Dashboards Don’t Show Who Feels Safe Learning
Most organizations have done the visible work of an AI rollout. Their dashboards tell a reassuring story about who has access and who completed the training. What they can’t tell you is who’s willing to actually experiment with AI in their day-to-day work—and risk looking incompetent while doing it.
When we search for the barriers to AI adoption, we reach for capability, relevance, or legitimate caution. All of these are real. But I’ve come to believe there’s another barrier we talk about far less: permission.
I recently tested this theory on a panel at the Dive In festival with HR and business leaders who were generous enough to challenge it. One conclusion stayed with me: AI doesn’t feel equally risky to everyone, which makes permission an inclusion issue as much as an adoption one.
It’s also a leadership issue. An organization can give people access to AI, but their managers tend to be the ones who influence whether they actually use it. And unlike a dashboard, a manager can see who feels safe using it and who doesn’t, and help close that gap.
What Is the AI Permission Gap?
By permission, I don’t just mean what’s written into your policies. I mean the permission that people read from the organization and the leaders around them. What happens here when someone experiments and it goes wrong? What happens when they admit that they don’t know?
Formal access answers the question, “Am I allowed?” Felt permission answers a harder one: “Is it safe for me to be seen learning?” The difference shows up in the calculation people make. Some ask what AI could help them do. Others ask what AI could cost them if they get it wrong: their credibility, their reputation, maybe even their sense of self.
The distance between access and that felt freedom is what I’m calling the AI permission gap.
Formal access answers the question, “Am I allowed?” Felt permission answers a harder one: “Is it safe for me to be seen learning?'”
Why AI Access Alone Doesn’t Necessarily Mean Adoption
Learning anything new means being visibly less capable for a while. With AI, that dip happens in public. People ask basic questions, share imperfect outputs, and admit uncertainty in front of colleagues who seem further ahead. For anyone whose reputation rests on having the answer, that can feel like a genuine risk.
Identity plays a part too. If you’ve spent years as the expert everyone comes to, AI can unsettle the very thing that earned you credibility. Fear sits underneath much of this. In our Great Leaders Think Differently research at GP Strategies®, 30% of leaders said fear of failure or making mistakes was their top barrier to a growth mindset. And Pluralsight’s 2025 AI Skills Report revealed that 91% of C-suite executives admit they’ve pretended to know more about AI than they actually do.
So what looks like employee resistance to AI adoption is often something more calculated. The permission gap is one lens, and it’s worth using precisely because it’s the barrier people are least likely to name out loud.
Where Cultural Barriers to AI Adoption Show Up
People rarely open a policy document to decide what’s allowed. They look for clues in what happens around them. So while many organizations tell people to go out and experiment, what they believe they can safely do doesn’t always match this message. And permission isn’t granted just once. People continually recalculate it from what they see happening around them.
For leaders, the permission gap shows up in small, everyday moments on their teams. What happens when someone gets it wrong, admits they don’t know, or spends time trying something that doesn’t immediately work? The manager’s response in those moments tells everyone watching far more about what is genuinely safe than the policy does.
When Someone Gets It Wrong
“Talk me through what you tested, what you learned, and what you would change next time.”
Early, imperfect attempts are where people learn the most. If those early attempts get shut down, people don’t just stop experimenting, they do something worse: they start pretending to use AI rather than genuinely adopting it to change how they work. My colleague Ella Richardson called it “performing AI theater,” and wrote about how fear drives this kind of performance, all the way up to executive leadership. Watch for the person who seems to be using AI but never talks about how.
When Someone Admits They Don’t Know
“That is useful to know. Which part feels unclear, and who else might benefit from us working through it together?”
Managers are the ones who tend to influence actual adoption because they are who their team members look to for that unwritten permission. Admitting you don’t know is where that permission gets tested in front of others. You might see this in the person who comes to talk with you one-on-one about how they’re struggling with AI, but wouldn’t dare to share that with the whole team.
When Something Doesn’t Work Right Away
“I’ve noticed our AI experiments are being led by a small group. I want to understand what would make it easier for others to take part.”
