Login

Psychological safety isn’t about being nice

Sustainable performance under pressure

Most HR directors I work with have, at some point, commissioned work on psychological safety. The survey gets run, the workshops happen, the team’s scores improve. Six months later, the executive team is still avoiding the difficult conversation they need to have. Or, worse, the team has become noticeably more pleasant to be in, and noticeably less honest. The CEO is still the only person in the room whose questions reliably land. The poor performer is still in post.

This is not a failure of the intervention. It is the intervention working precisely as designed, on a misreading of what psychological safety actually is.

Psychological safety is not comfort. It is not the absence of stress, the avoidance of conflict, or the warm feeling people get from being agreed with. It is the condition under which honest disagreement, smart risk and intelligent failure become possible. In a team that has it, more difficult things get said, not fewer. The concept has been so widely misused that its originator, Amy Edmondson, took the trouble to publish a corrective article in Harvard Business Review in 2025 setting out the misconceptions herself. UK organisations are accumulating the consequences of getting this wrong at the very moment the legal and economic environment is about to make those consequences sharper.

How can you tell if your team is actually safe, or just pleasant?

I do a lot of work with senior teams. What I have come to notice, in twenty years of this work, is that the gap between a team’s claimed safety and its actual safety shows up in the first ten minutes of any meeting I sit in, and almost never afterwards.

The signal is not in the words. It is in the rhythm. In a team with low psychological safety, contributions come fast, polished and aimed at agreement. People speak in order to confirm what they think has already been decided. The CEO speaks first or last, usually both. Disagreement is signalled obliquely, through silence or through follow-up emails the next morning. The meeting ends on time and nothing has been said that could not have been said in writing.

In a team with the real thing, the rhythm is different. There are pauses. People interrupt themselves. Someone says something half-formed and someone else builds on it. The CEO does not always speak first. A junior member contradicts a senior one, and the senior one’s face does not change. Decisions are arrived at, not announced.

A simpler diagnostic is what I call the corridor-room test. If colleagues consistently raise concerns to each other in the corridor that they do not raise in the meeting, the team is unsafe. It just looks polite.

A few years ago I was brought into a senior team at a UK financial services business where the CEO had recently introduced psychological safety as a board-level priority. The HR director had run a survey. The scores were respectable. The CEO believed he had a strong, candid team. In my first session with the executive group, I asked each of them, individually and confidentially, what topic they would not raise in the room. Six of the seven named the same one: a strategic decision the CEO had made the previous quarter that they all believed was eroding the business, and none of them had said so. The CEO heard this for the first time in front of the group, six months after the decision was taken. The survey had shown safety. The team had been demonstrating, in the only way the system permitted, that they did not have it.

What we worked on over the following months was not the team’s confidence. It was the CEO’s reactions, in real time, to bad news. The team did not need to be braver. They needed evidence that being brave was not going to be punished. That evidence is small, behavioural, and repeated. It takes longer to produce than any survey will admit, and once produced it has to keep being produced. The CEO described it, in our last session, as the hardest piece of leadership work he had done. It is also the work that changed the team’s actual decisions, not just its scores.

What I see most often, when a team’s safety is degraded, is a pattern the team has not noticed. The CEO interrupts in a particular way, or reacts to one specific kind of bad news with a particular tightness in the jaw, and over time the team has learned the precise topics on which the CEO is not, in fact, safe. The CEO is usually the last person in the team to know this. The team will not tell them, because telling them is the thing the team has learned not to do.

The Gestalt tradition I trained in has a useful vocabulary for this. The question in any room is not whether the team is safe, in the abstract; it is what the contact is like, right now, between these particular people. Contact is the live, fluctuating quality of how present and honest the encounter is. It can shift in a sentence. A leader who watches for contact, rather than for safety as a fixed state, has a chance of doing something about it in real time. A leader who is waiting for the survey results will be a year behind the data.

Why the stakes are rising in UK organisations now

The cost of getting this wrong is rising in the UK specifically, for three reasons.

