Flattening Organizations With AI: How flat is too flat?

Let me start by conceding the argument, because when it comes to flattening organizations, there is not much of one to have.

Nobody has ever walked out of a meeting wishing for more managers. Nobody has ever looked at an approval chain and thought, you know what this needs, another signature. The instinct to flatten is not a fad and it is not wrong. Flatter organizations move information better, up and down. People can actually see and understand the whole business instead of their own narrow slice. Decisions move faster because there are fewer sign-offs standing between a good idea and a shipped one. If you have spent any time inside a tall organization, you have felt exactly why the flat one looks so appealing.

And this is not hypothetical anymore. Span of control, the average number of direct reports per manager, rose from 8.2 in 2013 to 12.1 in 2025, and is forecast to reach around 25 by 2028. Manager headcount is already down 6.1% between 2022 and 2025. Gartner expects one in five organizations to use AI to flatten their structures, eliminating more than half of existing middle-management roles.  This is consensus, not a contrarian bet. I wrote earlier this year about the most vivid version of the vision, Jack Dorsey’s “intelligence layer” at Block, and I will not repeat all of it here. You can read that one here.

So if the direction is that obvious, why is this a debate at all?

And almost nobody is being honest about what gets cut along with the layer.

What you lose when you flatten too far

Every layer you remove has a bill attached. Most organizations only read the first line of that bill, the savings. Here is the rest of it.

You lose the people who develop people. A manager was the one who noticed you were ready before you knew it yourself, argued for your raise in a room you were not in, handed you the assignment that stretched you, and opened the door to the next one. Thin that layer out and nobody is doing that work, because it was never on anyone’s job description in the first place. It just happened, quietly, because someone was close enough to see you. Worse, the managers who remain are now absorbing more, validating AI output, catching its errors, coaching their teams on AI skills, all with bigger teams and no added support. We did not free them. We buried them. Yes, every individual should own their own career, but let’s be honest, career growth is not a one-person job. It takes a village, and the person standing closest to someone on their way up is their manager. Which is why the saying goes that people leave their managers, not their company. The company is a notion. The manager and the team are real.

You lose leadership alignment.  In Korn Ferry’s 2025 workforce survey, 41% of employees said their company had cut management layers, and 43% said leadership alignment had suffered as a result. Fewer layers can mean faster, but it can also mean everyone rowing in slightly different directions with no one close enough to notice.

You lose the trust layer, exactly when you need it most. I wrote last month about the forever-layoff era, the constant drip of small cuts that keeps everyone anxious. The manager is the person who holds a team together through that churn. Remove them in the name of efficiency and you accelerate the very attrition you were trying to save money on.

You lose your future senior bench. Middle management was the training ground for senior leadership. It was where people learned to lead before the stakes were existential. Cut it and you get the same problem I wrote about last week with entry level, the collapse of the apprenticeship, except one level up. It is a leadership shortage you will feel in five years, not next quarter, which is precisely why it is so easy to cause today. You cannot have a manager appear out of nowhere when one leaves, which is why we build succession benches. When you flatten drastically, you also thin out the bench.

None of this means keep every layer. It means every layer carries a cost beyond the one that is praised, and the organizations that will win at flattening are the ones honest enough to count the whole bill before they cut.

The biggest thing you lose is the ladder

Of everything on that list, the most expensive loss is the least discussed, because the layers were the ladder. Pull out the layers and you have pulled out the path people climb.

To see why that matters, remember what a promotion actually was. For decades it bundled three things into a single moment: progression (a new title, a bigger job), influence (a wider remit, a seat at more tables), and pay. That bundle was how an organization said “you matter, keep going.” Flattening pulls the three apart, and almost nobody has decided how each gets delivered now. We removed the ladder and left the promise dangling.

The fashionable answer is that growth comes from the work itself now. Bigger problems, more scope, cooler projects. There is truth in it, but notice what it quietly drops. It keeps a version of progression and silently removes the other two. Great work with a frozen title, frozen pay, and frozen standing is not a career. It is a holding pattern with good PR, and people see through it fast.

