Psychological safety is one of the best-known ideas in team research. Edmondson's 1999 study described it as a shared belief that a team is safe for interpersonal risk-taking.
Most leaders now know the definition. The harder question is what to do on Monday morning.
Three pieces of 2026 work add useful detail. None of them settles the question on its own, and two are small. Read together, they point to behaviours a team leader can practise.
Feedback and recognition matter everywhere
Mojarro-Durán, Rivera and Rodríguez-Aceves studied teams in three Mexican companies. They mapped how 54 people interacted, using social network analysis, and published the work in European Business Review in 2026.
Instead of asking people how safe they felt in general, they asked who they could bring up problems with. That makes the measure about real relationships.
Transactional behaviours, such as feedback and recognition, influenced psychological safety in all three companies. The effect was strongest in the smallest, youngest, project-based company.
Transformational behaviours, such as setting out a vision and role modelling, helped more in the mature companies with settled processes.
The sample is small and from one country. Even so, the message is practical: steady feedback and recognition may matter more than inspiring speeches, especially in new or fast-changing teams.
Five behaviours from experienced leaders
Resendiz and colleagues published a qualitative study in Frontiers in Psychology in February 2026. They interviewed eight senior leaders of research teams at one US university.
The behaviours those leaders described fell into five themes:
- self-awareness and growth
- mentorship and development
- amplification and empowerment
- communication and transparency
- accountability and integrity
Eight interviews cannot tell you how common these behaviours are or how much they help. They are useful as a checklist of what experienced leaders believe works, and the authors suggest building them into leadership training.
Safety and AI experimentation
Microsoft's 2026 Work Trend Index surveyed 20,000 AI users between February and April 2026. Only 13% said they are rewarded for reinventing work with AI even if results are not met, and 45% said it feels safer to focus on current goals than to redesign work with AI.
In a separate study of 1,800 workers, where managers created psychological safety for experimentation, readiness and value scores were up to 20 points higher. Those workers were 1.4 times as likely to be high-frequency users of agents.
Microsoft sells AI tools and the data are self-reported associations. Treat them as a signal, not as proof that safety causes AI use.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
Where the evidence is contested
The three sources do not fully agree on what drives safety. The Mexican study found that transformational behaviours helped in some contexts and not others. The interview study focuses on relational behaviours, and Microsoft's focuses on room to experiment.
They also measure different things. One maps relationships, one gathers stories, and one asks for self-ratings at scale. None is a controlled trial.
The fair conclusion is modest. Psychological safety looks like something leaders build through repeated, ordinary behaviour, and context changes which behaviours count most.
That has a practical consequence. A single leadership style will not suit every team. A new project team may need frequent feedback and visible recognition, while a settled team may respond more to a clear sense of direction.
Watch how your own team responds, and adjust. The research gives you options to test, not a script to follow.
What this means for team leaders
Turn the research into habits you can check each week.
- Give specific recognition. Name what someone did and why it helped. Do it often, not only at year end.
- Give feedback early and in private. Small, frequent feedback keeps problems discussable.
- Go first on mistakes. Name one of your own errors in a team meeting, including one involving AI.
- Protect experiments. Agree that a short AI experiment that fails is a result, not a failure.
- Ask who people can raise problems with. If the answer is nobody, that is your starting point.
The CIPD Good Work Index 2026 makes the same point from UK data. It recommends an open culture that allows for honest conversations, and it links leadership perceptions closely to business outcomes.
A four-week plan
Small habits are easier to start when they have a clear shape. This plan takes about ten minutes of extra effort each week.
- Week one: notice. Keep a note of every time someone raises a problem, asks a question or admits a mistake. Note how you responded.
- Week two: recognise. Give at least three pieces of specific recognition. Name the behaviour and its effect on the team.
- Week three: go first. Share one mistake of your own in a team meeting, and one thing you are unsure about.
- Week four: ask. Ask the team, anonymously if needed, who they could raise a tough issue with. Compare the answers with your week one notes.
Repeat the cycle and look for change over a quarter, not a week. Safety builds slowly and can be lost quickly.
Check where your team stands
Hidden AI use is often a sign that people do not feel safe to talk about it. The Shadow AI Exposure Check gives you a team-level picture in about four minutes.
Then agree shared rules with the AI Team Charter, including how the team handles an AI error without blame.
For the organisational side, see the Acceleration condition. For facilitated work, our psychological safety training and team events for 2026 and 2027 build these habits with the whole team.
Try this now
- Give one piece of specific recognition today, naming what the person did and why it helped.
- Name one of your own mistakes in your next team meeting.
- Run the Shadow AI Exposure Check to see whether AI use is hidden.
