The most useful thing I learned this month came from shutting up and listening.
For six months my co-founder and I were heads down building. Fast feedback, agents that do what you tell them, the rush of making the thing you wished existed. Then we went out and had a hundred real conversations in two weeks. Founders, operators, enterprise execs. People at three-person startups and people at three-thousand-person companies.
I expected to come back feeling behind. I came back feeling the opposite.
My biggest learning: almost no one is doing anything crazy with AI. The feed makes it look like everyone is ten-x-ing their output and running fleets of autonomous agents. The revenue numbers from the model companies make it feel like a rocket you already missed. But when you talk to actual people doing actual work, the picture is completely different.
Start with engineers, the group supposedly furthest ahead. They are ahead, and it is real. But the average engineer I talked to is getting maybe two to three times their old output, not ten. They have Claude Code or Codex open, they are shipping faster, and that is mostly where it stops. Their setups are simple. Their work stays narrow. A handful of people are doing genuinely impressive things. Almost everyone else is good, not superhuman. At bigger companies it is worse, because the company handed everyone a license and changed nothing else about how work happens.
Then look at the business side, and the gap is a canyon. I assumed marketing, sales, and ops were maybe six months behind engineering, with 3x output right around the corner. They are not close. Nearly everyone on the business side is doing little more than chatting with a model. The few who had something better, a daily brief that writes itself or an agent that analyzes deals, almost always had someone else build it for them. The rare exception was a person who genuinely loves the tech and spent hundreds of hours figuring it out alone.
For the executives at established companies, this has stopped being interesting and started being scary. I talked with a CRO at a four-thousand-person company who is, by his own description, panicking. They bought licenses for everyone. Engineering output went up. The entire business side of the house got nothing. He can see the gap, he knows he has to close it, and the tools to close it do not exist yet on the market.
That is the real finding. The enterprise leaders I talked to are spending somewhere between five hundred thousand and two million a year on AI, and their business teams are no more productive than they were six months ago. Leaders look at engineering getting 3x, look at their business teams getting nothing, and assume the business teams will catch up. They will not. Not with what exists today.
You can hand a business person Cowork, which is essentially Claude Code in a sandbox, and assume they will now operate like an engineer. They will not, for a simple reason: it has a cold-start problem. It is powerful if you know how to set it up and you are willing to invest the time. Business people have neither. So they open it, type a question like it is a chat box, and get nothing that changes their week.
Here is an important distinction: Business people do not work like engineers. Different rhythms, different mindsets, different days. Take a tool built for engineers and rename it, and the engineer's leverage does not transfer. It is why so many people tried Cowork once, used it to rename a few files, and quietly went back to chat. They are not lazy, and they are not short on ambition. They were handed a tool that was never shaped for how they work. What is missing here is not a smarter model. The models are already good enough. What is missing is anything actually built for how non-engineers spend their day.
You could read all of this as bad news, as proof that AI is overhyped. I read it the other way. The value is real. The money is there. The fear is real enough to drive urgency. The only thing missing is the product, and missing products get built.
So if you feel behind, sit with this: almost everyone feels behind, and almost no one is actually ahead. The race you think you are losing has barely started. The people who give non-engineers real leverage, not a renamed engineering tool, win the next wave of this. That is not a reason to panic. That is the whole opportunity.