Psychological safety is the belief that you will not be punished or humiliated for speaking up. It covers asking questions, admitting mistakes and raising concerns.
Amy Edmondson's research shows it is a strong predictor of team learning. Google's Project Aristotle found it was the most important factor in its effective teams.
Safety decides who starts
In 2026, Reich and colleagues surveyed 2,257 employees of a global consulting firm. Psychological safety reliably predicted whether people adopted AI tools at all.
It did not predict how often they used AI once they had started. The effect was the same across roles, experience and regions.
The message for leaders is clear. Safety matters most at the start, when people decide whether to try.
AI can erode the safety it needs
The relationship runs both ways. A 2025 study of 381 employees by Kim, Kim and Lee found AI adoption lowered psychological safety. Lower safety, in turn, was linked to higher depression.
Ethical leadership softened the effect. Leaders who reduced uncertainty and offered real support protected their teams.
So the teams that most need safety during AI change are the ones most likely to lose it.
What blocks safety in practice
Pressure arrives before skill. Productivity targets often rise before people have learned the tools. People then hide difficulties to keep up.
Knowledge gaps become status threats. Some people become fluent fast. Others fall behind and perform a confidence they do not feel.
Managers project false certainty. Many managers are unsure about AI themselves. Pretending otherwise teaches the team that pretending is safer than admitting.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
Why it matters for results
Without safety, people experiment less, stay quiet when AI output looks wrong, and hide their mistakes. They also use AI in secret.
Hidden use is a clear sign the change has not landed. Whether people take on the change and make AI part of their work is the Adoption condition in The Human Advantage Programme.
Six practices that build safety
Hold honest conversations. Name the opportunities and the worries. Go first, and do not rush to reassure.
Share your own learning. Tell the team what you do not yet understand and where you got it wrong.
Meet resistance with curiosity. Ask what someone finds difficult. Resistance often reveals a real problem with the tool or workflow.
Agree how to disagree with AI. Decide together what happens when someone thinks the AI is wrong. Then nobody has to defer silently.
Act on concerns. Investigate what people raise and report back. That shows speaking up is worth it.
Look after yourself. Leaders under strain struggle to steady others. Be honest about your own limits and get support.
Try this
Ask your team three questions anonymously. How safe do you feel raising concerns about AI? Do you feel heard? Does your expertise still matter?
Repeat them each quarter. For a fuller picture, take the Adoption Condition Diagnostic.
