Imagine noticing a mistake at work but hesitating to mention it. Perhaps you are worried about looking incompetent, upsetting your manager or being labelled difficult. That hesitation sits at the heart of psychological safety: whether people feel able to ask questions, raise concerns and admit mistakes without fear of embarrassment or punishment.

Leaders are often encouraged to create this kind of environment. The argument sounds straightforward: supportive leaders help people feel safe, and people who feel safe work more effectively. But how strong is the evidence behind that claim?

A new paper by our CBISS theme co-leads Geshwaree (Aartee) Huzooree and Melina Doargajudhur and colleagues, working with Corine de Ruiter and Vincent Dutot, examines this question. Published in The Leadership Quarterly, their review looks closely at what existing research can tell us about leadership, psychological safety and work outcomes.

The team reviewed 47 research articles published between 1990 and January 2026. Their aim was to assess whether the studies could establish cause and effect.

That distinction matters. If employees who describe their managers as supportive also report feeling safer and performing better, these things may be connected. However, this alone cannot establish that the manager’s behaviour caused the improvement. Other factors, such as working conditions, team relationships or previous success, could help explain the pattern.

The review found that none of the studies established the full causal chain from leader behaviour, through psychological safety, to better work outcomes. Some experiments provided partial evidence that changing leader behaviour could influence psychological safety. But they did not establish that psychological safety itself then caused better outcomes.

Another concern was how leadership was measured. In 82% of the review’s coded entries, researchers relied on employees’ ratings of their leaders. These ratings capture employees’ experiences, but they can mix what a leader actually does with how employees interpret that behaviour. This makes it difficult to identify which specific actions produce which effects.

So, what can we learn?

First, we should be careful about promises. The findings do not show that psychological safety is ineffective. They show that the research reviewed cannot yet confirm or refute the full explanation often used to support claims about improved performance.

Second, organisations need to be specific about leadership behaviour. “Be a supportive leader” is broad advice. Asking how a manager responds when someone raises a concern gives us something more concrete to examine. The review calls for research that separates observable actions from people’s judgements about those actions.

Third, leadership development needs thoughtful evaluation. A positive survey after a training session cannot, by itself, show that the programme improved performance or explain how any improvement happened. The authors recommend stronger experiments, clearer measures and studies in real workplaces to test each part of the proposed relationship.

For anyone managing people, the paper offers a useful question to ask when presented with a confident claim: what evidence shows that this particular action produces the promised result?

Through this work, our CBISS team encourages a more careful conversation about leadership and workplace improvement. Creating opportunities for people to speak openly deserves serious attention. Understanding how leaders can achieve that, and what follows from it, requires evidence we can trust.

Read the paper: A systematic and causal review of the link between leader-driven psychological safety and work outcomes.

The handover is complete. The files are saved. A replacement is on the way.

Then something goes wrong, and someone asks: “Who knows how we dealt with this last time?”

Everyone names the person who has just left.

It is a familiar workplace moment. It also reveals a weakness that a tidy handover document can easily hide. Organisations depend on knowledge that people rarely write down: which warning signs matter, when to escalate a problem, and who can bring the right people together when the usual process stalls.

A paper by Aartee Geshwaree Huzooree, our CBISS theme lead, published in Organizational Dynamics, explores what happens when organisations keep losing this experience while dealing with one disruption after another. It also considers how generative AI could help preserve some of it.

The starting point is simple. A crisis puts people under pressure. Some leave through exhaustion, retirement or restructuring. Their departure leaves the remaining team with less experience to draw on. If another crisis arrives before that experience has been rebuilt, the organisation starts from a weaker position.

The paper calls this a “knowledge depletion spiral”.

Replacing the person does not immediately repair the gap. A capable new colleague can learn the procedures quickly and still need years to develop the judgement their predecessor carried.

Think of an experienced project manager who notices that a supplier’s reassuring updates have become unusually vague. The report still looks fine. The deadlines have not yet slipped. But something feels familiar, and they know which questions to ask.

A handover might list the supplier’s contact details. It may never explain what made that manager pick up the phone.

That is the kind of experience the paper argues organisations should make a greater effort to preserve.

Its proposed approach starts by identifying where important knowledge rests with very few people. These may be senior specialists, but they may also be the colleague everyone calls when a problem crosses departmental boundaries.

The next step is a conversation about real events. Instead of asking someone to “write down everything you know”, ask them to revisit a difficult decision. What did they notice first? What would a newcomer have missed? Who did they contact, and why? When would the same approach be a mistake?

Generative AI could help a facilitator develop follow-up questions and organise the answers into short, usable records. The employee would check and approve those records, including the circumstances in which their advice applies.

The final step matters just as much: getting that experience back into use. A carefully maintained archive offers little protection if nobody consults it. The proposal is to bring relevant past experience to the team before its first major decision during a new crisis.

There are limits. An AI summary cannot capture everything a person knows. Some skills need to be learnt through observation and practice. Previous experience can also mislead when circumstances change, so the team still needs to question the advice.

Trust matters too. People need to know what is being collected and how it will be used. Participation in the knowledge-sharing conversations should be voluntary. A system intended to preserve experience could quickly lose its value if staff experience it as monitoring.

The paper presents a framework for organisations to test, rather than evidence that the full approach already works in practice. Its immediate value is the question it puts to leaders.

Before the next departure, ask whose experience the team would struggle most to lose. Then make time to understand how that person thinks through a difficult situation.

That conversation may reveal a gap no staffing spreadsheet can show.