AI Can Diagnose but It Can't Testify

Greg Crabtree has looked at more small business P&Ls than almost anyone alive. After decades of it, he noticed something nobody else had written down: companies stall at predictable revenue thresholds. He calls them black holes. Hit one, and you can orbit it for years without knowing why.

Thousands of CPAs had the same data Greg had. Same stalling clients, same plateaus, same financials crossing their desks quarter after quarter. None of them saw the pattern or at least none of them told anyone about it. Not because it was hidden, but because they were looking at the numbers to check if they were right, not to ask what they meant.

Greg was asking a different question. That question, applied across hundreds of private engagements over decades, produced an insight you couldn't find in any textbook. It came from a vantage point that required years of sustained access to the data, the mindset to do something different with it, and a deep knowledge of what drives business owners.

Greg wrote a book about the black holes. It's out there. You can ask Claude about it right now and get a perfectly competent explanation. The insight is no longer private.

And almost nobody acts on it without Greg (or someone he trained) in the room.

I have my own version of this. When business owners tell me they want to double or triple their companies, I tell them they'll probably have to replace 50 to 75 percent of their leadership team. The people who got you here can't get you there. It's not a judgment on them. It's a structural reality I've watched play out over and over.

Nobody wants to hear it. It doesn't show up in blog posts because it doesn't get clicks. It doesn't trend on LinkedIn because it makes people uncomfortable. Conference speakers don't lead with "you need to fire the people you built this with." AIs have a selection bias toward what's pleasant, popular, and encouraging. The hardest truths about running a business are none of those.

But let's say AI surfaces it anyway. Let's say you're using some dashboard and it flags that your COO is a growth bottleneck based on decision velocity, team turnover in her departments, and stalling revenue. The diagnosis is accurate. The data is clean. The recommendation is right there on screen.

Do you act on it?

You don't. Because the question was never "is this true?" You already knew. The question is "can I survive doing this?" And a dashboard can't answer that.

There's a difference between a system telling you your COO is blocking growth and a person sitting across from you who says: "I know this is real because I did it. I fired people I loved. I went to their weddings. And it was the right call."

That's not advice. It's testimony. It lands differently because the business owner's real problem isn't informational. It's emotional. They need proof of survivability from someone who bears the scars.

AI has no scars.

This is the part that gets lost in the discourse about what AI will replace. Everyone's focused on capability. Can it write the brief? Can it analyze the data? Can it spot the pattern? Yes. Increasingly, to all of it.

Knowledge was the bottleneck when knowledge was scarce. Pre-internet, you paid your accountant because she knew things you couldn't easily look up. The internet made knowledge accessible but AI makes it free, structured, and instantly applicable. You can get a competent answer to nearly any business question in thirty seconds.

So now the bottleneck has shifted. It isn’t knowledge, it’s the gap between knowing and conviction. It was always there, hiding behind the knowledge problem. Now that knowledge is solved, we can see it clearly. And it's enormous.

Greg Crabtree doesn't sell knowledge about black holes. He sells the ability to walk into your specific situation, recognize which stall pattern you're in, and push you to make the hard decision that’s as plain as day.

I don't sell the insight that your leadership team needs to turn over. I sell the credibility of having done it myself, the judgment to know when it applies to your situation, and the willingness to say it to your face when everyone else is being polite.

These aren't things AI threatens. They're things AI makes more valuable.

Rather than replace the person who gets you to act, AI makes that person dangerous.

Before AI, Greg could work with maybe fifty clients at a time. He had to manually review financials, spot patterns, prepare for conversations. Now he can have AI scanning every client's numbers continuously, flagging who's approaching a black hole, building the analysis before he walks in. People like Greg will spend almost none of their time on diagnosis and all their time on the intervention. The hard conversation. The part only they can do.

Same principle applies to anyone in a similar position. AI strips away everything except your highest-value activity and lets you do that activity far more often with far more people.

One person with hard-won insight plus AI does what used to take a firm.

Stop worrying about whether AI knows what you know. It does, or it will. That doesn't matter.

Ask instead: what do you know that's true, not because it's secret, but because it's unpopular, or because it requires a human to make you take it seriously.

What's the thing you tell clients that they resist hearing? What's the pattern you've seen so many times you stopped thinking of it as insight?

That's your edge. Not the knowledge itself but the fact that you can make someone act on it. Because you've done the hard thing. Because you have standing. Because the client's real question isn't "is this true?" but "can I survive it?" and you're the only credible answer in the room.

AI just made everything around that moment cheaper. Which means the moment itself — the one where a specific human gets another specific human to do the hard thing — just became the most valuable thing in business.

Figure out what yours is. Then use AI to do everything else.