The Death of Middle Management Is Greatly Exaggerated
The future will split middle management into two.
1Gartner says 20% of organizations will use AI to eliminate more than half of middle management positions by the end of this year. Every time I read a prediction like this, I think about what middle managers actually do all day.
They don’t spend their time on the things AI can automate — scheduling, reporting, performance tracking. Those are the tasks they do at 9 pm after a day spent doing the thing no one wants to talk about: absorbing ambiguity.
A CEO sets a direction. An individual contributor needs a task. Somebody has to convert one into the other. That somebody is a middle manager, and the conversion process is almost entirely illegible.
It involves understanding that when the CEO says “we need to move faster,” she means something different than the VP of Engineering does when he says the same thing. It means knowing which of your team members needs explicit direction and which needs room. It means having five definitions of success and knowing all of them are right.
———
I’ve watched this play out in content operations. You have a CEO who wants AI visibility proof. A CMO who wants messaging proof. A board that wants traffic growth. A team that wants to know if their work matters. Same company, same quarter, completely different scorecards.
No AI agent is going to sit in a room with those four stakeholders and navigate those competing truths. Not because the technology isn’t good enough. Because the job isn’t information processing. It’s meaning-making.
What will actually happen is something more interesting than mass elimination. The middle management layer will split into two types. The first type — those who primarily route information and enforce process — will indeed be compressed. AI is already better at this. It tracks tasks, surfaces blockers, and generates status updates. If your primary value is being a human router, you should be worried.
The second type — the ones who translate, interpret, and absorb contradiction — will become more important, not less2. Because as organizations adopt AI tools, the gap between “what the tool can do” and “what the organization needs done” will actually widen.
Somebody has to bridge that gap. Somebody has to figure out that the AI-generated analysis is technically correct but organizationally wrong, because it doesn’t account for the political dynamics of the last board meeting.
———
The real risk isn’t that companies will eliminate too many middle managers. It’s that they’ll eliminate the wrong ones. They’ll keep the process enforcers because their work is visible and measurable, and cut the ambiguity absorbers because their work is invisible and unmeasurable.
And then they’ll spend two years wondering why their AI transformation isn’t working, why strategy isn’t translating to execution, and why teams feel lost despite having better tools than ever.
The messy middle of an organization is called that for a reason. It’s where all the mess lives. You can automate a lot of things, but you can’t automate the willingness to sit in confusion until it turns into clarity. That’s a human function. It might be the most human function there is.
The Gartner prediction is doing a lot of rhetorical work by saying “more than half.” In practice, most organizations won’t eliminate a single middle manager in response to AI. They’ll just stop backfilling when someone leaves, and then wonder six months later why things are falling apart.
There's an irony in AI being used to flatten hierarchies: flatter organizations actually require more informal coordination, which tends to require more, not fewer, skilled translators between teams.