Psychological Safety at Work: What It Actually Means (and Why Most Leaders Get It Wrong)

A few years ago, I was coaching a senior leader at a multinational company — let's call her Diane. She had read all the right books, attended the right offsites, and genuinely believed she had built a psychologically safe team. Her direct reports liked her. The energy in team meetings felt warm and collegial. Nobody ever argued. Nobody ever pushed back.

She came to me because her team kept missing the signals. Problems that should have surfaced in week one were showing up in week eight. A critical product flaw that three engineers had privately worried about for months went undiscussed until it cost the company a client.

When I asked her whether people felt comfortable challenging her in meetings, she paused.

"I think so," she said. "Nobody seems uncomfortable."

That was exactly the problem.

The Term Everyone Uses and Almost Nobody Understands

"Psychological safety" has become one of the most overused phrases in leadership conversations. It appears in every culture deck, gets mentioned in every L&D workshop, and shows up in roughly every third LinkedIn post about modern management.

And yet, most organizations are getting it badly wrong.

Harvard Business School professor Amy Edmondson — ranked #2 on the Thinkers50 global list of management thinkers in 2025 — coined the concept in 1999 after a counterintuitive finding in her research on hospital teams. She expected the best-performing teams to report the fewest medical errors. Instead, she found the opposite: the highest-performing teams reported more errors than their lower-performing counterparts.

They weren't making more mistakes. They just felt safe enough to talk about them.

That discovery has since shaped decades of research on team performance, innovation, and organizational learning. And in 2025, Edmondson co-authored a piece in the May–June issue of Harvard Business Review specifically to address the misconceptions that have crept into the mainstream understanding of her work. The fact that she felt she needed to write it tells you something about how distorted the conversation has become.

So let's clear it up.

What Psychological Safety Is Not

This is the part that trips most leaders up.

It is not niceness. Perhaps the most damaging misconception of all. When teams confuse psychological safety with getting along well and keeping things pleasant, they end up with a culture where people stay quiet rather than say something that might cause discomfort. As Edmondson herself put it in Harvard Business Review, "Too many people think that it's about feeling comfortable all the time — that you can't say anything that makes someone else uncomfortable or you're violating psychological safety. Anything hard to achieve requires being uncomfortable along the way."

It is not the absence of accountability. A psychologically safe team is not a team where poor performance goes unchallenged or where standards slip in the name of making people feel good. The research consistently shows that psychological safety and high standards reinforce each other, not undermine each other.

It is not consensus. Leaders do not have to implement every idea or agree with every perspective to have a psychologically safe environment. The obligation is to genuinely listen and to make people feel heard — not to pretend all input is equal.

It is not a policy. Edmondson and her co-author Michaela Kerrissey made a sharp point about the Rhode Island Workplace Psychological Safety Act passed in 2024: "Telling people in a company or on a team that they must have psychological safety 'or else' will not produce it." You cannot mandate a climate. You can only model it.

What it is, in Edmondson's original 1999 definition, is simply this: the belief that you won't be punished or humiliated for speaking up with ideas, questions, concerns, or mistakes.

Simple. But far harder to build than it sounds.

Why It Matters More Than Ever Right Now

The data is not subtle.

In 2012, Google launched an internal research initiative called Project Aristotle — an attempt to figure out what separates their highest-performing teams from the rest. They analyzed over 180 teams, conducted more than 200 interviews, and reviewed 250-plus team attributes. The single most important factor they found was not talent, technical skill, or seniority. It was psychological safety. According to Google's own internal data, teams with high psychological safety outperformed their peers by 27%, showing higher productivity, more innovation, and greater employee satisfaction.

More recently, Gallup's 2025 engagement data showed that only 3 in 10 employees strongly agree that their opinions count at work. A separate 2025 report found that 1 in 4 employees still does not feel psychologically safe enough to speak up — meaning one in every four people on your team may be sitting on concerns, ideas, or early warning signs and saying nothing.

Think about what that costs. Not just in innovation lost. In problems not flagged, in mistakes that compound, in talent that quietly disengages and eventually walks out.

The research from Harvard shows that psychological safety is a particularly powerful buffer during hard times too. A 2024 study by Edmondson and Kerrissey — published in the International Journal of Public Health and drawing on data from over 27,000 healthcare workers — found that in environments of constraint and upheaval, psychological safety protects against burnout and staff turnover in ways that almost no other lever can replicate. In an era when global employee engagement has dropped to 20% — its second consecutive year of decline according to Gallup's 2026 State of the Global Workplace report — that is not a small thing.

A Note From the Rooms I Have Been In Lately

I was in Oslo, Norway at the INSEAD Forum titled Leadership, Balance and Disruption in a New Era recently, and unsurprisingly, AI dominated a large portion of the conversation. It does in almost every executive gathering right now. What struck me was not the enthusiasm for the technology — that was expected — but how many leaders are still trying to figure out the fundamentals of how to integrate it into their organizations while also keeping their people engaged, informed, and on board. The gap between how quickly AI is moving and how thoughtfully most organizations are bringing their people along for the journey is real, and it is worth being honest about.

