A COO I advised last spring slid a new org chart across the table and smiled. Four layers had become two. Eleven manager roles were gone, and she told me she had never felt closer to the work.
Seven weeks later, she learned that her second-largest customer was leaving. She didn’t hear it from her team. She read it in a LinkedIn post announcing the customer’s new vendor.
Nobody had hidden anything. The account manager had mentioned the frustration twice, once in a one-on-one and once in a hallway, to a director whose role no longer existed. That is the part of flattening middle management that almost nobody prices in. You remove cost, and you also remove the people who were quietly listening on your behalf.
What Is the Great Flattening?
The Great Flattening is the wave of companies cutting middle management layers, often justified by AI tools that automate reporting, scheduling, and performance tracking. Remaining managers take on wider teams. The stated goal is speed and lower cost. The unstated side effect is that executives lose the people who translated frontline reality upward.
Those numbers are not small. Gartner predicted that through 2026, 20% of organizations would use AI to flatten their structures and eliminate more than half of their current middle management positions. Meta, Amazon, and Google have all publicly trimmed layers.
So why are companies flattening middle management so aggressively? Partly because the math looks clean on a spreadsheet. Partly because AI dashboards make it feel as though the information those managers carried can now arrive by itself.
That second reason is the one I want to question.
Why Flattening Middle Management Feels Like Speed at First
For the first few weeks after a delayering, most executives feel sharper. Meetings shrink. Decisions move faster, and the noise drops.
I felt a version of this myself. A few years ago, I stopped using an editor on my ghostwriting projects and started working directly with every client. The first month was wonderful. Fewer handoffs, faster drafts, no one softening feedback before it reached me.
Then a long-standing client went quiet. When I finally asked why, she told me the last three pieces had felt “a little off,” and she’d assumed I was busy. My editor used to catch that tone. She’d notice a one-word reply and call the client before I ever knew there was a problem.
I hadn’t removed a layer of cost. I’d removed an early-warning system, and I didn’t know it existed until it went silent.
This is why the quiet after a flattening is so misleading. Fewer escalations reaching you does not mean fewer problems. It often means fewer people whose job includes noticing them.
Middle Managers Were Never Just Overhead
The common view is that middle managers pass information up and instructions down. If that were all they did, software could replace them tomorrow.
But the best ones do something harder to see. They interpret. They notice that a top performer has stopped volunteering ideas, or that a client’s emails have gotten shorter, and they decide which of those signals deserves a senior leader’s attention.
Gallup’s research has found that managers account for about 70% of the variance in team engagement. That figure usually gets quoted to argue for better manager training. I read it differently. It tells me managers are the ones holding the context that explains why a team is thriving or failing.
When you remove that layer, the context doesn’t move up to you. It stays at the front line, unspoken, or it leaves with the people who carried it.
A CEO I worked with described it well after his own restructuring. “I used to get the story,” he said. “Now I get the data.” The data told him attrition in one division had ticked up by two points. The story, which he pieced together four months later, was that a well-liked team lead had been passed over and half her team was quietly interviewing.
Is Flattening an Organization a Good Idea?
Flattening can work, and I don’t want to pretend otherwise. Some layers exist only because of historical accident, and removing them does speed decisions. Smaller companies often run flat structures well because everyone can still see everyone else.
The problem appears when the flattening is sized to a budget rather than to the work. Manager headcount at public companies fell by roughly 6% between 2022 and 2025, while the volume of coordination work didn’t shrink at all. Someone still has to notice the hesitation, resolve the conflict, and coach the struggling new hire.
In most delayered companies, that someone becomes you, or no one.
There is also a people cost that rarely shows up in the business case. Recent surveys suggest more than a third of employees in flattened companies say the loss of managers has left them feeling directionless. Directionless people don’t escalate problems. They wait, and then they leave.
So the honest answer depends on one question: did you redesign how information travels, or did you only redesign the org chart?
How Many Direct Reports Is Too Many?
For most knowledge work, more than 12 to 15 direct reports is too many for a manager to know each person’s work, pressures, and trajectory. Gallup data suggests manager engagement peaks at around eight or nine reports and declines beyond that, while team engagement is highest at smaller spans.
Yet the average span keeps widening. Gallup found the typical manager now oversees 12.1 people, up from 10.9 a year earlier and nearly 50% higher than when it first measured in 2013.
At twelve people, a weekly one-on-one of thirty minutes consumes six hours. Add preparation, follow-up, and the conversations that don’t fit a calendar slot, and you’ve spent most of a working day just keeping up. Nothing is left for noticing.
