A founder I know flattened his 80-person company last year after reading one too many LinkedIn posts about "removing bureaucracy." Six months in, three of his best people had quit, decisions were taking longer than before, and he’d become the bottleneck for everything because nobody else had clear authority. He called me frustrated, and I had to tell him what I’d seen play out a dozen times before: the problem was never the hierarchy. It was bad hierarchy. And what he’d built to replace it was worse.
Flat organizational structure is having a moment — a big one. Gartner predicts that 20% of organizations will use AI to flatten their structures by 2026, eliminating more than half of current middle management positions. Meta, Alphabet, FedEx, and Bayer have already started cutting layers. The narrative sounds clean: fewer managers, faster decisions, lower costs, more autonomy.
This essay digs into what actually happens when that narrative meets reality. Drawing on case evidence from companies that tried and retreated, McKinsey’s research on organizational health, and patterns I’ve watched unfold across dozens of leadership teams, I’ll walk through why the flatness trend keeps producing unintended consequences — and what the organizations that get structure right are doing differently.
The seductive logic of going flat
The case for flat organizations is easy to make, which is partly why it keeps gaining converts. Fewer layers means fewer approval chains. Fewer managers means lower payroll. More autonomy means faster innovation. AI can handle reporting, scheduling, and performance monitoring that used to require human intermediaries. On a whiteboard, it all makes sense.
And in certain conditions, it does work. Small teams with experienced operators who share context and trust can absolutely thrive without formal management layers. Early-stage startups often function this way by necessity, and the energy is real — everyone knows what’s happening, decisions move fast, and ownership feels shared.
The problem is that these conditions are specific and fragile. They depend on team size (usually under 20), shared context (everyone understands the whole picture), high trust (built through close proximity and shared experience), and experienced people (who don’t need much coaching or direction). Remove any one of those conditions, and flatness starts producing the opposite of what it promises.
What actually happens when you remove the middle
The pattern is remarkably consistent. I’ve seen it in tech companies, professional services firms, and manufacturing businesses. The specifics differ, but the failure modes are the same.
Invisible hierarchies replace formal ones
When you eliminate formal authority structures, informal ones fill the vacuum almost immediately. The loudest voices start driving decisions. People with the most tenure or strongest relationships with the founder accumulate de facto power that’s harder to check than formal authority because nobody explicitly granted it and nobody can explicitly revoke it.
This is arguably worse than traditional hierarchy, because at least traditional hierarchy is visible. You can see who has authority, who made a decision, and who’s accountable when things go wrong. In a flat organization, power still concentrates — it just concentrates unpredictably and without accountability.
Decision-making gets slower, not faster
One of the biggest promises of flat structure is faster decisions. In practice, the opposite usually happens. Without clear decision rights — who decides what, with what input, on what timeline — every decision becomes a negotiation. People who aren’t sure whether they have authority to act default to consensus-seeking, which means more meetings, more discussion, and longer timelines.
The founder I mentioned earlier described it perfectly: "Before, my directors made 80% of decisions without involving me. Now everyone comes to me because nobody’s sure if they’re allowed to decide." He flattened the org chart and accidentally centralized all authority in himself.
The mentorship gap opens wide
Middle managers do something that neither AI tools nor flat-structure evangelists have adequately replaced: they develop people. A good middle manager notices when someone’s struggling before it becomes a performance issue, coaches junior team members through ambiguous situations, and translates organizational context into career development conversations.
McKinsey’s research on organizational performance consistently identifies talent development as one of the strongest predictors of sustained competitive advantage. When you remove the layer that does most of the developing, you save on payroll in the short term and erode capability in the long term. AI coaching tools can provide information and frameworks, but they can’t provide the judgment, relationship, and contextual awareness that good mentorship requires.
Remaining managers burn out
When organizations cut management layers without redesigning the work, the surviving managers absorb everything the eliminated layer was handling. Span of control balloons from five or six direct reports to twelve or fifteen. Coaching conversations get compressed. Strategic thinking gets crowded out by operational firefighting.
This is a predictable failure mode, and it’s happening at scale. The math simply doesn’t work when you double someone’s span of control without halving their other responsibilities. The remaining managers become overwhelmed, their effectiveness drops, and the best ones — the ones with options — start looking elsewhere. This is exactly the kind of dynamic that a strong high-performance mindset can’t compensate for, because it’s a structural problem masquerading as an individual one.
