Most business owners think they’re using AI.

A few ChatGPT chats. Claude for emails. Maybe a YouTube video about agents.

That’s not using AI. That’s keeping up.

The actual capability is somewhere else. Agents that build software while you sleep. Browsers that fill out forms for you. Systems that pull data from five tools, write the report, schedule the meeting, and send the follow-up before you’ve finished your coffee.

Most operators haven’t touched any of it. They opened Claude Code, got intimidated, closed it. They heard the word “agent” and decided that part was for their team to figure out.

Then they got frustrated their team wasn’t figuring it out.

This morning a friend told me exactly this. Frustrated. His team wasn’t using it.

I asked him how he was using it.

The conversation got quiet.

He’s “using” AI. A few prompts a day. The same level his ten-year-old uses it for homework.

A few days earlier, another operator told me the same thing. Same complaint. Same shallow use under the surface when I pushed.

Their teams aren’t going deeper because nobody is going deeper. Including the owner.

This is the pattern.

Owners are mad at their team for not adopting “real” AI while sitting on top of the same shallow use. Surface prompts on both sides. Frustration projected downward. The team becomes the target.

The team isn’t the problem. The team is doing exactly what the owner is doing.

The frustration is the tell.

It’s psychologically cleaner this way. Being frustrated at your team is a manageable feeling. You can hold a meeting. Write a memo. Change the org chart. You feel like you’re doing something.

Being frustrated at yourself for ducking the depth of the most important transformation of your career is harder to sit with. So the frustration gets projected outward.

Here’s the part you can’t dodge.

You cannot push depth you haven’t pulled.

If you’ve only used the surface, you don’t know what the surface is missing. You don’t know what the real capability looks like. You don’t know what to ask your team to learn because you don’t know what’s available to learn. You’re frustrated they aren’t doing something you couldn’t accurately describe if pressed.

The agentic stuff is intimidating. Claude Code. Agents that run for hours and make their own decisions. Workflows that span multiple tools without your hand on the keyboard. The first time you see it, you don’t know where to put your hands. The interface is unfamiliar. The mental model is new.

Most operators see that wall and back off. The embarrassment stays unspoken.

Then they wonder why their team won’t climb a wall they themselves wouldn’t.

Most leadership decisions can be delegated. This one can’t. The AI transition is the rare moment in a leader’s career where the learning curve has to go through them. Not around them. Not below them. Through them.

While operators sit in this dynamic, other operators are moving.

This week Zeb Evans, the CEO of ClickUp, posted that he cut 22% of headcount. Not to cut costs. To rebuild the operating model. He’s introducing $1M salary bands for people who create outsized impact using AI. He restructured the entire org around what he calls 10x people doing 100x work. He’s not talking about “use ChatGPT more.” He’s talking about agent managers, judgment over output, restructuring around what AI actually does at depth.

(Read his full post.)

You don’t have to agree with the specifics. You don’t have to copy the move. Notice the shape. He’s the one making the structural call. He’s the one defining the new operating model. He’s the one going first.

He didn’t tell his team to start using AI and then get frustrated when they didn’t.

He used it himself. Saw what changed. Rebuilt the company around the new reality.

That’s leadership in this moment.

If the most you’re doing is being annoyed at your team, the gap between you and the operators actually moving is opening up faster than you realize.

It’s not “we lag a quarter.”

It’s structural. Who you hire. How you pay. What the org chart looks like. What next year’s revenue model actually is. Whether half the roles in your business still exist in eighteen months.

You don’t catch up by getting your team to use ChatGPT more.

You catch up by going past the surface yourself. Opening Claude Code. Building an agent. Hitting the wall, pushing through anyway. Doing the work yourself.

Then leading from what you actually learned.

Next time you catch yourself frustrated your team isn’t using AI, stop. Ask yourself the question I asked my friend.

How are you using it?

And then ask the harder one.

How deep are you actually in?

If the answer is “not really,” you’ve found the problem.

And it’s not your team.