The Real AI Adoption Gap Has Nothing To Do With Access

This voice experience is generated by AI. Learn more.This voice experience is generated by AI. Learn more.Andrew Antos is CEO & co-founder of Within, giving enterprises the organizational memory AI needs to understand how work actually happens.

getty​Most companies treat AI adoption as an access problem: Get everyone the tools, train them on prompt engineering, roll out Copilot companywide.

A senior BCG partner reframed how I think about that in one conversation.

A recent global workforce survey from PwC found that only 14% of employees use generative AI daily. That’s despite 54% saying they’ve used it for their job at least once in the past year. That’s not a story about access. Most of the workforce has already tried the tools; they just haven’t made it a daily habit.

What surprised me most: In every rollout I’ve been part of, daily usage doesn’t correlate with age. Younger employees aren’t uniformly ahead of the older ones. Access and training don’t explain it. The pattern has a different driver.

I’ve seen the same pattern play out across every organization I’ve worked with. There’s always a group, somewhere between 5% and 20% of the company, who use available tools 10 times more than everyone else. I call them Change Agents.

They’re not defined by title, tenure or technical skill. A psychological profile defines them. They’re the people who naturally tinker. They see something broken and feel compelled to fix it. They care about how work gets done, not just that it gets done. They exist at every level from individual contributors to managers to directors. Some have been at the company for 15 years, while others just started.

And right now, AI is giving these people a superpower.

In field after field, top performers already operate at multiples over the average. That gap existed long before AI ever entered the picture: Their output and quality sit in a different category altogether.

Now give that same group AI tools, and you’re stacking one force multiplier on top of another. The distance between your best and your average keeps growing.

This is why the AI adoption conversation is broken. Companies are optimizing for the wrong thing, trying to move everyone from 0% to some baseline of usage, when the real unlock isn’t getting 100% of people to use AI a little. It’s identifying the Change Agents and giving them room to work.

The people driving daily usage rarely need more AI training. What moves them is being seen and recognized for the improvements they’re already making, and given the permission and platform to share what’s working.

In my experience, people repeat behavior that gets recognized far more reliably than they stop behavior that gets punished. The problem has always been that you need to observe the behavior you want to reward, and in digital work nobody can see your screen. A manager can tell you what someone delivered, but they rarely know how.

AI changes that equation, though the framing matters. Don’t use it as a surveillance tool—people will reject that immediately, and they should. Show them a mirror instead that makes great work visible: what they actually accomplished and shared only if they choose to share it.

So if you’re an executive reading this, stop asking how we get everyone to adopt AI and start asking who the Change Agents are that we’re failing to unleash.

They’re already in your building. They’ve been there the whole time. They’re the ones who see opportunities and feel frustrated that nobody’s acting on them, and the ones who fix a process that was driving everyone crazy, though nobody notices.

AI didn’t create these people. But it will only widen the gap between organizations that unleash them and organizations that don’t.

The 14% aren’t early adopters. They’re your future.​

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?