Josh's firm has about eleven people. Three of them use AI, plus the owner recently. The other seven work the way they worked before.
I have seen this ratio nearly everywhere. When I was still inside a company, it was me and a couple of others, and everybody else came to us with questions. The most common answer we gave was a question: well, did you ask AI?
The split is remarkably consistent. What is worth understanding is what actually causes it, because most explanations are wrong.
It isn't age
The easy story is generational. It doesn't hold up. Josh's owner started using it. I've watched people in their sixties take to it faster than people in their twenties.
It isn't intelligence or skill
Also tempting, also wrong. The seven people at Josh's firm not using AI are competent accountants doing precise work correctly. Nothing about their capability explains it.
It is time and disposition
Josh's own explanation was better than mine. He said he and the others tend to look at the new shiny object and play with it, and try to come up with a case for it. And that so much of accounting work is cut and dry and systematized. If you're busy and don't have time to look into new tools, it's hard to find the space.
That is the whole thing.
Some people, when handed an unfamiliar tool, poke it. Others, when handed an unfamiliar tool, put it down and return to the work in front of them, which is urgent and which they already know how to do.
Neither is a flaw. The second group is why anything gets finished.
But the first group accumulates an advantage that compounds quietly, and the second group has no idea it's happening, because nobody is measuring it and nothing looks broken.
The self-description problem
Here's a detail I found revealing.
Josh described himself as a late adopter. I pushed back on air, because it isn't true. He is one of the three people at his firm using AI, he has tried to run local models, he wired up experiments, he blue-screened his own computer twice trying to get one working.
That isn't a late adopter. That is someone deep in the weeds who happens to be careful about what he deploys at work.
I think a lot of people misfile themselves this way. They are cautious about production and mistake that for being behind. It is worth checking whether the story you tell about yourself matches what you actually do, because "I'm not a tech person" becomes a real constraint the moment you believe it.
Closing the gap without a training program
The instinct is to schedule training. In my experience that produces attendance, not adoption.
What works better is smaller.
Make the one or two visible. They already exist. They already have workflows worth copying. Right now that knowledge is stuck with them because nobody asked. Fifteen minutes in an existing meeting, one person showing one real thing they did this week, does more than a workshop.
Give people a specific first task, not a tool. "Try AI" is not actionable. "Take that PDF you retype every month and ask it to pull the numbers into a spreadsheet" is. The busy group doesn't lack curiosity, they lack a reason that fits inside a workday.
Protect the time explicitly. Josh named this exactly. If the space to try things does not exist, it doesn't appear through encouragement. Someone has to give it.
Let them ask AI how to use AI. Both Josh and I landed on this independently as the advice for beginners. Describe your job, ask for a few safe ways to use AI in it, then do one. It is a strange loop and it works, because everyone's job is different enough that generic advice fails and specific advice is exactly what the tool is good at generating.
Why this matters beyond productivity
If you only care about output, this is an efficiency problem and a mild one.
If you care about the people, it is something else. The gap doesn't stay a tools gap. Over a few years it becomes a difference in what people are trusted with, and eventually in what they're paid.
That happens without anyone deciding it, which is exactly what makes it worth deciding about. The people in your building who are heads-down doing the work are not the ones who should be quietly falling behind, and they're the most likely to.
You almost certainly know who your one or two are. When did you last ask them to show everyone else what they know?
From a conversation with Josh on J&C Unscripted.