Psychological safety and AI adoption: why one determines the success of the other
Part 4 of a 4-part series: Activate psychological safety.
There is a question that every organisation navigating AI adoption will eventually face. Not whether the technology works, but whether the people using it feel safe enough to actually use it.
That is a psychological safety question. And the answer to it will determine more about the success of AI adoption in your organisation than any tool selection, implementation plan or training program you put in place.
What the data is telling us
The research on AI adoption failure is consistent and striking. Organisations are investing heavily in AI tools and seeing very little return. The technology is deployed. The training is delivered. And then, quietly, people do not use it.
The explanations that surface are usually framed as resistance, scepticism or a skills deficit. These are real factors. But they are also symptoms of something more fundamental. When people do not feel safe to experiment, to get things wrong in front of others, to admit they do not know how to use something yet, they do not try. They wait. They watch. They find workarounds that let them look competent without taking the risk of visibly learning something new.
That is not resistance. It is self-protection. And it is a predictable response to any environment where psychological safety is low.
The 2026 COE Psychological Safety Study, drawing on clinical insights from workplace wellbeing professionals across 47 countries, confirms that this pattern is both widespread and worsening. In 2025, organisational culture placed greater emphasis on authenticity, trust, fairness and psychological safety, but the gap between what organisations say and what employees actually experience remained significant. Research from Perceptyx found that executives report 87% positive perceptions of psychological safety, while individual contributors lag at just 69% and hourly workers at 66%. That 18 percentage point gap between those setting the agenda and those living it does not disappear during an AI rollout. It widens.
The specific ways low psychological safety undermines AI adoption
In the first article of this series, we explored why only 3 in 10 employees strongly agree their opinions count at work. In the second, we looked at how leadership behaviour either builds or erodes the conditions for people to speak up. In the third, we examined what happens after a mistake and why error culture is the clearest signal of psychological safety in any team.
Each of those dynamics plays out directly in AI adoption.
When employees do not feel their opinions count, they do not raise concerns about how AI is being implemented. Legitimate worries about workload, job loss, role clarity, data privacy and fairness go unspoken. Leaders interpret the silence as acceptance. They proceed on the assumption that people are on board. The gap between that assumption and the reality on the ground grows.
When leadership behaviour erodes psychological safety, the conditions for learning disappear. AI adoption requires people to be beginners in public. It asks them to try things that might not work, to ask questions that might expose gaps in their knowledge, to move through a learning curve in view of their colleagues and managers. In teams where that kind of visibility feels safe, people do it. In teams where it does not, they perform competence instead.
When error culture is poor, the feedback loops that make AI implementation better do not exist. The team that is struggling with a particular tool does not say so. The workaround that half the team has developed does not surface. The process that the AI was meant to improve but has actually complicated does not get raised. The implementation proceeds on the basis of incomplete information, and the gaps compound.
What this means for leaders
The implication is not that psychological safety needs to be perfect before an organisation can adopt AI. It never will be, and waiting for it would be a mistake.
The implication is that psychological safety is not separate from the AI adoption agenda. It is central to it. The organisations that treat these as parallel workstreams, one focused on technology implementation and one focused on culture, will find that both stall. The ones that understand them as the same workstream will find that progress on one accelerates progress on the other.
Practically, this means a few things.
Leaders need to model uncertainty openly. Not performatively, but genuinely. The leader who says they are still figuring this out, who asks questions they do not already know the answer to, who acknowledges that some of what they are implementing will need to change, is doing something important. They are demonstrating that it is safe to not know yet. That is the permission slip the rest of the organisation needs.
Teams need space to experiment without assessment of their performance. The brain cannot learn well when it is in a threat state. Creating conditions where people can move through the learning curve without fear is a hard requirement for adoption to actually work.
And organisations need to measure the right things. Licences activated tell you very little about whether adoption is happening. The more useful questions are whether people feel safe enough to use AI tools openly, whether they are raising concerns and having them addressed, whether the implementation is improving based on honest feedback from the people doing the work.
Where this leaves us
This is the fourth article in the Activate series. Over these four pieces we have explored what psychological safety is, why it is a leadership behaviour rather than a culture initiative, why error culture is its clearest signal, and why it is the precondition for AI adoption to work.
The thread connecting all of it is the same. Psychological safety is not a program. It is not a value on a wall or a box in a culture survey. It is the accumulated result of thousands of small leadership moments, each one either building or eroding the conditions in which people feel safe to be honest, to learn, to take risks and to tell you what is actually happening.
In an era of sustained uncertainty and rapid technological change, it is the most important organisational capability you can build.
Helena Kuo is Managing Director of Enmasse. This is the fourth article in her Activate series on psychological safety in the workplace, following on from:
- When only 3 in 10 employees speak up that’s a leadership problem
- Psychological safety is not a culture initiative, it’s a leadership behaviour
- The silence that should worry you most is the one after a mistake
If you would like to explore how Enmasse can help your organisation build genuine psychological safety, through leadership development, diagnostics or our Activate program, get in touch with us to discuss some options or explore our leadership and culture solutions.
References
Edmondson, A. C. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2), 350–383.
Gallup (2025). State of the Global Workplace Report. Washington, D.C.: Gallup Inc. Available at: gallup.com
Center for Organizational Effectiveness (2026). The COE Psychological Safety Study. Workplace Options. Available at: psychsafety.workplaceoptions.com
Perceptyx (2026). The Psychological Safety Gap: Why 30% of Employees Stay Silent. Available at: blog.perceptyx.com*
