Here’s a headline you’ve probably nodded along to: a company announces AI is replacing middle managers, removing layers of bureaucracy, flattening its org chart, empowers individual contributors. It reads like progress. Fewer meetings. Faster decisions. Leaner cost base. Boards love it, CFOs love it, and most employees at least the ones who survive the cut say they love it too.
Now here’s the question almost nobody in that boardroom is asking out loud: if the rung where leaders are normally made disappears, where does the next generation of leaders come from? That’s not a hypothetical. It’s the actual, measurable trade companies are making right now, and the bill for it isn’t due for another decade, which is exactly why almost nobody is pricing it in.
The Numbers Behind the Flattening
The scale of this is bigger than any single company’s PR-friendly “efficiency” announcement suggests. According to Korn Ferry’s most recent CEO and Board Survey, an overwhelming 82% of boards and chief executives say they expect to cut up to 20% of their workforce over the next three years specifically because of AI and the cuts are not landing evenly across the org chart. Korn Ferry separately found that 41% of employees globally, and 44% in the US, say their organization has already eliminated entire layers of management.
The pattern shows up clearly in who actually loses their job when budgets tighten. A Bloomberg-commissioned analysis by Live Data Technologies found that middle-management roles accounted for nearly a third of all layoffs in 2023, up from one-fifth in 2018 and that manager-level-or-above roles made up almost half of all job cuts that year. The safest seat in the building used to be a few rungs up from entry level. It isn’t anymore.
Why AI Is Replacing Middle Managers Now: The Economics of Span of Control
Every org chart is, underneath the titles, a bet on span of control, how many people one manager can actually oversee well. For most of the last century, that number sat stubbornly around five to seven direct reports, because coordination, performance tracking, and day-to-day problem-solving all required a human brain doing the work.
AI breaks that constraint, at least partially. Status updates, scheduling, first-pass performance data, routine escalation triage, a meaningful chunk of what a manager used to do to coordinate a team can now be handled by a tool. So the math companies are running is simple: if a manager can credibly oversee twelve people instead of six, you don’t need the second manager at all.
Amazon’s restructuring is the clearest public example. Its self-described “anti-bureaucracy” cuts eliminated roughly 16,000 positions explicitly to flatten management layers, and reporting on the changes shows the company’s manager-to-individual-contributor ratio moving from roughly 1:6 to 1:10 or higher in multiple divisions. Salesforce cut close to 1,000 roles concentrated in marketing, product management, and its AI-focused Agentforce and Heroku teams. Block went further still, cutting 40% of its workforce, with Jack Dorsey pointing directly at AI’s expanding ability to take on tasks that used to require a person.
Gartner had effectively predicted this: through 2026, the firm projected that one in five organizations would use AI to flatten their structure, eliminating more than half of their existing middle management positions. That forecast is no longer a forecast.
What the 7S Framework Exposes About This Rush
Run this through McKinsey’s 7S model and the problem becomes obvious almost immediately. 7S argues that an organization is really seven interlocking elements, that is Strategy, Structure, Systems, Shared Values, Skills, Style, and Staff and that changing one without the others creates drag, not progress.
Right now, companies are aggressively rewriting Structure. Org charts are getting flatter by the quarter. But Staff development and Skills pipelines, the systems that actually grow a junior analyst into a future VP are barely being touched, let alone redesigned for a flatter world. You can’t change the shape of the org and leave the model for how people grow inside it, completely unchanged and expect no consequences. That’s not lean management. That’s deferred organizational debt, and 7S is precisely the framework built to catch exactly this kind of mismatch before it compounds.
The Leadership Pipeline Nobody’s Pricing In
Here’s the part that should worry shareholders more than it currently does. Middle management was never just a coordination layer, it was the unofficial training ground for almost every senior leader a company would ever need. It’s where a 29-year-old gets their first real P&L exposure, their first experience managing people instead of just tasks, their first taste of cross-functional conflict that doesn’t get resolved by simply being right.
Strip that rung out and you haven’t just cut a cost center. You’ve cut the apprenticeship system that quietly produced your next decade of directors, VPs, and eventually your next CEO. Korn Ferry’s own talent strategists have flagged this directly, warning that eliminating middle managers and entry-level roles looks attractive on a spreadsheet today while quietly destroying the leadership bench a company will need in ten or fifteen years.
Some companies are already groping toward a fix, a handful are floating a new “AI manager” role designed to give early-career hires a foothold in overseeing AI-driven workflows instead of overseeing people. It’s an interesting idea. It is also, currently, mostly theoretical, and it does not obviously teach the same skills that managing a team of humans through a hard quarter used to teach.
There’s a regulatory wrinkle layered on top of all this, too. Colorado’s AI Act, effective this June, and the EU AI Act both now require a “human in the loop” for AI-influenced employment decisions like hiring, promotion, and performance management. Companies are simultaneously stripping out the humans who used to occupy that loop and being told by regulators that the loop still needs a human in it. Nobody has fully reconciled that tension yet.
This Isn’t the First Flattening Wave But It Might Be the First Permanent One
Delayering isn’t new. The most famous wave hit in the 1980s and 1990s, most associated with Jack Welch’s GE, which stripped out management layers in the name of speed and accountability and largely succeeded, before headcounts crept back up over the following decade as the pendulum swung back. Another smaller wave followed the 2008 financial crisis, driven by survival-mode cost-cutting rather than any structural rethink, and it also partially reversed once growth returned.
What makes this wave different is the driver. The 1980s and 2008 waves were both, fundamentally, cyclical, tied to a specific competitive shock or a specific recession, and they eased once conditions normalized. This one is tied to AI capability that keeps compounding rather than fading. There’s no obvious “growth returns, headcount comes back” mechanism this time, because the technological reason for the flattening doesn’t go away when the economy improves. It gets stronger.
The Bill Comes Due Around 2040
None of this means flattening is wrong. Plenty of these org charts genuinely were bloated, and faster decision-making is a real, legitimate gain. But “this is efficient” and “this is free” are two different claims, and right now companies are only measuring the first one.
The cost of this flattening wave won’t show up in next quarter’s earnings call. It will show up in roughly twelve to fifteen years, when a board goes looking for its next CEO and discovers a thin, under-prepared bench of candidates who never got the chance to run anything smaller first. By then, the executives who made the flattening call will mostly be gone, and the company that has to live with the gap will have no easy way to undo a decade of skipped leadership development in a single succession cycle.
The Ground Floor Take
The efficiency case for flattening is real, and it isn’t going away. AI genuinely does remove the need for a layer of pure coordination work, and there’s no obvious mechanism this time that pulls headcounts back up the way they did after the 1980s wave or 2008. Companies chasing this aren’t wrong about the short-term math.
But here’s the part no earnings call will ever mention: the 26-year-old who would have spent three years managing a team of six, learning how to handle conflict, hit a number, clean up a mess they caused themselves, increasingly doesn’t get that job anymore. And no AI tool teaches a person how to lead other people. That’s only learned by doing it, on actual humans, with actual consequences attached.
The companies that come out ahead a decade from now won’t be the ones that flattened fastest. They’ll be the ones that flattened and simultaneously built some new mechanism like stretch assignments, micro-P&Ls, rotational leadership tracks, whatever it ends up being called to replace the training middle management used to provide for free.
The cuts are happening now. The system to replace what those cuts used to teach isn’t being built yet. That gap is this decade’s most underpriced corporate risk.
