Most companies are trying to solve the AI transformation puzzle by buying something that can’t be bought.

They treat AI like any other kind of software. You pick a tool, roll it out, and wait for the work to get better. Usually it doesn’t, and the usual explanations don’t really explain it. It isn’t that they bought the wrong tool, or didn’t train people enough. I think they’ve misread where the value is. The AI is a tool, and by now everyone has roughly the same ones. What decides the result is the capability to use them well, and that capability is the thing you can’t buy or spend your token budgets on.

That difference does more work than it looks like. A tool sits outside you. You adopt it, swap it, drop it, and you’re the same person afterward. A capability is yours: leadership, or judgment about money, or knowing how to read a room. Nobody installs those in you. You get them by doing the work, and you’d never hand one to a vendor and still call yourself good at your job. AI itself belongs with the tools. But it’s a strange tool, because what it gives back depends almost entirely on the person aiming it. So the tool spread to everyone, and the capability to use it well became the thing that separates people. Most are still paying for the tool and skipping the capability, which is why they keep buying and waiting, and why the waiting doesn’t end.

How I came to this

I didn’t just wake up one day with this thinking or reason my way to this. I watched it happen firsthand. A few years ago, I was a Chief of Staff at a rapidly growing B2B SaaS company. And then one day, right around the time ChatGPT first came out, my CEO handed me a different job: figure out how the entire company itself could work differently with AI. He didn’t mean putting AI in our product. He meant changing how we actually worked. I spent the next couple of years on that, from the inside, first directing a team of engineers and then, once it got easier to build than to explain, building it myself.

What surprised me was what didn’t work. Every time we built something for a team and handed it over, almost nothing stuck. They used it at arm’s length, it slowly went out of date, and the people were no more capable than before. Whatever capability showed up, showed up in the person who did the building, never in the thing they built. My team and I stuck out like a sore thumb in what we were doing, how we solved problems, and the impact we were making. What changed us was doing the work ourselves, clumsily at first, until it was theirs.

More about Harry+

Before OneTwo Growth, I spent five years as Chief of Staff at a B2B SaaS company, from seed funding through Series C. The work was whatever fell between the org chart: rebuilding the financial model before a board meeting, writing a product spec, sitting in on sales calls to understand why deals stalled, fixing the handoffs where leads went missing. Most of it came down to building systems that didn’t exist yet — compensation frameworks, planning processes, reporting infrastructure — and connecting strategy to execution across product, GTM, finance, and ops.

When ChatGPT arrived, that turned into a single mandate from the CEO: figure out how the whole company could work differently with AI. Two years of doing it — first directing engineers, then building myself — is where the conviction behind this work came from. I’m based in Seattle.

What actually works

Once you’ve seen that, a lot follows. If building it for people doesn’t transfer the capability, then most of what the market is selling can’t really work. Not the tool you roll out, not the automation a consultant hands you, not the course you don’t finish. They leave you about where you started, just poorer and a little more anxious. The only thing I’ve ever seen actually work is developing the capability in the person who needs it. So that’s the one thing I do, and it’s why I probably won’t build your AI solution for you, even when you ask me to.

What the capability is

It’s fair to ask what the capability even is, if it isn’t knowing the tools. As far as I can tell it’s three things, and they hold together.

1. 01

   Thinking with AI. Framing a problem well, and knowing a good answer when you see one.

2. 02

   Handing off and directing. Turning your real work into something you can hand off and direct — closer to managing than to using software.

3. 03

   Knowing when to trust it. Understanding enough of how these systems work to know when to trust them and when they’re confidently wrong.

None of those is a feature anyone can show you. They’re judgment, and judgment only comes from doing the work.

You can tell who has it. They’ve stopped chasing tools and know that building a solution is more than vibe coding yet another web app. When a new model comes out, they’re curious rather than worried, because what they have is judgment, and judgment carries over to whatever ships next. They spend less time afraid of falling behind and more time deciding where to point the thing. The calm is real. It comes from having taste, judgment, and a technical understanding of the technology — combined with their own expertise and IP — that holds while the tools underneath it keep changing.

Where I’ve seen it work

I’ve watched this happen in companies that look nothing alike.

A COO at a legaltech wanted her operations team to work differently, not to be handed one more system to babysit. We did a couple of workshops and some coaching, built a few things alongside the team, and within about three weeks they were working in a genuinely new way. She put it better than I could:

The one I think about most is more subtle. A forty-year-old industrial company started a project the normal way, by outsourcing a build. Halfway through we stopped and decided the better, more durable investment would be to help them learn how to do it themselves. The rules of their business now live in a file their own team maintains, without me. They own it because they built it, and it kept working after I left. The work changed, and it stayed changed, because the capability ended up in the people.