In a room of 2,000 people, one is building with AI

Hi there,

You're one of about 50 people getting this. I built the list by hand, which is fitting, because this issue is about how small the room actually is.

If it landed in your inbox we've met, worked together, studied together, or you asked me for it. Between you there are founders, executives, and people running things in mining, government, health, property, banking, sport and tech. A few of you are already in the middle of a build with me.

I left Google last month and have been building AI systems for businesses since, mostly around growth and how the work actually gets done.

A few times a month: what I built, how it works, what broke. No listicles, no selling.

Now the number.

Put 100 people in a room

Eighteen. That surprises most people I say it to, because if you work in tech it feels like everyone is using this stuff.

Now take just those 18 and put a hundred of them in a room.

Three. Ninety-seven are on the free tier and stopping there.

Now take those three, and put a hundred of them in a room.

Eight. That last group is the frontier: agents, APIs, systems wired into how a business runs, rather than typing into a chat box.

Multiply the three rooms together and you get one number worth sitting with. In a room of 2,000 people, one is building with AI.

That is not a prediction about where this ends up. Nobody credible knows. It is where the room is standing this month, and the ground where the leverage sits is almost entirely empty.

Sources: UN World Population Prospects 2025, DataReportal Digital 2026, ITU Facts and Figures 2025, and reported user numbers from OpenAI and Microsoft. The paid and builder counts are deliberately conservative.

Most of what you're reading doesn't matter

Claude versus Gemini, open weight versus closed, whoever topped a benchmark this week. Interesting, fast moving, and almost none of it changes what you should do.

Which model you pick is close to the least important decision here. What decides it is whether the work around it gets rebuilt. Models are the cheap part now. The rebuild is the expensive part, and nobody writes about that, so that's what I write about.

Now find your business on it

Go back through those three rooms and look for yourself. Nearly every business I talk to is in one of three places.

If you're inside a large organisation rather than running your own, read "your business" as the function you own. The three rungs are identical, only the budget conversation changes.

The jump from the second to the third is the one that matters, and almost nobody has made it. It isn't a productivity tweak. It changes what you can take on at all.

Here's what that looks like, from builds I've run this year.

Different industries, same underlying shape.

1. What all three have in common
2. Inputa recording, a job spec, an enquiry
3. The rules run firstthe standard applied before the work, not checked after
4. The system does the workdrafts, scores, files, routes
5. A person signs offevery time. The part people skip, and the part that matters
6. Outputpublished, priced, delivered

None of that is exotic. It's ordinary business process, rebuilt so the system does the work and a person signs it off. The reason to start now isn't that AI is impressive. It's that the businesses one rung up are quietly changing what they can deliver, and the gap compounds.

Building got cheap

A platform that needed a product manager, a designer and two engineers for six months last year takes one person about a month now.

Roughly twenty to one. The hard part moved though: planning at the front, sign-off at the back, and a short stretch of building in between.

You don't need to be code native to do this. You need the plan.

The part that surprised me most

Rebuild a workflow properly and you haven't only solved your own problem. You've built something other businesses with the same problem would pay for. The engine I built for one business isn't specific to that business. It's specific to that problem.

Inside a large organisation the same thing holds, except the other businesses are other departments, and the currency is headcount you don't have to ask for.

That's productisation, and until recently it needed a dev team and a year of runway. Now it's a version of the thing you already run.

Then the constraint moves again. If building is cheap and everyone can build, making the thing stops being the hard part. Getting anyone to know it exists becomes the hard part. Distribution is the new bottleneck, and it's the one AI helps with least.

Pick what comes next

And if you have questions or ideas about AI, I'd be glad to share what I've built.

Talk soon,
Jason