Your team doesn't need to be good at AI. It needs to be top 0.1%

I told my team they need to be in the top 0.1% of AI power users. Not top 10%. Not top 1%. Top 0.1%. In practice that means two things: cutting-edge technique rather than better prompting, and automating ruthlessly. The bar we're building toward is 5x the quality of output in a fifth of the time, which is a target we set ourselves, not a measured result.

I told my team this week: you need to be in the top 0.1% of AI power users.

Not top 10%. Not top 1%. Top 0.1%. Full stop.

Here's the uncomfortable part for anyone whose job is to be smart for a living. AI is already more capable than most experts in most domains. As a growth adviser I feel that pressure acutely, and it only gets more intense from here.

The people who thrive aren't going to be the most naturally intelligent. They're going to be the best operators of AI. Big winners and losers are being made right now, and I fully intend to be on the right side of that. 🔥

"We use AI" stopped meaning anything

Everyone says use AI. Fine. But "our team uses AI" now tells you about as much as "our team uses the internet." It's table stakes. It's not a strategy and it isn't a differentiator.

Most people use AI like a faster search box. They ask it a question, paste the answer, move on. The top 0.1% are doing something different in kind, not in degree. They rebuild the workflow around it: chaining steps, running agents, building their own small tools, putting verification loops around the output so they can actually trust it.

One group saves a few minutes. The other group changes what a single person can ship.

So what does "top 0.1%" actually mean in practice

Right now, two things.

1. Cutting-edge technique, not just prompting. Living in tools like Claude Code and Cowork. Running agentic workflows rather than one-shot chats. Using leading prompting practices, and tracking every major model release and capability jump like it's your job. Because it is. The half-life on this stuff is short enough that a technique you learned six months ago is often no longer the good way to do it.

2. Automating ruthlessly. Your own work. Your team's work. Your clients' work. If you're not building automation into everything you touch, you're building on a foundation that's already cracking.

The second one is where most teams stall, and it isn't a skill problem. It's that automating your own work feels like arguing yourself out of a job, so people quietly don't. That instinct is the thing to manage directly, which I'll come back to.

Why 0.1% and not 1%

Because the returns are nonlinear at the tail.

The distance between not using AI and using it competently is real, but it's small. A modest speedup. The distance between competent and elite is enormous, and it's enormous for a specific reason: the elite user isn't doing the same task faster. They're doing a bigger task that wasn't possible before.

For my own consulting business the goal I've set is aggressive on purpose: 5x the quality of output in a fifth of the time of what someone gets doing it themselves with AI. To be clear about what that number is: it's a bar we set for ourselves and are building toward, not a result we have measured. You don't get near it by being pretty good. You get there by being at the edge, and then by spending the time you save on iteration, specificity and execution confidence instead of coasting on the speedup.

That last clause is the part that gets skipped. A team that gets 3x faster and pockets the 3x has bought itself a slightly earlier finish. A team that gets 3x faster and puts it back into more iterations has bought itself better work.

The reason "get good at AI" doesn't work as an instruction

Telling a team to be top 0.1% and leaving it there is a wish, not a plan. I've watched that fail: everyone agrees, nothing changes, because the actual blocker is that nobody has time to figure out which of forty techniques matter this month.

So I made the material rather than the exhortation. Everything I've made mandatory for my team, the advanced techniques, the agentic workflows, the whole thing, exists as a package they can work through: a podcast version, a video explainer, slides, and a reference doc. Built with NotebookLM in an afternoon, which is itself the point. There's really no excuse not to be an expert when the curriculum takes less time to produce than one person spends flailing at it alone.

That's the honest cost of this standard, by the way. If you set the bar at top 0.1% you owe your team the path to it, and if you don't build the path you're just applying pressure.

What it looks like on a team

Reps, mostly. People building real workflows, sharing what actually worked, and AI becoming the default way work gets done rather than a thing you reach for occasionally.

Two things I'd do deliberately. Hire and train for it, so it's a stated expectation and not a nice-to-have someone discovers in month four. And reward the person who automated their own job instead of letting them worry about it, because the alternative is a team that hides its best leverage from you.

It's a practice, not a purchase. Buying the tools is the easy part and it's where a lot of teams stop.

The takeaway

"We use AI" is a baseline, not a strategy. The advantage lives out in the tail, where a few people have rebuilt how they work and are shipping at a level the rest can't match.

Aim your team there, give it the reps a real practice takes, and hand them the path rather than the pressure.

Common questions

What does a top 0.1% AI user actually do differently?

They rebuild the workflow around AI instead of using it as a faster search box: chaining steps, running agents, building small tools, and adding verification loops so the output can be trusted. It's a difference in kind, not degree.

Why aim for 0.1% instead of just being competent with AI?

Because the returns are nonlinear. Competent-to-elite is a far bigger jump than not-using-to-competent, since the elite user is doing a bigger task that wasn't possible before, not the same task slightly faster.

How does a team actually get there?

Build the curriculum instead of just setting the expectation. Then reps: real workflows, shared openly, with AI as the default way work gets done. Hire for it, and make sure automating your own job is rewarded rather than quietly punished.