I watched this happen to a sales director I coached. She went from seven reports to nineteen after her company flattened. Within two months, her one-on-ones had shrunk to fifteen minutes and turned into status updates. She told me, a little embarrassed, that she no longer knew which of her people were struggling. She only knew whose numbers were down.
Numbers tell you where a problem has already landed. People tell you where one is forming.
The Leadership Pipeline Nobody Budgeted For
There is a second loss, and it shows up years later. Middle management roles were where most future executives learned to lead. They were the place to make small mistakes with a team of six before making large ones with a team of six hundred.
Gartner flagged this risk alongside its flattening prediction, noting that removing these roles disrupts mentoring pathways and leaves junior employees with fewer places to develop. I’d go further. Middle manager layoffs don’t only cut today’s managers; they cut the practice field for tomorrow’s senior leaders.
A founder I worked with discovered this when she tried to promote from within for a new VP role. She had strong individual performers everywhere. But after two rounds of delayering, not one of them had ever led a team through a bad quarter. She hired externally, paid a premium, and lost eight months while the new VP learned the culture.
So when you model the savings from a flatter organizational structure, add a line for the leaders you won’t have in five years. It rarely appears in the business case, and it is often the most expensive line of all.
Where AI Helps, and Where It Goes Deaf
I use AI tools every day, and they have made parts of my work far faster. Summaries of long client calls, first-pass research, draft outlines: all of that now takes minutes instead of hours.
That said, AI has a specific blind spot that matters here. It summarizes what gets said and written. It cannot report what people chose not to say.
The account manager at my COO client never wrote “this customer is about to leave” in any system. She said it with a sigh in a hallway. No dashboard caught it, because it never became data.
This is the trap in using AI to justify flattening middle management. The tools are strong at the parts of a manager’s job that were already visible: reporting, tracking, scheduling. They are weak at the invisible parts: reading hesitation, sensing that a meeting went badly even though no one objected, knowing which silence is dangerous.
So when you cut the layer and rely on the tools, you keep the visible work and lose the invisible work. And the invisible work was where most of your early warnings came from.
How to Rebuild Your Hearing After You Flatten
If your company has already delayered, the answer is rarely to rebuild the old hierarchy. Instead, rebuild the listening that the hierarchy used to do by accident. Here is the approach I now walk leadership teams through.
1. Keep a surprise log for ninety days. Every time you learn something important later than you should have, write it down, along with who knew first. After three months, patterns appear. You’ll see which parts of the business have gone quiet and which roles used to carry that signal.
2. Name your listening posts. Pick three to five people close to customers and frontline teams, not by title but by judgment. Meet each one monthly with a single question: “What are people worried about that isn’t showing up in the numbers?” Then protect those meetings the way you’d protect a board call.
Designing Translation Into a Flat Structure
The first two steps show you where hearing has gone. The next three rebuild it.
3. Write translation into the job. In a flat structure, translating frontline reality upward is nobody’s formal role, so it becomes nobody’s work. Add it explicitly to senior individual contributor roles and reward it in reviews. One founder I advised gave her principal engineers a standing fifteen minutes in the monthly leadership meeting, with no slides allowed. That slot surfaced two product risks in its first quarter.
4. Set a span ceiling by type of work. Routine, well-defined work can handle wider spans. Complex, ambiguous, or client-facing work cannot. A single company-wide ratio treats both the same and quietly starves the teams where early warnings matter most.
5. Let AI handle the visible layer, deliberately. Use the tools to absorb reporting and status tracking, then give the time you’ve freed back to managers for conversation. If the hours saved simply become more direct reports, you’ve automated the paperwork and eliminated the listening.
None of this is complicated. It asks you to treat information flow as something you design, rather than something you assume will keep happening after the people who carried it are gone.
The Question Worth Asking Before the Next Restructure
The COO I mentioned at the start didn’t reverse her restructuring. She did something smaller and, I think, wiser. She went back to the account manager and asked her directly: “What else do you know that I don’t?”
The answer took forty minutes. Two more accounts were wobbling. One product feature was confusing nearly every new customer. None of it had shown up anywhere she could see.
Flattening changed the chart. It didn’t change the fact that the people closest to the work still know things you don’t. So before your next round of cuts, look at your own org chart and ask one thing: if something started going wrong at the front line tomorrow, who would tell you, and how long would it take?
If you’re working out which parts of leadership to hand to AI and which to keep firmly in human hands, my book The Practical AI Playbook goes further, particularly on building simple workflows that free time for judgment instead of consuming it. Available on Gumroad: https://5402694886686.gumroad.com/l/pqzjdf