What Companies Lose When They Cut Middle Managers
What do companies lose when they flatten their org structure?
When companies remove management layers, they lose the functions middle managers perform that never appear on an org chart: translating strategy into daily priorities, developing people, absorbing and resolving conflict, and carrying culture through change. AI can genuinely replace the coordination work of management, such as routing information and tracking decisions. It cannot replace the judgment, development, and human accountability that determine whether teams actually perform. Companies that flatten without a plan for those functions do not eliminate them. They redistribute them, unstaffed, to whoever is left.
The boldest org chart in corporate America
Jack Dorsey has proposed the flattest large company in modern memory. In an essay this spring co-authored with Sequoia's Roelof Botha, the Block CEO argued that AI tools and smaller teams can replace traditional middle management, and he has said publicly that he wants Block reduced from roughly five organizational layers beneath him to two or three by the end of this year. On a podcast, he described his ideal as all six thousand employees reporting directly to him. This followed Block cutting about four thousand roles, roughly 40% of its workforce, in February.
Block is not alone. Amazon cut fourteen thousand corporate positions with the stated aim of reducing bureaucracy and removing organizational layers, and its CEO has been explicit that well-intentioned middle managers who slow decisions are the target. Meta, Shopify, and others have run their own versions. The flattening movement now has its manifesto, its data platform, and its poster companies.
We want to engage with the argument seriously, because parts of it are right. And then we want to name what the argument consistently ignores, because we have spent decades inside organizations watching exactly the layer these companies are deleting, and we know what it actually does.
What the flatteners get right
The honest version first. Some of what has accumulated in management layers deserves to go. Information routing, status aggregation, decision logging, the meetings that exist to prepare for other meetings: this is coordination work, and AI is genuinely good at it. Dorsey and Botha's proposal for a persistent operational record that replaces managers as the conduit of information is not absurd. Much of it is overdue.
It is also true that layers can slow organizations badly, that some management roles exist because they always have rather than because they should, and that speed matters more than most incumbents want to admit. If flattening meant only removing coordination overhead, we would be for it.
But that is not what is being removed, because coordination was never most of the job.
The four functions that do not appear on the org chart
Here is what we observe inside client organizations, across industries, when we look at what capable middle managers actually spend themselves on.
Translation. Strategy does not execute itself. Someone converts "we are repositioning for enterprise" into what Tuesday looks like for a team of eight, decides what to stop doing, and answers the question every team asks of every announcement: what does this mean for us? AI can distribute the announcement. It cannot do the sensemaking, because sensemaking requires knowing the people, the history, and what this team heard the last three times leadership announced a new direction.
Development. People become capable through observed practice, timely feedback, and someone with standing in their career who cares whether they grow. Gallup's research attributes about 70% of the variance in team engagement to the manager, and our own coaching work shows why: development is relational or it is nothing. A company with no management layer has no apprenticeship structure, which means it is consuming capability it has no mechanism to replace. It will feel fine for four quarters and expensive forever after.
Conflict absorption. A meaningful fraction of management work is invisible precisely because the manager did it: the tension resolved before it became a dispute, the workload rebalanced before someone quit, the misunderstanding between two functions settled in a hallway conversation. Flatten the structure and none of this disappears. It lands on individual contributors, who now do it without authority, or on the executive, who now does it at a distance of six thousand direct reports. (We wrote more about this hidden layer in the role of mid-level leaders in culture and organizational performance.)
Human accountability. People do not perform for a world model. They perform for people whose judgment they respect and whose disappointment they would feel. That is not sentimentality; it is the entire behavioral engine of commitment. The essay's answer, that customer signal and revenue will substitute for managerial judgment, mistakes measurement for meaning. Metrics tell you what happened. A leader is how a team decides what to do about it.
The timing problem nobody is discussing
There is a bitter irony in the flattening movement's timing. It arrives at the exact moment the management layer is at its weakest in recorded history. Gallup's latest global report shows manager engagement fell nine points since 2022, the steepest decline of any employee group, and managers have lost the engagement premium they always held over their teams. We wrote about that collapse last week, and the connection matters: a decade of treating managers as overhead produced disengaged managers, and disengaged managers now serve as the evidence that the layer is not worth keeping.
That is not a discovery. It is a self-fulfilling prophecy completing its loop. An underinvested function performs poorly, and its poor performance justifies its elimination. Any organization could run the same play on any function it chose to starve.
The honest question for the executives watching Block is not whether your middle layer is performing. It is whether you ever gave it the conditions to. In Gallup's best-practice organizations, 79% of managers are engaged, nearly four times the global average, and those organizations are not flattening their way to it. They are developing their way to it.
What we would tell a leadership team considering this
If you take one thing from the practitioners' side of this debate, take this: automate the coordination, and double down on the judgment.
Use AI aggressively for the parts of management that were always administration in disguise. Then take the capacity you free and reinvest it in the parts that were always the actual job: fewer managers, perhaps, but genuinely developed ones, with reasonable spans, real coaching, and explicit accountability for translation, development, and team health. That is a defensible modern structure. Deleting the layer and hoping an intelligence platform absorbs the human functions is not a structure. It is an experiment with your culture as the collateral, and the results will arrive on a lag, after the people who understood what was lost have left.
Watch what happens at Block with genuine curiosity; we will. But when a leadership team asks us whether to follow, our answer comes from inside organizations rather than from an essay: the companies that outperform over decades are not the ones with the fewest layers. They are the ones whose layers are worth what they cost, because they invested in making them so.
If your organization is rethinking its structure, or trying to make a leaner one actually work, that is a conversation we have been having with leadership teams for a long time, and we would be glad to have it with yours. Reach out to start it.
Key Takeaways
Block's Jack Dorsey wants the company reduced from five layers to two or three this year, with AI replacing middle management functions; Amazon and others are running similar plays.
The flatteners are right about coordination work: AI genuinely can replace information routing, status tracking, and decision logging.
What flattening actually removes is translation, development, conflict absorption, and human accountability, and those functions get redistributed unstaffed, not eliminated.
The movement's timing compounds the problem: manager engagement is at its weakest on record after a decade of underinvestment, making the layer's poor performance a self-fulfilling justification.
The defensible modern structure automates coordination and reinvests in fewer, better-developed managers rather than deleting the layer.
Frequently Asked Questions
Is Jack Dorsey really eliminating middle management at Block?
Dorsey has stated he wants Block reduced from about five organizational layers to two or three by the end of 2026, and has described an ideal in which all six thousand employees report directly to him, with AI systems handling the coordination work managers historically performed. Block cut roughly 40% of its workforce in early 2026 as part of this restructuring.
Can AI actually replace middle managers?
AI can replace the coordination portion of management: routing information, tracking decisions, aggregating status, and surfacing context. It cannot replace translation of strategy into team priorities, development of people, conflict resolution, or the human accountability that drives performance. Those functions persist after flattening and land on whoever remains.
Do flatter organizations perform better?
Sometimes, when the layers removed were genuinely redundant and the remaining managers are strong. But manager quality drives roughly 70% of team engagement, so organizations that flatten by cutting capable managers typically see delayed costs in development, retention, and culture that do not appear in the first year's financials.
How should we strengthen our management layer instead of cutting it?
Automate administrative and coordination work, right-size spans of control, and invest in sustained development with coaching and peer practice rather than one-off training. Loeb Leadership works with organizations on exactly this through Managing for Impact and leadership coaching engagements.
Give your managers the tools to lead through what's next.
Managing for Impact develops front-line and mid-level leaders through live practice, peer cohorts, and coaching from senior practitioners.
Explore the program