Return-to-office mandates, AI rollouts, and budget freezes share a hidden constant: pressure to centralize. Boards want coherence, regulators want accountability, and managers, already stretched, default to approval layers. Yet the numbers on engagement keep sliding. Gallup reports global engagement at 20 percent in 2025, with lost productivity estimated at 10 trillion dollars, or 9 percent of world GDP, and manager engagement down 9 points since 2022, including a steep fall from 27 to 22 percent in the last year (Gallup, 2025). If oversight were the cure, morale and performance would look better by now.
A more useful reading of the moment is that uncertainty raises the value of local knowledge. That calls for a different reflex: distribute decisions to the edge, then manage the system, not every choice.
The empirical case for pushing power out in rough seas
We have unusually strong evidence on what happens when firms decentralize before a shock. Using two large micro datasets, the World Management Survey across ten OECD countries and the US Census Bureau’s MOPS, Aghion, Bloom, Lucking, Sadun and Van Reenen study how firms that had delegated more authority from headquarters to local plant managers fared during the Great Recession (Aghion et al., 2021). Their quasi-experimental design compares firms within sectors and shows a clear pattern. In industries hit hardest by the downturn, firms that had pushed decisions down before the crisis outperformed those that had stayed centralized. The advantage is not a generic love letter to empowerment. It shows up particularly where turbulence is high, measured by product churn and stock market volatility, consistent with the idea that local information becomes more valuable when conditions shift quickly.
The authors probe alternative explanations and find the decentralization effect robust to concerns about coordination changes or managers’ fear of bankruptcy. The mechanism that fits the data is simple and practical. Local teams see the demand blips first, feel supplier constraints sooner, and can adapt without waiting for a chain of authorizations that is calibrated to yesterday’s market (Aghion et al., 2021). In calm waters, central playbooks can work. In storms, latency kills.
What distributed decision-making is, and what it is not
Pushing authority down does not mean managerial anarchy. Lee and Edmondson define self-managing organizations as those that decentralize authority in formal, systematic ways across the whole firm, eliminating the manager-subordinate relationship as the default structure (Lee and Edmondson, 2017). Their synthesis is careful on scope. Classic managerial hierarchies are effective for reliable execution of known tasks in stable environments. They become brittle when work is cross-functional, nonroutine, and cognitively demanding. Status gradients and approval bottlenecks can mute ideas from the edge, slow problem solving, and stunt development.
This is not a manifesto for abolishing bosses everywhere. It is a design choice about where expertise lives and how quickly the firm learns. The evidence base here is theoretical and case-derived rather than experimental, which should caution leaders against cargo-culting job titles or copying rituals out of context (Lee and Edmondson, 2017). The claim that survives scrutiny is narrower and more durable. When the work is complex and conditions are changing, distributing decisions to those with the best information can raise adaptability, provided the organization supplies guardrails that keep the whole coherent.
The human prerequisite: people must be safe to surface bad news
Authority can be delegated by chart, but judgment only flows when people speak up. Edmondson’s field study of 51 work teams in a manufacturing company introduced the construct of team psychological safety as a shared belief that the team is safe for interpersonal risk taking (Edmondson, 1999). In that study, psychological safety was associated with learning behaviors such as asking for help, speaking up about errors, and experimenting. Team efficacy by itself did not predict these behaviors once psychological safety was accounted for. Learning behavior in turn mediated the link to performance. The design was cross-sectional, confined to one company, and should not be read as causal proof across all settings. Its practical read is nonetheless direct. If local teams fear blame, distributing decisions simply shifts risk without unlocking better information.
In other words, decentralization without psychological safety is a bet on faster silence. The firm gets neither the speed nor the insight it hoped for.
How to distribute decisions now, without losing the plot
Leaders facing 2025’s volatility and manager fatigue can treat distributed decision-making as an operating system upgrade, not a slogan.
- Decide where variance is high. Use your own indicators of turbulence, such as frequent product changes, supplier churn, or volatile demand, and push more authority to the people closest to that volatility. This matches where decentralization paid off during the Great Recession, particularly in sectors with high product churn and market volatility (Aghion et al., 2021).
- Specify decision rights and escalation paths. Write down who decides what, what information they must consider, and when to escalate. Self-managing organizations that work do not erase structure, they replace role authority with process clarity designed for complex, nonroutine work (Lee and Edmondson, 2017).
- Invest in learning behaviors. Train leaders to frame work as learning, to invite input, and to respond productively to bad news. Create forums where teams can surface small failures quickly. Psychological safety is associated with the very learning behaviors that link local judgment to better performance in Edmondson’s field study, and it is a tractable managerial practice even if the original evidence was in one firm (Edmondson, 1999).
- Measure speed and quality, not proximity. Track decision cycle time, rework, and customer impact. In Aghion and colleagues’ analysis, the payoff of decentralization arrives through faster and better local adaptation under turbulence, not through costless consensus (Aghion et al., 2021).
- Watch the load on managers. Gallup’s survey puts global manager engagement at 22 percent, with the largest annual drop between 2024 and 2025, and shows best-practice organizations at 79 percent, which implies design matters more than fate (Gallup, 2025). Distributed decision-making is one way to redesign the manager job from bottleneck to coach. Do not claim that this move alone will fix engagement, since the Gallup data are survey correlations. Do recognize that concentrating every decision at the top is already failing many managers.
The test for any governance change is whether the next surprise gets handled faster and better. Centralized control can still make sense where uniformity is the point, such as regulatory responses or brand standards that do not benefit from local variation. Elsewhere, particularly in product, operations, and customer response, the current environment favors firms that can move authority without losing alignment.
The practical implication is simple. Treat the organization chart as a distribution network for judgment. Push decisions to where the information is richest, create psychological safety so that information is voiced, and keep the lattice connected through clear decision rights and rapid escalation. The next shock will not wait for a steering committee.
Sources
- Aghion, Bloom, Lucking, Sadun and Van Reenen (2021). Turbulence, Firm Decentralization, and Growth in Bad Times. Firms that delegated more authority before the Great Recession outperformed centralized peers in the hardest-hit sectors, consistent with the value of local information under turbulence. https://www.aeaweb.org/articles?id=10.1257%2Fapp.20180752
- Lee and Edmondson (2017). Self-managing organizations: Exploring the limits of less-hierarchical organizing. Defines SMOs and argues they fit complex, nonroutine work while hierarchies suit stable execution, clarifying when and how to decentralize. https://www.sciencedirect.com/science/article/abs/pii/S0191308517300059
- Edmondson (1999). Psychological Safety and Learning Behavior in Work Teams. In a field study of 51 teams in one manufacturing firm, psychological safety was associated with learning behaviors that mediated the link to performance, highlighting a key enabler of distributed decisions. https://journals.sagepub.com/doi/10.2307/2666999
- Gallup (2025). State of the Global Workplace 2025. Reports global engagement at 20 percent, a productivity cost of 10 trillion dollars or 9 percent of global GDP, and a sharp drop in manager engagement, underscoring the need for redesigned decision systems. https://www.gallup.com/workplace/659610/state-global-workplace.aspx