In Defense of the Middle Manager
Key Takeaways
Prominent tech leaders argue AI can replace management layers, and companies are quickly flattening org charts as a result.
The argument treats managers as information routers. Routing can be automated; the trust managers create cannot.
Decades of Great Place to Work data show performance rising with the leader’s level: the manager is a multiplier, and the gap between the lowest and highest levels is measured in hundreds of percent.
Manager engagement is collapsing at the very moment spans of control are widening.
The case for the middle layer is a performance case backed by shareholder-return data, not nostalgia for hierarchy.
Middle management is having a hard year. Organizations across industries are removing layers from their org charts, and the most influential voices in technology are supplying the rationale: with AI in the picture, the argument goes, the coordinating work that managers do can finally be automated away. I understand the appeal. I also think the argument takes a narrow view of what managers actually do, and I want to make the case for the defense.
To be clear, this is a performance argument, not a sentimental one. The evidence I will walk through here suggests that the layer being cut is the layer that carries trust through an organization, and that the quality of that layer predicts productivity, innovation, and retention to a degree most executives would find surprising. Before deciding that software can replace these people, it is worth being certain about what they have been holding up.
The argument for eliminating the layer
The clearest statement of the case comes from Block CEO Jack Dorsey and venture investor Roelof Botha, who argue in an essay titled “From Hierarchy to Intelligence” that organizational hierarchy exists primarily to route information and pre-compute decisions, work that AI systems can now do faster and without the overhead. The essay arrived weeks after Block cut 40% of its workforce, so this is not an abstract thought experiment.
I want to be fair to the argument, because parts of it are right. Layers do slow information down. Bureaucracy is a real tax on speed, and anyone who has waited three weeks for a decision that needed three days knows it. The critique also lands on how the role actually functions in many organizations: in McKinsey’s research on middle managers, surveyed managers reported spending almost three-quarters of their time on work other than developing people. If that is what the job has become, it is no wonder it looks automatable.
But a role misused is different from a role unneeded, and this is the distinction the flattening argument leaves unaddressed. The fact that we have buried managers in administrative routing tells us something about how organizations assign work. It tells us very little about what the role is for.
Routing is not the full role of the manager
Strip away the status tracking, the scheduling, and the information relay, and what remains is the part of the job that’s actually the most important: translation and meaning-making, so that strategy becomes something a team can act on; psychological safety, so that people can raise problems while the problems are still small; and the daily relational work that determines whether people bring their full effort or the minimum. None of that is routing and it certainly can’t be automated.
The numbers bear this out. McKinsey finds that workplace relationships account for 39% of employees’ job satisfaction, and that the relationship with management accounts for 86% of satisfaction with those relationships. “People don’t leave bad jobs, they leave bad managers” is a cliché for a reason.
And the performance stakes are just as concrete: organizations with top-performing managers deliver multiple times the total shareholder returns of those with average or below-average managers over a five-year period.
The manager is also the organization’s early-warning system: the one person positioned to notice disengagement, confusion, or simmering struggles while it is still relatively easy and cost-effective to address. Remove that vantage point and problems do not disappear. They simply become apparent later, more ingrained, and extraordinarily difficult to change.
The leader is a performance indicator for the team
Here is the deeper reason I believe the flattening argument is a miscalculation. I have argued that trust is what animates workplace culture, engagement, and ultimately performance. And for most employees, trust in the organization is not experienced in the abstract. It is experienced through a manager. If you take out the layer through which people experience trust, you take out the mechanism itself, and the performance cost follows over time.
The research from Great Place to Work (GPTW) speaks directly to how much this varies with the person in the role. GPTW’s For All Leadership Model identifies five distinct leadership levels: from the Level 1 Unintentional Leader, unaware of their impact on others, up to the Level 5 For All Leader, who creates a consistently great experience for everyone on the team.
What makes the model more than a typology is what happens to performance at each step. Moving up the levels brings measurable improvement in productivity, innovation, retention, and organizational agility. Compared with teams under an Unintentional Leader, employees working for a For All Leader show 353% higher productivity, 325% greater readiness to innovate, 300% greater agility, and 128% greater desire to stay. Under a For All Leader, 93% of employees report a consistently positive experience of their workplace; under an Unintentional Leader, 28% do.
Even a single step matters. Moving from a Level 4 Good Leader to a Level 5 For All Leader still yields 21% higher productivity, 32% more innovation, 18% better retention, and 17% greater agility. Read that alongside the flattening debate and the conclusion I draw is this: the manager is a multiplier, and the highest-return move available to most organizations is developing that layer, not removing it.
Managers are under stress
None of this means the layer is healthy as it stands. Gallup’s State of the Global Workplace finds that manager engagement fell from 31% in 2022 to 22% in 2025, the steepest decline of any employee group and the primary driver of the global engagement drop. The capability picture is no better: in LifeLabs Learning’s 2026 mid-year trends research, 56% of HR professionals rated their managers’ coaching and feedback capability at 5 out of 10 or lower, and only 14% rated it highly.
Watch what happens after each round of cuts. Surviving managers inherit wider spans of control and less time per person, which degrades exactly the parts of the role that matter most, which in turn makes the next cut easier to justify. Underinvestment produces underperformance, and the underperformance is then presented as proof that the role was dispensable all along. It is a self-fulfilling case, and I see organizations building it against themselves right now.
Questions to ask before flattening your teams
If your organization is weighing a flatter structure, I would sit with four questions first:
What is required of managers — decision traffic, or relationship load? Only one of those can be handled by technology.
Who absorbs the trust work when the layer goes? Senior leaders rarely have the span. Peers rarely have the standing.
If the layer stays, what will you do with the capacity AI frees up? The best answer I know: move managers away from administrative routing and toward coaching, and build that skill deliberately. The For All data suggests every level gained pays for itself.
If restructuring proceeds anyway, how will you communicate it? Those choices, more than the org chart itself, determine whether trust survives the change. I laid out that playbook in Rebuilding Trust After Organizational Change.
FAQs
Can AI really do what middle managers do?
It can absorb the low-level coordination tasks, like status tracking, information routing, scheduling. It cannot hold a career conversation, notice someone struggling, or make it safe to disagree. Those were always the highest-value parts of the role.
Is flattening ever the right call?
Sometimes. Layers added for control rather than support can go. The test is determining the purpose of each layer: removing decision bottlenecks strengthens an organization; removing relationship capacity degrades it.
What happens to trust when layers are removed?
It concentrates in fewer, busier, less closely connected people. Wider spans mean less access, slower feedback, and weaker early-warning signals, unless the remaining managers are deliberately freed up for people work.
How do I strengthen the managers I keep?
Reduce their administrative load before adding expectations, train coaching and feedback as skills rather than assuming that they naturally exist, and measure them on team health. What gets rewarded is what gets done. The payoff is measurable: even moving a manager one leadership level up brings double-digit gains in productivity, innovation, and retention.
Closing Reflection
Every organization runs on a layer of people who translate strategy into daily work and carry trust in both directions. Before deciding that software can replace them, it is worth being certain about what they were holding up.
Explore More
• Rebuilding Trust After Organizational Change
• Managing Leadership Transitions
Download
Assess how your own leadership patterns support or strain the managers around you: download the Leadership Pattern Reflection Workbook.