A mentor once told me that opportunity can’t be equal if the cost of failure isn’t. That line has been going around in my head. Spending time on something that might go nowhere is a risk not everyone can afford equally. People who are new to the organization, or who feel they have to prove themselves, may not feel able to overstep, let alone fail in public.
If the invitation to experiment becomes clearer but only those who already feel confident act on it, you get pockets of strong adoption instead of adoption across the board. A useful check is to notice who always volunteers for AI pilots and champion networks, and who never does. And before assuming someone lacks confidence, ask: what is making this difficult? Is it uncertainty about the rules, concern about the outcome, lack of time or support, or the risk of being seen getting it wrong?
Opportunity can’t be equal if the cost of failure isn’t.
How Leaders Build Confidence to Experiment with AI
So, as leaders, how do we mind the AI permission gap? In my opinion, through small, visible moments, and these are the ones I’d prioritize.
Learn out loud, and go first. None of us has all the answers about using AI well. With my own leaders, we open every meeting with a learn-out-loud moment: what I tried this week, what worked, what didn’t, and where I’d love your perspective. It takes a few minutes and shows that the learning journey, not just the finished result, is worth sharing. Try opening your next team meeting the same way.
Protect early attempts. The way a manager responds when something doesn’t work is one of the moments when the team learns whether experimentation is genuinely safe. When someone shares an imperfect output or spends time on an approach that goes nowhere, get curious about what they tried and what they learned.
Make the expectations clear. Explain what AI can currently be used for, where human judgment remains essential, and what is still undecided. Talk about how you expect their work to change, what AI will take on, and the value you’re counting on them to bring. Don’t manufacture certainty you don’t have. Take unresolved questions upward and tell your team what you are doing to get answers. Clarity makes it easier to know where trying something new is welcome.
Notice who isn’t joining in. The permission gap doesn’t just affect quieter people. Your most confident and vocal team member may still feel exposed using AI. Look at who isn’t experimenting, sharing, or asking questions (whatever the reason), and invite them in directly.
What Managers Can Influence, and What They Need from the Organization
Managers shape felt permission, but they can’t create it alone. They can make it safer to ask questions, admit uncertainty, test new approaches, and share work that is not yet polished. They can respond to mistakes with curiosity, make learning visible, and ensure that experimentation is not limited to the most confident voices.
But managers also need clear guardrails, access to approved tools, time for people to practice, and honest communication about how AI may affect roles and expectations. If the organization rewards only immediate productivity or leaves managers to interpret an unclear strategy, even the most supportive will struggle to make experimentation feel safe.
Closing the AI permission gap therefore requires shared responsibility. Managers shape the experience within their teams. Senior leaders in the organization must create the conditions that allow them to do it well.
Managers shape felt permission, but they can’t create it alone.
BRAVE as a Practical Tool to Support Inclusive AI Adoption
At GP Strategies, we use the BRAVE framework to help managers distinguish between the conditions they can influence and the support they need from the wider organization. Use these five questions as a quick check:
- Belonging: Can people learn here without losing credibility or standing?
- Relevance: Can they see where AI adds value to their work?
- Access: Do they have the approved tools, time, and support to practice?
- Visibility: Do they see leaders and colleagues learning openly, not just presenting polished results?
- Empowerment: Do they know where they have freedom to experiment and where the boundaries are?
You may be able to address some of these conditions within your team. Others will require clearer policies, resources, decisions, or support from elsewhere in the organization. The purpose of the check is to help identify what can change, what needs to be escalated, and where people may be experiencing barriers that adoption data won’t show.
Look for progress in the work, not just in license activations. Who is experimenting, sharing what they are learning, and participating where they previously held back?
Who Has Permission to Experiment on Your Team?
The value of AI isn’t about the technology alone. It comes from people bringing their judgment, expertise, creativity, and context, and being willing to challenge what AI produces. If your people don’t feel they have permission to experiment, your team won’t get the full value of human+AI work.
This week, I challenge you to think about the last time someone showed you an imperfect AI output. Did your response make them more likely to experiment openly next time, or more likely to hide the process and show you only the finished work?
If you’re ready to close the permission gap, explore how we help leaders guide their teams through AI-driven change.
About the Author
