How to build trust on a team

Trust and psychological safety are related but distinct. Trust is one-to-one: your read on whether a specific person will come through. Psychological safety is a group-level belief that the team is safe for taking interpersonal risks, like admitting a mistake. You build both through repeated leader behaviors, not a one-off pep talk.

Most guidance here stops at “create a safe space,” which is not an instruction anyone can follow. Safety is an outcome, produced by what people repeatedly do in small moments: how a manager reacts the first time someone admits a mistake, whether the quiet person on the call ever gets asked directly. This chapter is about those moments, and the trust they add up to.

Trust and psychological safety are not the same thing

Trust is dyadic. It runs between two people and points outward: will they do what they said, are they good at their part. Psychological safety is a property of the group, and it points inward: if I ask this question or admit I am lost, how will these people judge me. You can trust a colleague privately and still not feel safe raising a concern in the full team meeting.

Patrick Lencioni, whose model of team trust puts its absence at the base of the dysfunctions pyramid, draws a distinction worth stealing: predictive trust (I have worked with you long enough to anticipate what you will do) versus vulnerability-based trust (I can say “I was wrong” or “I need help” without it being stored up and used against me later). Predictive trust arrives with time. Vulnerability-based trust has to be built on purpose, and it is the kind teams run on.

Why building trust and safety matters

The case rests on two well-known bodies of research, and it pays to be precise about what each shows.

The first is Amy Edmondson’s. Her 1999 study of 51 work teams defined psychological safety as “a shared belief held by members of a team that the team is safe for interpersonal risk taking.” In that study, safety was associated with more of what she calls learning behavior (teams that asked questions and talked openly about their errors), which was in turn associated with better performance. This is correlational work, so the honest verb is “associated with,” not “causes.” Edmondson later brought the idea to a wide audience in The Fearless Organization and a 2014 TEDx talk, where she made the human mechanism plain:

“Don’t want to look ignorant? Don’t ask questions. Don’t want to look incompetent? Don’t admit weakness or mistake.” — Amy Edmondson, TEDxHGSE, 2014

The safe move is to stay quiet, which is exactly what costs teams the information they need.

The second is Google’s Project Aristotle, its internal study of why some teams outperformed others. Google found psychological safety to be the most important of the five dynamics it identified, and the foundation the other four rest on. This is observational research on Google’s own teams, so “Google found” is the right frame, not a law of physics. We cover all five keys in what is team dynamics. Two independent lines of evidence pointing the same way is a strong enough signal to act on.

How to build trust: the behavior list

Edmondson frames the leader’s job as a toolkit with three parts: set the stage, invite participation, and respond productively. Each part maps to a specific moment where trust is built or broken. None of it requires a personality transplant, only a handful of unglamorous things done consistently.

Set the stage

Framing does more work than most managers expect, and two moves matter most.

Name the work as a learning problem, not just an execution problem. When you say out loud that the work is uncertain and interdependent, that no one has done exactly this before and the parts depend on each other, you make questions and course-corrections read as the job rather than as incompetence.
Say that you need everyone’s input, and say why. “I do not have the full picture, and I will make worse calls if you sit on what you see” is a sentence that hands people explicit permission to speak. Say it, then behave as though you meant it.

Invite participation

Setting the stage is talk. Inviting participation is what you do once people are in the room.

Model fallibility first. A leader who says “I may miss things here, so tell me when I do” resets the odds for everyone else. Edmondson calls this situational humility, and it opens a room faster than any amount of encouragement.
Ask real questions, then be quiet. Proactive inquiry means asking specific questions you do not already know the answer to, and leaving enough silence that someone can fill it. “What are we missing?” dropped into a two-second pause is theater. The same question with a real ten-second pause is an invitation.
Build structures so speaking up is routine rather than heroic. Do not rely on courage; design the meeting so input is the default: round-robins where everyone speaks once, write-first prompts that capture views before the loudest voice sets the tone, anonymous retrospective entries, and pre-mortems. Structure carries the people who will never be first to raise a hand.

A short check-in question at the top of a meeting is a low-stakes version of the same idea: it gets every voice into the room once, before the stakes are high. Our list of check-in questions is built for that. To practice small disclosures as a group, you can run something like Two Truths and a Lie live with your team.

Respond productively

This is the part that decides everything, because people learn what is safe from what happens after they take a risk, not from what you announced you wanted.

Treat bad news as the most consequential moment you get. When someone brings you a problem early, your first visible reaction trains the whole team. Thank the person who flagged it before you react to the problem itself. The engineer who says “I think I shipped the bug” is doing exactly what you need more of.
Replace blame with curiosity. “Whose fault was this?” ends the flow of information. “What happened, and what can we learn?” keeps it open. Laura Delizonna, writing in Harvard Business Review, frames this as approaching conflict as a collaborator rather than an adversary, and suggests asking your team for feedback on how you deliver hard messages.
Destigmatize honest mistakes without dropping standards. Safety is not immunity from accountability. Edmondson is explicit that a preventable, blameworthy act still gets sanctioned; what you destigmatize is the intelligent failure, the well-reasoned attempt that did not work, and the honest “I do not know.” The goal is a team that is high on safety and high on standards at once, not a comfortable one that has quietly stopped pushing.