The graveyard of flat experiments
The case evidence is piling up, and it should give any leader pause before reaching for the flattening playbook.
GitHub abandoned its flat structure during a period of internal turmoil, ultimately concluding that the absence of management created more dysfunction than management ever had. Treehouse tried a no-boss experiment and its founder later called the approach "naive." Zappos offered a severance package to anyone unhappy with its holacracy experiment and lost nearly a third of its workforce. Bayer’s attempt to cut layers across 100,000 employees is still playing out, but early reports suggest the transition is producing exactly the kind of confusion and informal power struggles that the research predicts.
These aren’t obscure companies making rookie mistakes. These are well-resourced organizations with smart leaders who genuinely believed flatness would improve their operations. The pattern suggests the problem isn’t execution — it’s the premise.
When flat actually works
All of this said, flat structure does work in specific conditions, and it’s worth understanding what those conditions are so you can evaluate whether they apply to your situation.
Flat works when teams are small enough that everyone can maintain full context (roughly 8-15 people). It works when the work is highly creative or exploratory and benefits from maximum autonomy. It works when team members are experienced enough to self-manage and skilled enough to resolve conflicts without formal authority. And it works when the leader is willing to invest heavily in explicit decision-making frameworks, communication systems, and cultural norms that do the work that hierarchy would otherwise handle.
That last condition is the one most flat-org advocates underestimate. Removing hierarchy doesn’t eliminate the need for coordination — it just shifts the coordination burden from structure to culture. And building the cultural infrastructure to support flat operations is significantly harder than building a traditional management layer. It requires a genuine ownership mindset from every team member, not just the leaders.
The design principles that actually matter
The organizations that get structure right tend to focus on design rather than doctrine. They’re not ideologically committed to flat or hierarchical. They’re committed to whatever structure best serves their people and their mission, and they’re willing to adjust as conditions change.
Clarify decision rights before changing structure
Before removing any management layer, map every significant decision type in the organization and explicitly assign ownership: who decides, who provides input, who gets informed. This is unglamorous work, but it’s the single most important thing you can do to prevent the decision paralysis that plagues flat organizations. When everyone knows who owns which decisions, you can remove layers of approval without creating ambiguity.
Redesign the work before you redesign the org chart
Most flattening failures happen because leaders remove roles without removing or redistributing the work those roles performed. Before cutting a management layer, catalog every function that layer serves — not just supervision, but mentoring, context translation, conflict resolution, talent development, and upward communication. Then deliberately assign each function to a new owner, a new system, or explicitly decide it’s no longer needed.
Invest in coordination systems that replace hierarchy
If you reduce formal hierarchy, you need to increase the sophistication of your coordination systems. This means better meeting design, clearer communication protocols, more explicit delegation frameworks, and more intentional feedback loops. The organizations that successfully flatten are the ones that treat coordination as a design problem rather than assuming it will happen organically.
Preserve development pathways
Career growth is one of the top reasons people stay at organizations and one of the top reasons they leave. McKinsey’s data shows that 41% of people quit because of limited career growth opportunities. If you flatten your org chart, you need to create alternative development pathways — lateral moves, skill-based progression, project leadership — that give ambitious people a sense of forward motion even without a traditional promotion ladder.
The real question isn’t flat vs. hierarchical
The flat-vs-hierarchical debate is a false binary. The real question is whether your organizational structure is intentionally designed to serve your people and your mission, or whether it evolved accidentally through years of hiring and org chart inheritance.
Most organizations that reach for flatness are solving the right problem (too much bureaucracy, too many approval layers, too little autonomy) with the wrong tool. The answer usually isn’t removing structure. It’s redesigning structure — clarifying decision rights, expanding autonomy within clear boundaries, eliminating bureaucracy that doesn’t serve a purpose, and investing in the coaching capability of the managers you keep.
Structure is a tool. Like any tool, it works when it’s designed for the job and fails when it’s applied ideologically. The organizations that will thrive through this period of AI-driven restructuring aren’t the ones that flattened the most layers. They’re the ones that asked the hardest question: what does our organization actually need in order to make good decisions, develop great people, and execute at a high level? And then they built exactly that — no more, no less.