Stop Hiring AI Celebrity Advisors — Here's What Bosses Should Actually Be Using AI For

Celebrity brain trusts are, frankly, playacting

A whole category has popped up in the market lately, aimed squarely at bosses, called AI celebrity brain trusts, sometimes celebrity private board meetings. The pitch always sounds the same. Spend a little money and you get your own dream team, Buffett, Jobs, Musk, ready to advise you at any moment.

Sounds impressive, doesn’t it. I tried them all. Let me be straight with you — this thing is cyber playacting, and what you’re buying is emotional comfort. If you want to make a real business decision with it, it’s nowhere close.

It’s not completely useless. It has been fed at least some of what these figures have said publicly, and there’s a top-tier model behind it helping you think, so chatting with it to find some inspiration is mildly interesting. But put a genuinely complex business problem in front of it and ask it to call the shot for you, and it folds immediately. Because it’s missing the single most important thing: a framework and tools for analyzing a problem. What it learned is one person’s cadence and turns of phrase — not the real craft of taking a business apart.

Real capability comes from tools: the Lean Canvas

So what does doing business, doing analysis, actually rely on? Frameworks. Tools. The most classic one is the Lean Canvas.

It grew out of the Business Model Canvas, created by a Swiss academic named Osterwalder — originally his 2004 doctoral thesis at the University of Lausanne, later famous worldwide and used by the likes of Coca-Cola and LEGO. But that canvas was built to “describe” a company that already exists. Later Ash Maurya pointed out that the biggest hurdle in a startup isn’t description at all — it’s that you have no idea whether anyone actually wants what you’re making. So he modified the canvas, and that became the Lean Canvas, better suited to founders.

The canvas is just nine boxes, which are really nine questions you have to answer. Who exactly is your customer, what unsolved pain do they have, what makes you different from everyone else, what’s your solution, through what channels do you find them, where does the money come from, where do the costs go, which single key metric are you watching, and what edge do you have that nobody can take from you.

But filling in the nine boxes is not the point at all. The real soul of the Lean Canvas is that it forces you to hunt down the one killer “key assumption” — which box is the shakiest and simultaneously the most life-or-death, the one that if it’s wrong, the whole business goes under. Once you’ve found it, don’t bet everything you own on it. Verify that one assumption at the lowest possible cost. Find the key assumption, then verify it cheaply. That’s where the real value of this canvas sits.

Xiaomi: build a system first, see if anyone will flash it

Here’s an example you all know, Xiaomi. In 2010 Lei Jun wanted to make phones — supply chain, factories, hundreds of millions of yuan sunk in, and if nobody bought them it was all gone. The box he was least sure about: did a crowd of enthusiasts actually exist who would go all in on a phone they could tinker with down to the guts? Get that box wrong and everything downstream is wasted.

How did he verify it? He didn’t build a factory and make phones — that’s the most expensive way to bet. He took the cheapest route available and built a free phone OS, MIUI, dropped it into the forums, and asked who wanted to flash it. The first version had exactly 100 people. But from those 100, with zero yuan spent on advertising, it doubled to 200 the second week, 400 the third, 800 the fourth, and hit over 300,000 within a year. The key assumption was verified, and only then did he dare put real money into phones.

The most expensive verification is building out the whole factory and supply chain and then waiting for the market to pronounce life or death. The cheapest verification is shipping an OS first and seeing whether anyone will brick their phone for you. The smaller the company, the more this applies — a one-person company has the fewest bullets of all, so it can least afford to spray them in the wrong direction. Yet far too many companies start out immersed in a grand strategy they invented in their own heads, with dazzling slide decks and a vision announced at full volume, and they forget to stop and ask the one question that matters. That killer assumption in my business: has it been verified or not.

Not a one-time exercise: turn the canvas into an agent you can re-run

The canvas used to be a one-shot deal. You fill one in carefully when you start the company, then lock it in a drawer and never look at it again. But business changes daily. A new competitor shows up, something you bet on doesn’t pan out, a different kind of customer arrives, and your canvas has long since expired.

