I’ve had a couple hundred conversations about AI so far this year with founders/owners, ops people, marketers, CS leads, support managers, salespeople, and other GTM leaders. At some point I expected a pattern to show up. In particular, I expected the more technical roles within an org would be the ones deep in it and everyone else would be a bit slower to adopt.
That’s not what I’ve found. There’s basically no correlation between how technical someone’s role is and how enthusiastic about AI they are.
Ignore software engineers for a second, they got pulled in whether they wanted to be or not because the payoff was too obvious. With everyone else, though, the spread is pretty random if you look at existing role or skillset. I’ve talked to fairly technical ops people who still barely use anything past ChatGPT for basically search queries and people in sales roles running three tools I’ve never heard of.
What actually predicts it
Willingness to pick up new stuff. That’s it.
It reminds me of who was first on MySpace or AOL Instant Messenger back in the day. Not the technical crowd exactly. Just people who thought the new thing looked interesting and went down a rabbit hole for a weekend. Hell, I remember the least technical people I knew asking me about HTML/CSS tips for their MySpace pages. This feels like the same vibe.
So when I walk into a company, I’ve stopped guessing who my early adopters will be based on titles. I ask who’s been messing around with something. That list rarely looks like the org chart, and it’s almost never the people leadership predicted.
Workera landed on something similar about why adoption stays lumpy:
The gap isn’t about access to models or tools. It’s about how work is designed, governed, and led.
Access isn’t the thing holding anybody back. Nobody’s stuck because they can’t get a login.
The gap inside one team is enormous
Here’s the number that made me change how I run my AI engagements. On most teams I see, the gap between the most sophisticated AI user and the least is something like 100 to 1 in output.
Same job title. Same tools available. One person shaved six hours a week off their workload and the other one has never opened it.
That gap is where all the available value is, and it’s not in building one deeply automated workflow. If you’ve got 100 people and you get each of them 30 minutes a day back, that’s a bigger number than any single automation project you’ll run this year. And it’s cheaper, because the tools are already paid for.
The deep automation stuff is still worth doing. It’s the most fun and complex part of what we do. But it usually shows up second, once the team has enough sense of what AI does to tell you which of their tasks are worth automating.
The floor moved up on everybody
There’s a side effect of this that I don’t think people have fully absorbed yet.
You used to be able to run a business with a surface-level understanding of the areas outside of your own experience. You had a specialty from whatever career track got you there, and for everything else you hired specialists or consultants. Hire for your weaknesses, and it worked fine.
AI is making those specialist functions accessible to non-specialists, which sounds like it lowers the bar. But it does the opposite for whoever’s in charge. If your marketing person can now produce work that used to require an agency, you need to know enough about marketing to tell whether what came out is any good or if they’re copy/pasting the first garbage AI spits out. Same for ops, finance, support, etc.
And if you’re the one deciding what to automate, you need a cross-functional read on the whole operation down to what an individual contributor actually does all day. Without that you’ll automate a process that doesn’t work the way you think it works. This is the #1 situation I run into when cleaning up some AI messes.
What I do about it personally is boring. I have a running file on the exact details of how every process works, mostly gathered from transcripts of internal calls. Then I pull that into my AI tool of choice when I want to wear that hat. But you need the context to pull from and that starts with interviewing your boots-on-the-ground team to REALLY understand what their workflow looks like, not just at a high level.
If you’re starting with 20 people and no plan
Don’t start with a complicated rollout. Find your three people who already play with this stuff, whatever their titles are, and get them showing everyone else what they do with it. Enable everyone with access and permission to experiment. Then look at where the time is going across the teams and pick the boring repetitive things first to automate.
The complicated projects can wait a quarter or two. Getting your least skilled people from zero to 30 minutes a day is the most valuable place you can start.