse at thinking for themselves?
It can, if it’s used to skip the thinking rather than sharpen it. Research from Harvard Business Review points to measurable cognitive offloading when people accept AI answers uncritically, which is exactly why Josh frames the useful habit as asking AI to question you, not just answer you.
How does EAW Consulting use AI when working with clients?
The team now asks founders to bring a version of their plan that’s already been stress-tested against hard questions before a strategy session, rather than the polished version they’d present to a room. It changes what the session can actually achieve.