This is where AI is fundamentally different. It takes that canvas out of the drawer and turns it from a dead sheet of paper into a living agent that will re-run it with you as many times as you want. Hiring a consultant to analyze the business used to be expensive and slow, once a year at most. Now you can drop the latest situation in every month, even every week, have it re-run, and tell you which box has become the thing you most need to verify right now.

So I used Claude Code to build myself a little assistant, a skill. It doesn’t dump a blank canvas on you to fill in. It asks you one question at a time like a real coach, and by the time it’s walked you around the loop, a perfectly respectable canvas has grown itself. At the end it also forces you to mark which box is your key assumption. Come back next month, describe what’s new, and it runs the whole thing with you again from the top.

The Lean Canvas is really just an example. Porter’s Five Forces, SWOT — all these battle-tested classic frameworks can become agents that re-run with you like this. Stop wasting money having AI roleplay some mogul. What’s actually valuable is having the very best model sit with you inside a solid classic framework and think your own business all the way through, together. That is the real way to use AI in business.

This is the full voiceover transcript of this digital-avatar episode. Timecodes are derived from the audio ground truth.

[00:00] There’s something blowing up lately, aimed squarely at business owners — the AI celebrity brain trust, sometimes called a celebrity board of advisors. The pitch is that for a little money you get your own dream team on call: Buffett, Jobs, Musk, ready to give you advice any time. Sounds impressive, right? I tried every one of them. And I’ll be straight with you: this stuff is basically cyber make-believe. What you’re buying is emotional value. If you actually want to make business decisions with it, it’s nowhere close.

[00:19] So how should a boss actually use AI? I’ll tell you: the thing that really works is something else entirely — take the business analysis tools that are genuinely solid, and use AI to turn them into an agent you can run again and again. What does that mean? Hold onto it, I’ll break it down for you in a minute.

[00:31] You might say, so the celebrity brain trust is completely useless? Not quite. It has been fed some amount of what these figures have said publicly, and there’s a top-tier model behind it doing the thinking, so chatting with it for fun, fishing for inspiration — that’s actually kind of interesting. But put a genuinely complicated business problem in front of it and ask it to make the call for you, and it folds immediately.

[00:46] Why? Because it’s missing the one thing that matters most: a framework and a toolkit for analyzing the problem. What it learned is how a person talks, the mouth — not the real skill that takes your business apart inside and out.

[00:55] So what does running a business and doing analysis actually rely on? Frameworks. Tools. Let me give you the most classic one: the Lean Canvas. Its ancestor is the Business Model Canvas, invented by a Swiss guy named Osterwalder — it started as his 2004 doctoral thesis at the University of Lausanne, and then it went global, with big companies like Coca-Cola and Lego using it.

[01:12] But his canvas is for describing a company that already exists. Then along came a guy named Ash Maurya, who said the biggest hurdle in a startup isn’t description at all — it’s that you have no idea whether anyone actually wants the thing you’re building. So he reworked the canvas, and that became the Lean Canvas we have today, which fits founders much better.

[01:26] What does this canvas look like? Nine boxes. Which really means nine questions you have to answer. Who exactly is your customer. What unsolved pain do they have. What makes you different from everyone else. What’s your solution. What channels do you use to reach them. Where does the money come from. Where do the costs go. Which key metric are you watching like a hawk. And the last one: what do you have that nobody can take away from you.

[01:46] But note this: filling in the nine boxes is not the point at all. The real soul of the Lean Canvas is that it forces you to pull the deadliest key assumption out of those nine boxes — which box is the shakiest and at the same time the most life-or-death, the one that, if it’s wrong, sinks the whole business. Once you’ve pulled it out, you do the most important step: don’t bet everything you own on it. Verify that one assumption first, at the lowest possible cost. Find the key assumption, then verify it at minimum cost. That’s where the real value of this canvas is.