Responding productively includes closing the loop. If people surface problems and nothing visibly changes, they learn that speaking up is pointless, and the safety you built leaks away. Follow-through is where this shows: a team that keeps agreeing to changes that never ship is quietly teaching itself that candor is pointless. For a facilitator-focused version aimed at retrospectives, see how to build a psychologically safe retrospective.

How to measure whether your team feels safe

Safety is easy to assume and hard to see. Edmondson’s original study measured it with a seven-item survey answered on a scale, and it still holds up as a starting point. The reverse-scored items are marked, since a “yes” there is a bad sign.

If you make a mistake on this team, it is often held against you. (reverse-scored)
Members of this team are able to bring up problems and tough issues.
People on this team sometimes reject others for being different. (reverse-scored)
It is safe to take a risk on this team.
It is difficult to ask other members of this team for help. (reverse-scored)
No one on this team would deliberately act in a way that undermines my efforts.
Working with members of this team, my unique skills and talents are valued and utilized.

Two practical notes. First, run it anonymously: a person who feels unsafe is the least likely to admit it out loud, so a show-of-hands reading will flatter you. Second, watch the trend rather than a single number, since what matters is whether safety is climbing or sliding as you change your behavior. A recurring team health check does both. To read the signals off behavior instead, our guide to telling whether your team feels psychologically safe covers what to watch for. Written team agreements help too: making expectations explicit removes much of the guesswork that keeps people quiet.

Building trust when the team is remote

Everything above gets harder when the team is distributed, because the micro-moments that build trust in an office do not happen on their own. The hallway “quick question,” the read on someone’s face when a plan lands badly: remote, you have to engineer them.

Default to written and async, with explicit triggers for going live. Write-first input levels the field for quieter people and non-native English speakers who lose out in fast verbal meetings. Decide together which decisions are worth a synchronous call.
Keep camera-on optional. Mandating cameras reads as surveillance and taxes the people already most drained by video. Invite it and model it, and let people opt out without owing anyone a reason.
Over-communicate context. Co-located teams absorb the why behind a decision by proximity; distributed teams do not, so write the reasons down or people fill the vacuum with worst-case guesses.
Protect low-stakes human connection on purpose. The chat that happens for free in a kitchen has to be scheduled online: a standing check-in, a few minutes of non-work talk at the top of a call. It feels inefficient, and it is where a lot of trust is quietly made. Running something like Common Ground live gives a distributed team that moment.

The deeper version of this sits in remote and distributed team dynamics. The short form: distributed trust is built on cadence, not on a single offsite.

None of this is fast. Trust and safety are made in small, repeated moments and lost in one bad reaction to bad news, which is why the behavior list matters more than any workshop. Start with the moment you control most: what you do the next time someone tells you something you did not want to hear.

Frequently asked questions

What is the difference between trust and psychological safety?

Trust is dyadic: your read on whether a specific person will come through. Psychological safety is a group-level belief that the team is safe for taking interpersonal risks, like admitting a mistake or asking a naive question. You can trust a colleague one-to-one and still not feel safe speaking up in the full group. They reinforce each other, but they are different things.

How do you build trust in a team?

Through repeated behaviors, not a slogan. Amy Edmondson’s toolkit is a useful spine: set the stage by framing the work as a learning problem and saying you need everyone’s input; invite participation by modeling fallibility, asking real questions, and building structures like round-robins and anonymous input; and respond productively by thanking people who flag problems and replacing blame with curiosity.

How do you measure psychological safety on a team?

Amy Edmondson’s original seven-item survey is the standard starting point, asking whether mistakes are held against people, whether it is safe to take risks, and whether it is hard to ask for help. Run it anonymously, since people who feel unsafe are the least likely to say so out loud, and watch the trend over time rather than any single score. A recurring team health check does both.

Does psychological safety mean lowering standards?

No, and that is the most common misread. Safety is not immunity from accountability, and it is not about being nice. Edmondson’s point is that you destigmatize honest mistakes and the words “I do not know,” while still holding a high bar and sanctioning careless work. The strongest teams are high on safety and high on standards at once.

How do you build trust in a remote team?

You engineer the micro-moments that happen for free in an office. Default to written, async input so quieter and non-native English voices count equally; keep cameras optional rather than mandated; over-communicate the context behind decisions; and protect low-stakes human connection with a standing check-in question. Remote trust is built on cadence, not on a one-off offsite.