Last week, I had the privilege to speak on a panel about AI and sales pipeline at an event organized by Cuvama. The conversation kept circling back to the same question: what does AI actually change about how deals get done? And the honest answer, which a few of us on the panel kept returning to, is that the mechanics change but the fundamentals do not. People still buy from people. They still want to feel understood, not processed. They still choose partners and vendors based on something that cannot be automated: the sense that the person across the table actually gets them, that there is genuine trust in the room, that they are not just another entry in a CRM.

The same logic applies inside organizations. People still report into people. They still need to feel like they belong somewhere, that their contribution matters, that the person leading them is someone they can be honest with. No AI system can create that. What it can do — if it is implemented without care — is quietly erode it.

What AI Cannot Fix (and May Make Worse)

This is where I want to be direct, because there is a version of the AI conversation happening in boardrooms right now that makes me a little uncomfortable.

The conversation goes something like this: we can use AI to analyze engagement data, flag at-risk employees, predict attrition, surface communication patterns that suggest low trust. And yes, technically, some of these tools exist and some of them have real value.

But there is a meaningful difference between measuring psychological safety and building it.

And the belief that better data will solve a human climate problem is one of the more seductive and dangerous ideas circulating right now.

Research published in 2025 by Kim, Kim, and Lee — a peer-reviewed three-wave study of 381 employees published in Humanities and Social Sciences Communications (Nature portfolio) — found that AI adoption, when handled without care, actually reduces psychological safety in organizations, increasing employee anxiety and, in some cases, contributing to depression. The mechanism makes intuitive sense: when people do not understand how AI tools are being used to evaluate them, when decision-making feels opaque, when they are uncertain whether their role still matters, they pull back. They go quiet. They protect themselves by saying less, sharing less, and taking fewer risks.

A separate piece in the May 2026 issue of Harvard Business Review — co-authored by organizational psychologists at Boston University and the University of Canterbury — warned specifically that AI cannot replace the benefits of human connection and may erode collaboration, trust, and social skills over time when organizations lean on it for roles it was not designed to play.

A December 2025 global report by Infosys and MIT Technology Review Insights found that 83% of executives believe a culture that prioritizes psychological safety measurably improves the success of AI initiatives. Read that the other way around: the organizations least likely to get ROI from their AI investments are the ones with low-trust, low-safety cultures. You cannot automate your way out of a trust deficit.

AI is a tool. A genuinely useful one, used well. But it needs the human infrastructure around it to work — and that infrastructure is what psychological safety is. The organizations that get this right will use AI to free their people up to do more of the relational, judgment-intensive, trust-building work that no algorithm can replicate. The ones that get it wrong will find that their AI tools are running on top of a culture that is quietly falling apart underneath them.

The Trust Problem Sitting Underneath It All

Psychological safety does not exist in a vacuum. It is built on trust — and right now, trust in leadership is in short supply.

The 2025 People Element Employee Engagement Report found that only 32% of employees trust their leaders to make the right decisions. Only 29% trust their managers. Gallup puts an even sharper point on it: just 20% of US employees strongly trust their leadership.

This is not a numbers problem. It is a behavior problem. And it sits squarely in the lap of leadership.

Trust is not built through town halls, culture workshops, or well-worded values statements. It is built through thousands of small moments: how a leader responds when someone raises bad news, whether they credit ideas to the right person, how they behave when something goes wrong, and whether the way they act in private matches the way they speak in public.

In my work with leaders across industries, the trust gap usually traces back to one of three patterns:

1. The leader who punishes candor without realizing it. Not through anything dramatic — no shouting, no demotion. Just a slightly cooler response to the person who questioned a decision in a meeting. A subtle shift in energy. A tendency to remember the person who disagreed. Over time, the team notices, adjusts, and stops sharing.

2. The leader who created a "tell me anything" culture but never changed a single thing based on what people told them. Feedback became theater. People eventually stopped bothering.

3. The leader who is safe to be around in a calm quarter but unrecognizable under pressure. The version of them that shows up when the numbers are bad is withdrawn, blame-focused, and short. People don't trust the good version because they've seen the other one.

Any one of these patterns can hollow out a culture of psychological safety faster than anything a training program can build.

When It Goes Beyond Dysfunction: Gaslighting and Fear as Leadership Tools

I want to name something that does not get discussed nearly enough in polite leadership circles, which is what happens at the far end of low psychological safety. Because there is a difference between a leader who is simply not great at creating openness and a leader who is actively — even if unconsciously — using confusion and fear to maintain control.

Gaslighting in the workplace is no longer a fringe concept. Research published in Healthcare in December 2025 documented it as a measurable, distinct leadership behavior — defined broadly as a pattern in which those in positions of power undermine their subordinates' sense of reality through denial, dismissal, minimization, and behavioral inconsistency. The study found gaslighting to be significantly predictive of anxiety, depression, quiet quitting, and reduced work engagement among employees who experience it.