[02:08] Take an example you all know: Xiaomi. Drop Xiaomi into those nine boxes. Who’s the customer? Android power users. What’s the pain? Stock ROMs are hard to use and get updated slowly. What makes it different? A system that updates every week and actually listens to you. The solution: MIUI. The channel: word of mouth on forums. Where the money comes from: selling phones later on, with cost being one small software team. Key metric: whether users double week over week. The thing nobody can take: a crowd of die-hard Mi fans.

[02:35] Lay out those nine boxes and the key assumption surfaces immediately. What Lei Jun was least sure about wasn’t whether a phone could be built. It was that nobody had any idea whether a crowd of power users like that even existed — people who’d genuinely go all in on a phone like this. Get that box wrong and hundreds of millions go straight down the drain.

[02:45] So how did he verify it? He didn’t build a factory and make phones — that’s the most expensive way to bet. He went for minimum cost: he built a free ROM, MIUI, dropped it on the forums, and asked who was willing to flash it. The first version got 100 people willing to risk bricking their phones to try it. And from just those 100, without spending a cent on ads, week two doubled to 200, week three 400, week four 800, and in under a year it was past 300,000. Key assumption: verified.

[03:00] Only then did he dare put real money into a phone. When the Xiaomi phone launched in 2011 at 1,999 yuan a unit, 320,000 of them sold out in 34 hours. See the difference? The most expensive way to verify is to build out the whole factory and supply chain and then wait for the market to decide whether you live or die. The cheapest way is to ship a ROM first and see whether anyone will brick their phone for you.

[03:20] And don’t think this is something only big companies can afford to play with. It’s the opposite — the smaller the company, the more you need it. Even if you’re a one-person company, you’ve got very little ammo to begin with, so you really can’t afford to blow it firing in the wrong direction. But in reality, way too many companies get lost right away in some grand strategy they’ve made up in their heads. Gorgeous slides, a vision shouted from the rooftops, and the one thing they forget is to stop and ask themselves: the biggest risk in this business, the deadliest assumption — have I actually verified it or not? It’s usually that one step that gets skipped, and by the time the money’s burned they find out that the thing they should have spent a little money verifying first doesn’t hold up at all.

[03:46] But note, there’s a big trap in here too. In the past this canvas was a one-shot deal. You’d fill one out carefully when you started up, then lock it in a drawer and never look at it again. But the business changes every day. A new competitor shows up, something you bet on doesn’t pan out, a different crowd of customers turns up — and your canvas expired a long time ago.

[04:07] And this is where AI changes everything. It can turn that canvas from a dead sheet of paper in a drawer into a live agent that reruns with you again and again. Hiring a consultant to do one business analysis used to be expensive and slow — once a year, tops. Now you can drop the latest situation in every month, even every week, have it rerun the whole thing, and tell you which thing has become the one you most need to verify right now. Business analysis goes from a one-shot deal to something you keep doing and keep refreshing.

[04:24] So how do you make this canvas actually come alive and rerun? I put together a little helper for myself with Claude Code — a skill. It doesn’t hand you a blank canvas and tell you to fill it in; that’s useless, nobody can fill one in properly. It works like a real coach, asking you one question after another. Who exactly were your earliest customers. How are they making do right now without you. And what makes you think your solution will work. After it’s walked you around the loop like that, a pretty decent canvas grows on its own. And at the end it forces you to mark one thing: out of these nine boxes, which one is your key assumption — the one that, if it collapses, your company is done immediately. Then next month you come back, tell it what’s new, and it walks you through the whole run again.

[04:59] The Lean Canvas is really just one example. Want competitive market analysis? Porter’s Five Forces. Want a strategy review? SWOT. All of these battle-tested classic frameworks can be turned into an agent that reruns with you like this. So the one thing I actually want to say today is: stop wasting money getting AI to play some famous figure. What’s genuinely valuable is having the most capable model in the world sit with you inside a solid classic framework and think your own situation all the way through, over and over, until it’s clear. That’s the real way to use AI in business. I built this Lean Canvas coach as a skill and open-sourced it — you can grab it from the comments or from the playbook on my WeChat account. Which box in your own business is the one that would least survive a test? Once you’ve figured that out, let’s talk in the comments.