The Employment Rights Act 2025 received Royal Assent in December 2025 and begins to take effect across 2026 and 2027. Its provisions include earlier unfair dismissal protection, reducing the qualifying period from two years to six months, day-one rights for statutory sick pay and flexible working requests, and stronger protections against fire-and-rehire practices. The CIPD’s Labour Market Outlook in early 2026 found 55% of UK employers expect workplace conflict to increase as a result. Whether or not you think the legislation is well-judged, the practical implication is the same: managers will need to be having more honest, earlier, better-documented conversations about performance, conduct and expectations, not fewer. A team culture trained in pleasantness will fail at this task.

The second is the squeeze on managers. CIPD and HSE data consistently show that UK managers are operating under sustained pressure. The HSE’s most recent statistics reported 17.1 million working days lost in 2022/23 to stress, depression or anxiety, the largest single cause of work-related ill health. CIPD’s Health and Wellbeing at Work surveys repeatedly identify heavy workloads as the leading cause of stress-related absence. People operating at capacity cannot do the work of safety. They default to avoidance, because avoidance is cheaper.

The third is AI. Generative AI is automating large parts of the analytical and informational work that used to fill leaders’ days. What remains is disproportionately the relational, judgement-based work. The questions a leadership team needs to ask each other in 2026 are harder, more ambiguous and more consequential. The team’s capacity to ask them well is now load-bearing in a way it was not five years ago.

The implication is uncomfortable for HR directors whose psychological safety work has produced the “nicer but less honest” outcome. That work is not neutral. In the current environment it is actively making the organisation less able to function.

What did Edmondson actually say about psychological safety?

The term is older than the current management fashion would suggest. Edgar Schein and Warren Bennis used it in the 1960s in the context of organisational learning. Amy Edmondson’s 1999 paper in Administrative Science Quarterly defined it as “a shared belief held by members of a team that the team is safe for interpersonal risk-taking”. The risks she had in mind were specific moments where someone might admit a mistake, ask what they fear is a stupid question, challenge a senior colleague’s reasoning, or surface an inconvenient fact. The question is whether the team is the kind of place where that move is survivable.

What gave the concept its corporate prominence was Google’s Project Aristotle. Between 2012 and 2015, Google studied around 180 of its own teams and found that what distinguished high-performing ones was not personalities, seniority, office or project, but the dynamics of how the team worked together, with psychological safety as the strongest single predictor. The finding was popularised by Charles Duhigg in the New York Times Magazine in 2016, and within a decade psychological safety had become a fixture in leadership development catalogues, employee surveys and consultancy decks.

Two things were lost in the translation.

The first is that Edmondson’s concept was always interpersonal, not individual. It describes a property of a team, not a personality trait of its members. You cannot fix it by sending the quieter ones on assertiveness training, any more than you can fix a leaking roof by drying the carpet.

The second is that it was always tied to learning and performance. Edmondson’s early work emerged from studies of medical teams where psychological safety predicted who reported errors and who did not, and therefore who learned and who did not. The point of the concept was never feeling good at work. It was: you cannot perform in any complex, uncertain environment if people are unable to say what they actually see.

Why does psychological safety training so often backfire?

In a May 2025 Harvard Business Review article, Edmondson and her co-author Michaela Kerrissey set out six misconceptions that have done most of the damage. Three of them sit underneath nearly every misfiring psychological safety intervention I see in UK organisations.

The first is that psychological safety means being nice. It does not. A team that has prioritised pleasantness over honesty has bought comfort at the cost of safety. The clue is what gets said in the corridor that does not get said in the room.

The second is that psychological safety lowers the bar on performance. It does the opposite. In safe teams, standards can actually be enforced because feedback is possible and corrections do not feel like an existential threat. Where safety is missing, performance management collapses into either avoidance, where no-one says anything, or hostility, where everyone says it all at once, in writing, to HR. Edmondson’s preferred framing, set out in her 2018 book The Fearless Organization, is high safety paired with high standards. The popular version produces low fear and low standards, which is a different and worse thing.

The third is that psychological safety is the leader’s gift to give. It is not, although leaders have outsized influence. Status differences, hierarchy, and the cumulative history of how the senior person has reacted to bad news shape the field much more than any single declaration that “this is a safe space”. I have lost count of the number of CEOs who have told their team that it is safe to challenge them, in a tone of voice and with a facial expression that announced precisely the opposite. Bodies read bodies. The verbal claim is almost irrelevant by comparison.