Titles still mean something. They are not vanity. Inside the org, a title signals scope and standing: who gets invited to which room, whose opinion carries weight. Outside it, a title is the shorthand the market reads on someone’s next move. Some people leave for nothing more than a better one, and it is easy to roll our eyes at that until you remember the title is often the only portable, legible proof of growth a person can carry out the door. Stop handing them out because the ladder is gone, and do not act surprised when people go somewhere that still will.

Pay needs the biggest rethink. For decades the only real jump in compensation came with a promotion. Do well at the same level and your pay stagnated, or crept up at the margins. That was fine when a promotion every few years reset your number. But if promotions are becoming rarer because the rungs are gone, a pay model that only rewards level changes quietly tells people that growing in place is a financial dead end. This has to shift, and shift drastically, not just for the top performers we scramble to retain, but for the rest, the solid, steady people who now have nowhere to climb.

And we broke the ladder and cut the ladder-builders at the same time. The person who used to notice you were ready and go argue for you? Their span just doubled and their development budget got cut. We removed the ladder and thinned out the very people whose job was to help others climb what was left of it. You cannot solve a growth problem and a management-capacity problem with the same decision that created both.

The data already shows the gap. In one 2026 study, only 28.5% of employees said the skills they built with AI had shortened their time to a promotion or a pay change, and more than a third said those skills had not helped them advance at all. People are growing faster than ever. The system is just not set up to notice.

The questions we have to answer

I do not know how flat is too flat. Neither does anyone else, yet. But I know the three questions we have to answer before we can, and I think naming them matters more than pretending to have the number.

One. What replaces the promotion bundle? If “up” is mostly gone, how do we deliver progression, influence, and pay through lateral growth, concretely, not as a story we tell at review time? Comp for lateral moves? Titles that reflect scope rather than headcount? Recognition the system can actually see and reward? Until this is answered, every “we’ve grown your scope” conversation is a promise we have not funded.

Two. Where does development go when the developers are gone? Managers coached, sponsored, and grew people. If we cut the managers, who owns growth now? The organization? The individual? An AI coach? A smaller, dedicated cadre whose actual job is developing people? “Everyone owns their own growth” sounds empowering and usually means no one does. You cannot flatten responsibly until you have a real answer, not a slogan.

Three. How flat is too flat? This is the question we keep treating as new, and it is not. When the person nominally managing all the directs has no time left for actual information flow, for coaching, for judgment, they are not a manager anymore. They are a bottleneck with a title, and the whole thing quietly becomes a mess. We have run this experiment before. I have written about it, the Holacracies and self-managing teams we have been trying for decades, and the lesson keeps repeating: coordination does not disappear when you delete the coordinator. It just becomes everyone’s unassigned problem. The span math says the quiet part out loud. Twelve direct reports is a stretch. Twenty-five is a fantasy for anyone expected to genuinely develop the people under them.

And this is where AI enters, because AI is the reason all of this feels newly possible, and to be fair, it genuinely is more possible than it was. Agents can carry the routing, the reporting, the status-chasing that used to eat a manager’s entire week. That frees real capacity, and it can push the “too flat” line further out than it sat five years ago. But more possible is not the same as solved. AI can hold the information flow. It cannot yet own someone’s growth, argue for their raise, or model judgment under pressure the way a good manager does without being asked. So AI moves the line. It does not erase it. And treating it as magic is exactly how you flatten past the point of no return and call it strategy.

So here is where I land. The direction is right: fewer layers, yes, and without apology. But we need real answers to those three questions before we cut, not after. What replaces the promotion bundle, where development lives now, and the span at which a manager stops being able to manage. Answer those, and the right number of layers falls out of them almost on its own. Skip the questions and “we flattened” is not a strategy, it is a bet, and the bill for getting the depth wrong does not arrive for years, which is exactly why it is so easy to get wrong today.

Your turn

So I will ask you what I am still asking myself. In your organization, who owns someone’s growth now that the managers are stretched thin? And if you have flattened, how is it going? Tell us, for we would love to learn.

Tell me in the comments.