It does not always look like what you might expect. It does not require a villain. Some of the most sophisticated versions of it come from leaders who are charming, intelligent, and genuinely liked by their peers. What it looks like in practice is a team that is constantly second-guessing themselves. Talented people who suddenly seem uncertain. High performers who have stopped contributing in meetings. People who leave conversations feeling vaguely confused about what just happened, or who have started to doubt their own read on situations they understood clearly before.

Fear-based cultures sit in the same territory. The research is unambiguous: fear and psychological safety cannot coexist. When people are worried about the consequences of speaking up — whether that fear is rational or not, whether it comes from overt threat or just accumulated experience of how things tend to go — they protect themselves. And a team of self-protecting people is a team that cannot learn, cannot surface problems early, and cannot grow.

The cost is not just human. A 2025 narrative review of peer-reviewed literature from 2018 to 2025 — drawing on research across PsycINFO, MEDLINE, Scopus, and Web of Science — found that exposure to toxic leadership was significantly associated with anxiety, depression, burnout, and post-traumatic stress in employees, and that diminished psychological safety intensified every one of those outcomes.

None of this is fixed by a values poster on the wall or a mental health awareness week. It requires leadership accountability, a willingness to name what is actually happening, and, in many cases, external support for both the individuals affected and the leaders whose behavior — however unintentional — is driving it.

What Leaders Can Actually Do

Psychological safety is not a climate you create once. It is a standard you maintain daily through the choices you make in your smallest interactions.

Here is where to start.

Model the fallibility you want to see. Edmondson's research consistently finds that leaders who acknowledge their own uncertainty and mistakes create significantly safer environments than those who project confidence and certainty at all times. This is not about performing vulnerability — it is about being honest when you do not have the answer, or when you got something wrong. Your team will take their cues from you.

Change how you respond to bad news. When someone brings you a problem, the first 30 seconds of your response will teach your entire team whether honesty is safe. Thank them. Ask questions. Resist the urge to immediately assign blame or to reassure them with premature certainty. If your first instinct is to look for who caused it rather than what it means, you are training people to hide things from you.

Separate safety from standards. One of the most useful distinctions Edmondson draws is between a team that is psychologically safe and a team that is complacent. High psychological safety paired with low performance standards produces a comfortable team that does not improve. High psychological safety paired with high standards produces learning, growth, and resilience. Your job as a leader is to hold both, not to trade one for the other.

Make space for disagreement to be a normal part of decision-making. If every meeting ends in consensus, it is worth asking whether you are getting genuine agreement or social compliance. Not every team needs structured debate tools — sometimes it is as simple as asking "what am I missing here?" or "who sees this differently?" and then genuinely sitting with the answer.

Watch the patterns, not the policies. Your team will not read your Slack announcement about psychological safety. They will watch what happens to the person who brings up the awkward question in the senior leadership meeting. They will notice whether the colleague who disagreed with you last month got assigned to the interesting project or the thankless one. Psychological safety lives in patterns, not proclamations.

Be honest about what AI can and cannot do for your culture. If you are rolling out AI tools — and most organizations are right now — the way you bring your people along matters enormously. Transparency about what is being measured, what is changing, and why creates safety. Opacity destroys it. The leaders who get this right will find that their teams engage with the technology with curiosity rather than fear.

Back to Diane

When I challenged Diane on the silence in her meetings, it was uncomfortable for her. She had genuinely believed she was doing the right things.

We spent the next few sessions looking at the small patterns she had not noticed: the slight impatience when someone raised a concern late in a meeting, the way she would signal confidence in a direction before the room had really had a chance to think, the gap between her stated "open door" and the scheduling reality that made her almost impossible to reach.

None of it was intentional. Almost none of it was dramatic. But it was costing her.

The shift she made was more subtle and intentional than signing up to a leadership program or launching a culture initiative. It was a decision to get genuinely curious — in the moment, in the meeting, in the conversation — about what she was not hearing. To ask better questions. To slow down before she concluded. To notice the silences as information rather than agreement.

Her next 360 review told a different story. And so did her team's results.

Psychological safety is not a management trend. It is the foundation on which all meaningful leadership work rests. You cannot have honest feedback without it. You cannot have real accountability without it. You cannot build teams that learn, adapt, and grow through difficulty without it. And you absolutely cannot delegate it to a dashboard.

No AI system can create the conditions where someone feels safe raising a hard truth. No tool can replace the feeling of being genuinely heard by someone who has skin in the game. People work for people. They trust people. They follow people. That remains stubbornly, beautifully human — and it is the part of leadership that no amount of technological capability will ever make redundant.

You can only earn it. One interaction at a time.

---

Merve K. Hokamp is the founder of Leadrise Coaching & Consulting, a global executive coaching firm specializing in leadership development, team effectiveness, and organizational change. She has coached leaders across 37 nationalities, holds an MBA from INSEAD, and spent 11 years in senior roles at Google before founding Leadrise. She works with individuals and organizations seeking positive change.

If you are working on building more honest, high-trust cultures in your organization and want to talk through what that looks like in practice, get in touch.