Companies Are Cutting Middle Managers to Fund AI. Something Load-Bearing Goes With Them.

Gartner is now predicting that one in five organizations will eliminate more than half their middle management positions by the end of 2026. Oracle has cut 20,000 to 30,000 roles, most of them management. Amazon laid off 14,000 middle managers. Cloudflare said the vast majority of its recent layoffs came from management, finance, legal, and internal audit. The justification is almost always the same: agentic AI can now handle the scheduling, reporting, and performance tracking that used to require a layer of people to run, so why keep paying for the layer?

Here is the number that should slow that logic down. Forrester’s 2026 Future of Work report found that 55 percent of employers already regret AI-related layoffs. More than half the companies that made the cut are already wishing they had not. Is cutting middle management for AI a mistake? The regret data says a lot of leaders are quietly answering yes, after the org chart has already changed.

Why are companies cutting middle managers to fund AI?

The pitch is efficiency. A manager’s job has always split into two kinds of work. There is the administrative half: scheduling, status reports, forecasting, tracking who is behind on what. AI is genuinely good at that half. It can summarize a project’s status faster than a human can compile it, flag a budget variance before a manager would have noticed, and draft the update that used to eat an hour of someone’s afternoon.

Then there is the other half of the job, the half that does not show up on an org chart line item: reading a report’s tone in a one-on-one and knowing something is off before performance drops. Translating a strategy decision made three levels up into a version their team can actually act on. Noticing the person who is quietly cracking under a workload two months before they quit. Vouching for someone in a room they were not invited to. None of that is administrative. All of it is invisible until the person doing it is gone.

The mistake in the flattening logic is treating those two halves as one job. Automate the first half, and a lot of organizations are assuming the second half either was not that valuable or will somehow keep happening on its own. It will not. It was never a task on a list. It was a person paying attention.

What actually disappears when you flatten the middle?

The people who report Forrester’s regret number are not describing an abstract loss. They are describing specific, measurable damage: slower decisions because nobody is translating strategy into daily priorities anymore, remaining managers stretched across double the headcount with no more hours in the day, and a visible drop in the kind of early-warning signal a good manager used to catch before it became a resignation letter or a compliance problem.

Fast Company’s reporting on the trend calls this “The Great Flattening,” and the pattern it describes is consistent: the administrative load genuinely shrinks, but the judgment load does not disappear, it just falls on whoever is left. A remaining manager who used to lead eight people and now leads twenty has not been handed an AI tool that replaces the six conversations they no longer have time for. They have been handed a bigger spreadsheet and less time to notice anything that is not already a number on it.

Is cutting middle management for AI a mistake?

The honest answer is: it depends entirely on which half of the job a company thinks it is cutting. Organizations that use AI to strip out the administrative drag and then reinvest the saved hours into the human half of management, coaching, one-on-ones, cross-team judgment calls, are not flattening anything that mattered. They are correcting a job description that had gotten bloated with busywork.

Organizations cutting the headcount and assuming the human half will simply keep happening at scale are the ones showing up in Forrester’s regret number. Newsweek’s reporting on the trend put it plainly: when AI cuts middle management, companies lose the layer that used to catch problems early, mentor the next generation of leaders, and hold the organization’s institutional memory. None of that was ever on the task list AI was hired to automate. It was the reason the role existed in the first place.

What should leaders do instead?

Start by separating the two halves of the job before making any cut. List what a manager’s calendar actually holds for a typical week, and sort it honestly into “AI can do this” and “a human has to be the one paying attention here.” Most leadership teams have never done this exercise, which is exactly why the flattening decisions get made on vibes and quarterly targets instead of a real accounting of what the role covers.

Second, if AI genuinely frees up hours in a manager’s week, protect that time for the human half of the job instead of using it to justify a bigger span of control. A manager freed from status reports should be spending that time in more one-on-ones, not managing twice as many people with the same number of them.

Third, invest in training the managers who remain, deliberately, before assuming they can absorb a flattened structure by instinct. The judgment work, reading a team, translating strategy, catching the quiet signs before they become an exit interview, is a skill that gets stronger with the right framework, not something every manager already has fully formed.

Where this leaves a leadership team right now

The Great Flattening is not going to reverse itself, and AI genuinely can strip a lot of dead weight out of a manager’s week. The mistake is not using AI. The mistake is assuming the invisible half of the job disappears along with the visible half, and then acting surprised when the regret numbers show up a year later.

Aziz Aghayev built The Spotlight Machine on exactly this tension: AI should make more room for the human judgment a leadership team actually needs, not quietly erase it while nobody is watching the org chart shrink. If your organization is weighing what to cut and what to protect as AI reshapes the management layer, that is the conversation a keynote from The Spotlight Machine is built to start.