With appreciation to Naomi Lippin and members of the AI and coaching discussion forum
Psychological safety is suddenly everywhere in AI conversations — but I’m not sure we’re asking the harder questions yet. Are we really considering how AI changes power, voice, visibility and risk at work? Or are we layering new tech onto old habits and hoping for the best? A recent MIT Technology Review Insights paper makes it clear: where people don’t feel safe to question, experiment or admit uncertainty, AI innovation either stalls or goes underground. This isn’t a technology issue; it’s a human one. If AI adoption is accelerating fear, silence or performative confidence, psychological safety isn’t optional — it’s fundamental. Read the MIT paper and access the report
And the new book Data is Labor by James Felton Keith also raises some challenging thoughts.
Why we need to talk about this
AI adoption in organisations isn’t just a technical shift — it’s a cultural transformation.
The MIT report highlights that psychological safety is one of the strongest predictors of whether AI initiatives succeed or stall.
Many leaders say their organisations encourage experimentation, yet fear, ambiguity, and mixed messaging still block progress, according to the MIT paper.
For coaches, this creates a unique opportunity: helping leaders navigate uncertainty, communicate clearly, and create environments where people feel safe to learn, question, and experiment with AI.
What the research found
Psychological barriers outweigh technical ones
Nearly 1 in 4 leaders has hesitated to lead or suggest an AI initiative from fear of failure or internal backlash. Even in companies reporting “high psychological safety,” people still hold back when the stakes feel high.
Experimentation comes with an asterisk
Many organisations say experimentation with AI is welcome. But the reality is that culturally, missteps still carry personal reputational risk. And while teams often believe they can fail… they can, just not on anything visible.
Communication clarity is the biggest safety lever
The factor leaders say would most improve psychological safety is clear communication about how AI will — and will not — impact jobs. The danger is organisations emphasise the benefits of AI without addressing limitations or risks, and with vague or no guardrails, trust erodes. We therefore need to focus on and agree boundaries and understanding of what we are contributing and where this information is implicitly shared.
Fear and danger hides in unexpected places
Interesting stories are emerging of organisations trying to adopt a considered approach to the use of AI. Using a closed Google system, a small pilot prevented some academics being able to access it (the closed system). In the meantime, other academics keen to adopt the new technology ended up using an open system that meant data was uploaded to the LLM. The result, people were:
-
- concealing their AI use (not wanting to appear “cheating”)
-
- struggling with ‘future shock’ as new tools roll out faster than they can adapt they don’t want to be left behind.
-
- a fear of failing and blame when things go wrong
All of these create behavioural patterns that coaches see every day.
Consideration of who benefits
Real concerns are that LLMs are being trained on real data and users are oblivious to this. This is laid out clearly in Data is Labor by James Felton Keith. Felton Keith raises an important challenge that many of us are skating past too lightly: every time we use widely available AI platforms, we are quietly contributing value to mega‑corporations, often without pausing to consider the scale or implications of that transfer. Our prompts, patterns, thinking, and expertise are not neutral inputs—they are data, insight, and learning fuel. The question Keith invites us to sit with is not whether AI is useful (it clearly is), but whether we are being sufficiently deliberate about who benefits, how value flows, and what we are trading in the process.
High‑performing AI organisations do one thing differently: they actively track the link between psychological safety and AI outcomes. Among companies that moved 76–100% of AI projects into production, 63% formally monitor this connection.
-MIT Report
Why this matters for coaching
As coaches, our work is to help leaders face uncertainty without being paralysed by it. We support them in saying “I don’t know” without fear, in modeling curiosity rather than chasing perfection, and in framing AI as an opportunity for growth rather than a threat to identity or status. We help leaders create the conditions where some discomfort is not only tolerated but useful, which is of course, the space where new thinking emerges. And critically, we work with them to build environments where learning and iteration are visible, valued, and celebrated, rather than hidden away to avoid judgement. Open dialogue facilitates growth.
An important message of the MIT paper is that AI is not going to magically shift (anything) workplace cultures for the better. In fact, the reverse. It will simply amplify existing norms. And the impact for psychological safety is clear.
AI amplifies existing cultural dynamics. It doesn’t build new ones.
Coaching enables leaders to build trust, clarity, safety, and adaptability — the conditions under which AI adoption can thrive.