The other three misconceptions are worth reading in full, but they share a common structure. Each treats psychological safety as something the team either has or does not have, like a feature you can install. In reality it is a moving condition that is being produced and degraded in every meeting, every email, every reaction to someone making an honest attempt.

What changes when leaders take it seriously

A few shifts follow from this.

The first is that you stop trying to install psychological safety and start trying to notice when it is present and when it is not. This is a different kind of attention. It requires being in the room, not running a survey from outside it.

The second is that you accept the leader’s centrality. Not because the leader is the only source of safety, but because the leader’s reactions are the strongest single shaper of what the team will and will not say in their presence. This is not a moral observation. It is a description of how power works in human groups. The leader who wants their team to be safer has to start by paying close attention to their own reactions, and to what those reactions teach the team about what is and is not sayable. Most leaders find this much harder than running an offsite.

The third is that you separate safety from agreement. A team in which everyone agrees may be safe or may be terrified. You cannot tell from the surface. The diagnostic question is what happens when someone disagrees. If the disagreement is heard, examined, sometimes taken on, sometimes rejected with reasons, the team is safe. If the disagreer’s career trajectory subtly changes over the following months, the team is not.

The fourth is that you stop measuring psychological safety primarily through self-report surveys. Self-report surveys catch what people are willing to say about how safe they feel. In an unsafe team, people are not willing to say. The data is corrupted at source. Better signals are behavioural: who speaks in meetings and who does not, who challenges whom, how bad news travels (or does not), how long it takes for emerging problems to reach the senior team.

The fifth is that you build it through the small, repeated moments that nobody puts on a slide. The leader who thanks someone, on the day, for bringing them a piece of unwelcome information, and then acts on it, is producing safety. The leader who runs a Q&A session about psychological safety on Friday and then snaps at the person who flagged a budget risk on Monday is not. The body keeps the score, in the team as in the individual.

What’s harder than the research admits

A few things are worth being honest about.

Psychological safety has been over-claimed in some quarters. The Project Aristotle finding is real but should not be treated as the only thing that matters. McKinsey and others have found team effectiveness depends on a wider set of conditions, including clarity of role, alignment of goals, and the quality of the team’s information environment. A team that has psychological safety but no clarity is still going to underperform.

It is also harder than the literature sometimes suggests to disentangle safety from status. In hierarchical UK organisations, what looks like a safety problem may be a power problem. A junior person not speaking up may be making a rational calculation about their career, not failing a confidence test. The intervention required is structural, not interpersonal. Asking them to be braver puts the cost of the system’s design onto the least powerful person in the room.

And there is a contested edge in the research itself. Methodological reviews have noted that psychological safety has been measured in inconsistent ways across studies, and that the self-report instruments now in widespread use may not capture what Edmondson originally described (see Frazier et al., 2017, for a meta-analytic review). The research base is real and substantial, but it is not as settled as the consultancy market sometimes implies.

None of this invalidates the concept. It does mean leaders should be sceptical of any vendor offering them psychological safety as a packaged product with a guaranteed outcome.

What this leaves you with

Psychological safety is not the goal. It is the condition that lets the work happen.

Treating it as an end in itself produces teams that are pleasant, agreed and slow. Treating it as a condition for honest exchange produces teams that can disagree well, learn from their mistakes, and make decisions that hold up under pressure. The difference shows in the room, not on the survey.

In the UK, in 2026, with new employment legislation, a stretched managerial population, and AI removing the work that used to make leadership feel productive, the cost of the soft version is rising. Senior teams that can have the hard conversations will perform. Those that can’t, won’t. That is not a moral judgement. It is a description of the field.

The work, for leaders and for the HR and OD teams supporting them, is not to install safety. It is to keep paying attention to the conditions that produce it, and to act on what they notice. That is a different job. It does not finish.

Most of mine happens with senior leadership teams, in the rooms where the patterns actually show up.

References

Submit a Comment

Your email address will not be published. Required fields are marked *

Share This