Is middle management dead? What the data actually shows

Marcus was four months into his search, multiple final rounds in, still no offer, when I asked him a different question than usual. Not how the interviews went. Walk me through the last real judgement call you made for your team this month. Not a status update. Not something two levels up already decided. A call that was his.

He went quiet for a second. Then: "I just relay what my skip level already decided."

That sentence is the real middle management story, and almost nobody is telling it that way.

In March 2026, Atlassian cut 1,600 roles to fund its shift to AI. About 500 of those, close to a third, were in Australia. I still sit on the hiring side there. Within a day, the headlines had already decided what it meant: middle management is dying, and here was the proof.

Nobody checked whether the cuts actually hit management. Atlassian never released a breakdown of those cuts by level, department, or function. Not one number. The layoffs are real. The "middle management" framing sitting on top of them is a headline doing the reader's thinking for a company that left the question open on purpose.

What the harder evidence actually shows

Australia doesn't have a manager headcount study at this scale yet. The clearest number available is American. Live Data Technologies tracked 6.5 million white collar job changes in North America between 2018 and 2023. Layoffs at manager level or above made up almost half of every layoff in 2023, a jump of nearly 58 percent against the five year average before it. That's a real, measured shift. It isn't ours.

Then there's the stat everyone quotes and almost nobody reads properly: Gartner's prediction that a fifth of organisations will use AI to eliminate more than half their middle management positions by the end of 2026. It's a forecast about stated intent, not a body count. Most of the flattening so far isn't firing. It's a manager leaving and nobody backfilling the role. Which is a very expensive way of saying nobody actually knows yet.

Here's the part the North American number can't tell you, because it isn't measuring what I'm measuring. Across my coaching clients this year, the ICs land a role in one and a half to two and a half months, on average. The engineering managers and above take four months or more, chasing a much smaller pool of roles, and the roles that do exist get swarmed with applications. If middle management were simply vanishing, there would be nothing left to swarm. What I'm watching is slower and more specific than extinction: hiring managers running a longer filter over a narrower pool, because they're trying to tell two very different candidates apart, and both of them currently hold the same job title.

The split nobody's naming

Call it the Router/Owner Split, because that's what it is. Every management job sold under one title is actually two jobs stitched together. One half is routing: status reported upward, process enforced downward, the update, the check in, the "did you finish the form" chase. The other half is owning: reading that someone on your team is about to leave before they tell you, deciding what the roadmap actually means when two priorities collide, telling a senior stakeholder no and making it stick.

Software can already do most of the first half without a person in the loop. Nothing else in the org can do the second half at all. That's the whole split. Everything else is commentary.

So here's the claim, and it's a harder one than "middle management is changing." AI didn't come for management. It came for whichever half of the job could already be described in a Jira automation rule, and it turns out a lot of managers, plenty of good, hardworking people, were spending most of their week almost entirely inside that half. The layoff list making the rounds on LinkedIn isn't a trend piece about the fate of a job category. It's an audit result. Most of the people forwarding it with a nervous caption would fail the same audit if you ran it on their last month, this one included.

The companies that already cut a layer aren't celebrating either. Deloitte's 2025 Global Human Capital Trends survey, close to ten thousand people across 93 countries, found 36 percent of managers already in the job feel unprepared for it, and 40 percent say their mental health got worse after they took it on. Cut the layer above them and that strain doesn't disappear. It moves up, onto whoever's left holding the team. That's the Router/Owner Split scaled up to the org chart: flatten the layer and the routing work doesn't vanish, it lands on whoever didn't get cut, still unnamed, still unautomated, now just heavier.

The Last Call Test

Here's the test I ran on Marcus, and the one worth running on yourself before you run it on someone else's layoff list. Write down the last real judgement call you made for your team, a call that was yours, not routed down from above, not a status update mislabelled as a decision. If nothing from the past month comes to mind, you're not being modest. You're a router wearing an owner's title, and a Jira rule is already a cheaper version of you at that half of the job.

Pick one purely routing task this week and kill it before you defend it. A Jira automation rule can post your team's weekly status straight off the board, nobody typing it. An AI note taker, an Otter or a Granola, can turn your own standup into that same update without you writing it either. Either one closes the loop with your fingerprints nowhere near it. Spend what you get back on the half that's actually yours.

That's the gap between passing this audit on paper and passing it live. You can know exactly which calls were yours and still go blank the moment a panel asks, the way Marcus did with me, because rehearsing it in your own head isn't the same as saying it out loud under pressure, at the level you're interviewing for. That's what we built Interview Preparation in Careersy AI to close: the rehearsal itself, the same question that caught Marcus, run until the owning story is the one that comes out first, not the routing one.

Marcus didn't have a router problem so much as an owner problem he'd been mistaking for one, because that's the safer story to tell yourself. Looking at your calendar and counting the weeks since you made a call nobody handed you is a worse feeling than blaming AI for a market you can't control. Run the audit anyway, before someone else runs it on you and calls it a layoff.

FAQ

Does Atlassian's 2026 layoffs prove middle management is dying in Australia?

No. About 500 of the 1,600 roles Atlassian cut in March 2026 were in Australia. Atlassian never released a breakdown by level, department, or function, so there's no confirmed link to management roles specifically. It's evidence of AI-driven cuts hitting Australian tech, nothing more specific than that.

Is the Gartner stat about half of managers disappearing by 2026 real?

It's a real prediction from a real survey, but it measures what a fifth of surveyed organisations say they intend to do, not what has already happened. Most of the flattening actually observed comes from companies not replacing a manager who leaves, not from active firings.

Should I avoid becoming a manager because of this?

No. The part of the job that's shrinking is the routing half: status reporting, process enforcement. The part becoming more valuable, the owning half, absorbing ambiguity, making the calls nobody else can, is exactly what most managers already spend too little time on. Bet on that half, and run the Last Call Test on yourself monthly to check you actually are.