Attention AI Transformation Leaders: Half Your Adoption Is Invisible.

What hidden AI usage says about your culture (and how to change it)

Last week I was chatting with the head of operations at a mid-sized company. Forty minutes into a discussion about her team’s adoption challenges, said something that caught me by surprise..

“Honestly, I’ve been running my prep work through Claude for the last six months. I just haven’t told anyone.”

She wasn’t bragging. She wasn’t asking permission. She was confessing something that should have been surfaced to her organization, but she chose to keep it quiet.

I have been hearing a version of this conversation since I started working with AI. While shadow AI usage isn’t as pervasive as it once was, this person still described their workflow in lowered tones, the way you might mention you’ve been secretly feeding steak to the dog. As if it might count against them.

The AI split underneath the obvious one

I have written about the bifurcation problem, where there is a split between power users and everyone else. To solve this, I share why naming AI Leads isn’t enough. This obvious split is the one most leaders can describe in their sleep. Twenty people in the organization are doing all the interesting AI work. Two hundred are still asking someone else to do it for them. We name AI Leads. We fund training. We hope the gap closes.

But there is another divide underneath the visible one. According to the Awareways Trend Report 2025, 48.8% of employees who use AI at work hide that usage from their employers. Their stated reason is fear of judgment. Almost half. Hiding.

That number shines a light on why it’s so hard to measure AI impact. The split is not only between power users and everyone else. The split is between the people who can use AI out loud and the people who use it in private. The first group gets named. The second group gets missed.

Which means most of what we measure when we measure AI adoption is wrong. The dashboards in the AI program office count the visible users. The actual user base coud be roughly twice that size, and the second half is invisible by design.

Why people hide their AI usage

If you have not had this conversation directly, you might assume people hide their AI use because they are unsure whether it is allowed. That is part of it. It is also the safest version of the answer, and it isn’t the full one.

The deeper reasons have nothing to do with the tools.

People hide because their manager has not named AI as something the team is supposed to use. The WRITER 2026 AI Adoption Survey is sharp on this. Only 35% of employees say their manager is an AI champion. And 80% of Gen Z employees in the same survey say they trust AI more than their manager for work tasks. (Sit with that one for a second.) When the person you report to has not opened ChatGPT, you do not bring it up. You quietly use it and ship faster work without explanation.

People hide because they fear judgement. AI amplifies skills - it can make a deliverable look more polished. But those same individuals have watched what happens when someone admits they used AI for a deliverable. The work gets discounted. The question shifts from ‘is this good’ to ‘is this real.’ Once you have seen that happen, you stop signing up to share what’s behind your good work..

So, I ask: If half of your team was using AI privately and never told you, what signal would even reach you? What metric do you have in place that would catch this? And..most importantly, does it matter? If so, why or why not?

Most organizations I talk with measure AI use through licenses purchased, official tools rolled out, training hours logged. None of those measurements pick up the person doing real work in a personal ChatGPT tab on the side.

What the hiding AI usage costs organizations

This is where it gets interesting and potentially, a little sad.

The ActivTrak 2026 State of the Workplace report names a pattern that we should be talking about more often. Burnout risk in 2025 fell 22%, to just 5% of employees. Disengagement risk, in the same period, rose 21% to nearly one in four. Their framing was precise. ‘These aren’t employees who are checked out. They’re employees whose capacity isn’t being used.’

I want you to hold that next to the hiding number.

If half your workforce is using AI privately to get their existing work done faster, and the time savings go invisible because they cannot acknowledge how they got there, that capacity has nowhere to go. It does not get reinvested into the next big thing. It does not show up as a contribution in their next one-on-one. It sits there, quietly, and over months it becomes disengagement.

We were told AI would free people up to do their most meaningful work. For a meaningful share of employees, what it actually freed up was time they cannot tell anyone they have. And capacity you cannot name is capacity you cannot use.

The standard response to the divide problem has been to train harder, more, run another lunch and learn. None of that addresses the actual diagnosis. You can train every employee on every tool, and if your culture rewards visible non-AI work and inadvertently punishes visible AI work, you have half your workforce hiding and won’t know what is going on.

The cultural work around AI is much harder and slower than learning to prompt. It looks like managers modeling AI use in front of their teams. It looks like sharing hours where people can showcase their wins and their failures with AI. It looks like designing the AI Activation Hubs to create safe surfaces for everyone’s usage. These are not technology decisions. They are leadership decisions.

Final Thoughts

When the head of operations told me she had been using Claude in secret for six months, she inadvertently revealed the culture in her organization. The interesting question to ponder is what her organization did over those six months that made hiding feel like the safer move.

Has someone in your organization told you something they had been doing with AI for months that you did not know about? Do you disclose your AI usage? Why or why not?

I would love to hear from you in the comments.

P.S. A Limited Opportunity to Address This

If you are leading AI transformation and want a working room to figure out how to make AI use visible and valued in your specific organization, we have 15 spots to learn from each other and the author. Running Hyperadaptive Organizations. brings the concepts of Hyperadaptive to life, providing a learning arena to work through your hardest issues. The cohort runs June 18 and June 25. Founding rate $797 through June 11. hyperadaptive.solutions/class

Based on the book Hyperadaptive: Rewiring the Enterprise to Become AI-Native (IT Revolution Press, May 2026) is available at hyperadaptive.solutions/book.