Executives are restarting office mandates and adding town halls while engagement slides and managers switch off. Global employee engagement fell to 20 percent in 2025, with Gallup estimating about 9 percent of global GDP lost to low productivity, roughly 10 trillion dollars (Gallup, 2025). The sharper alarm is managerial: engagement among managers has dropped nine points since 2022, including a fall from 27 percent to 22 percent in the last year alone (Gallup, 2025). Some organizations still manage 79 percent manager engagement, which suggests the gap is not fate but design (Gallup, 2025).
The signal behind the slump
Treat the engagement slide as a systems warning, not an attitude problem. The best evidence on psychological safety is clear about what it predicts and what creates it. In a meta-analysis of 136 independent samples, covering more than 22,000 people and nearly 5,000 groups, psychological safety is strongly associated with learning behaviors, information sharing, engagement, and task performance, with corrected correlations of 0.62, 0.52, 0.45, and 0.43 respectively, and only weakly with creativity at 0.13 (Frazier et al., 2017). Translation for managers: expect fewer avoidable errors and faster coordination before you expect breakthrough ideas.
Even more useful, the same review surfaces design antecedents. Autonomy, interdependence, role clarity, and supportive context all show sizable associations with psychological safety, with corrected correlations of 0.47, 0.86, 0.63, and 0.49 respectively (Frazier et al., 2017). One caveat matters. The very high interdependence effect at the individual level rests on a small number of samples, and the group-level estimate is more modest at 0.40 (Frazier et al., 2017). The practical read still holds. People speak up when their jobs are structured to require one another, when they know who does what, when they have room to act, and when asking for help is supported by the context.
The mechanism is learning, not cheerleading
The classic field study that defined team psychological safety showed why the correlations above show up in performance. In a single manufacturing firm with 51 work teams, psychological safety was linked to team learning behaviors, and those learning behaviors mediated the relationship to performance. Team efficacy by itself did not predict learning when psychological safety was accounted for (Edmondson, 1999). The study is cross-sectional and limited to one company, so treat causal language with care. Still, the mechanism is consistent and practical. Teams that can surface errors, seek feedback, and discuss problems without fear learn faster. Faster learning shows up as better performance.
If you want more initiative, do not tell people to be brave. Make it easier, safer, and more useful to raise issues in the flow of work. That is the leverage psychological safety offers.
Flattening raises the bar
Many firms are cutting managerial layers and granting more local authority. The theory is appealing and, for certain problems, correct. A peer-reviewed synthesis defines self-managing organizations as those that decentralize authority radically and eliminate the manager–subordinate relationship by design, formalizing decision rights across the system (Lee and Edmondson, 2017). The literature they review argues that traditional hierarchy is effective for reliable execution of known tasks in stable settings, but it tends to slow or block complex, cross-functional problem solving and can harden status differences that stunt development (Lee and Edmondson, 2017). This is a theoretical synthesis, not a randomized trial, so do not mistake aspiration for proof of outcomes. It does, however, clarify the managerial trade-off.
As authority pushes outward, the costs of silence rise. Without the protection and routines that create psychological safety, decentralization can turn into diffusion of responsibility. With it, cross-boundary learning and rapid error detection become normal. The design question is no longer whether people feel comfortable. It is whether the system requires and rewards the behaviors that make distributed decision making work.
Design moves for the next 90 days
None of the following are culture pep talks. They are structural prompts derived from what the evidence links to psychological safety and from how team learning drives outcomes.
- Make roles explicit and revisited. Publish a one-page map of who owns what and what good looks like. Update it monthly. Role clarity is a strong antecedent of psychological safety (Frazier et al., 2017).
- Create real interdependence. Set at least one shared outcome that crosses functions and can only be achieved jointly. Pair it with a cadence for joint problem solving. Interdependence correlates with psychological safety, with the caveat that individual-level estimates rest on a small number of studies and group-level effects are lower (Frazier et al., 2017).
- Increase local autonomy where risk is bounded. Replace blanket approvals with decision thresholds and peer review for routine calls. Autonomy is associated with higher psychological safety (Frazier et al., 2017).
- Normalize help-seeking with process. Add a standing agenda item called “stuck points” to team meetings and require at least one ask per session. In Edmondson’s field study, teams that treated help-seeking and error discussion as normal learned more and performed better through that learning pathway (Edmondson, 1999).
- Run blameless, specific debriefs. After incidents or sprints, examine process and assumptions first, not personalities. Capture two changes and assign owners. This makes the learning mechanism visible and repeatable, consistent with the mediating role of learning behaviors (Edmondson, 1999).
These moves are not silver bullets. They are scaffolding. Expect the first gains where the evidence points most strongly. You will likely see faster information flow, more proactive problem surfacing, and steadier execution before you see a flood of novel ideas. The creativity link is weaker in the data, which is a useful guardrail for expectations (Frazier et al., 2017).
The managerial bottom line
In a year when only one in five employees say they are engaged, and when managers themselves are disengaging, the burden of proof has shifted to design. Some firms already show it can be different, with dramatically higher manager engagement than the global average, which implies that context and systems matter more than slogans (Gallup, 2025). The evidence base does not promise that psychological safety fixes everything. It does show a reliable path to better learning, better coordination, and better task performance, and it specifies the job features that invite those behaviors (Frazier et al., 2017; Edmondson, 1999). If you are flattening spans or decentralizing decisions, the stakes are higher still. Build the conditions that make voice routine. Do that, and you give people a reason to speak up that does not depend on courage.
Sources
- Gallup (2025). State of the Global Workplace 2025. Reports global engagement at 20 percent, a roughly 10 trillion dollar productivity cost, and a nine-point drop in manager engagement since 2022, framing the design problem addressed here. https://www.gallup.com/workplace/659610/state-global-workplace.aspx
- Frazier, Fainshmidt, Klinger, Pezeshkan and Vracheva (2017). Psychological Safety: A Meta-Analytic Review and Extension. Finds strong links between psychological safety and learning, information sharing, engagement, and task performance, and identifies antecedents like autonomy, interdependence, role clarity, and supportive context, with caveats on small-sample estimates. https://doi.org/10.1111/peps.12183
- Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams. In a single-firm field study, team psychological safety predicted learning behaviors, which mediated the link to performance; team efficacy did not when controlling for safety. https://journals.sagepub.com/doi/10.2307/2666999
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Provides the canonical definition of self-managing organizations and argues, in a theoretical synthesis, that hierarchy suits reliable execution while hindering complex cross-boundary problem solving. https://static1.squarespace.com/static/5b3cd0cb1137a6d4c0c877c9/t/5c919b0afa0d60694eebc96b/1553046287400/Final+ROB.pdf