Is Management Dead?

Every few months someone says management is dead, and now they have numbers. Companies are cutting layers. AI is taking over work managers used to do by hand. "Do we even need this role?" used to be a hot take. Now it's a budget question. I have a stake in the answer. I'm an engineering manager, still am, and I'm watching my own job change while I do it. So this isn't a guess from the outside. It's me trying to figure out what's happening to my own role.

I don't think management is dying. But one common version of it is, including part of mine, and pretending otherwise doesn't help anyone in the job. Here's where the pressure is coming from, why the human part survives, why "be the team's therapist" is the wrong bet, and what I'm doing about it.

What the data is actually saying

The trend is clear. Org charts are getting flatter. Spans of control are getting wider. The layer taking the hit is middle management, including engineering management. One projection has one in five companies cutting more than half their middle managers by the end of 2026. And a big share of recent tech layoffs gets blamed directly on AI, not the usual cost-cutting.

It helps to be clear about which work is at risk, because "management" is really several jobs in one. A middle manager was traditionally a conduit: take strategy from above, break it into tasks, collect status, report it back up. That conduit work, the status rollups, the "can you put together a deck on where we are," the translating between layers, is exactly what a model does cheaply and tirelessly. So when leaders look for a layer to cut, the one whose main output was moving information around is the obvious pick.

Engineering has a second twist, and I feel this one directly. As AI makes engineers more productive, the manager and senior-engineer roles are converging. Managers are expected to stay closer to the code. Senior engineers are expected to coordinate more. The pure people-manager who hasn't touched the system in three years is the most exposed person in the building. I feel it in my own week. More of my time is going back to the actual work: reviewing designs, using the agent tooling, having opinions about the system again. Less of it is the relay work I used to think was the job. This isn't a prediction. It's already showing up in who gets cut and who gets told to "go back to building."

So the honest read: the hands-off, coordination-heavy version of management is shrinking, and AI made cutting it easy to justify. I'm confident about the direction. I'm less sure about the size. The "death of the manager" headlines are way ahead of what I'd bet on, and ahead of what I see on my own team.

Sorting my own job

It's easy to nod along to all this in the abstract. It got real when I wrote down what I do in a week and sorted each item into two columns: what a model now does about as well as I did, and what's still clearly mine. Here's roughly how it came out.

The split is almost embarrassingly clean. Everything in the "AI does it" column is moving information from one place to another. Everything in the "still mine" column is judgment about people and priorities when the answer isn't clear. The week I thought of as "my job" is really two jobs under one title, and only one is at risk.

The more interesting part isn't what moved to the left column. It's that the same shift creates work that wasn't on my list two years ago, work I think is becoming the real point of the job.

That second table is the part I'd underline. The role isn't just losing its bottom layer. It's growing a new top layer, and that work is harder and higher-leverage than the coordination it replaces.

The part that doesn't automate

Here's the jump I want to avoid: seeing the coordination work disappear and assuming the whole role goes with it.

Teams never had managers just for logistics. People working on hard things together need someone whose job is the team: noticing when a strong person has quietly checked out, starting the conversation nobody wants to start, deciding what not to do so people aren't drowning, taking the ambiguity from above so the team can keep building. When I look at my calendar and ask what a model could have done, the answer is uncomfortable for the status updates and reassuring for the 1:1s. A model can write my status update. It can't sit across from an engineer who's about to quit and work out whether they need a new challenge, a real break, or just to feel heard for ten minutes.

The data backs this up more than you'd think. Employees who feel their mental health is supported are roughly twice as likely to report no burnout. Manager support measurably weakens the link between emotional labor and burnout. Even people predicting deep management cuts add the same caveat: algorithms can't provide psychological safety, and companies that cut too far hit a culture-and-mentorship gap they didn't plan for. The connective tissue of a team is load-bearing. You just don't see the load until it's gone.

So no, the human part isn't going anywhere. As the busywork burns off, it's a bigger share of what's left.

But a manager is not a therapist

It would be easy to stop there: machines take the spreadsheets, humans keep the feelings, everyone wins. It's a tempting story for me, because the human part is what I'm good at and enjoy. I still don't buy it, for two reasons.

First, I don't actually want to be a therapist, and most managers I know don't either. The guidance here is clear about the line: my job is to notice work patterns, start a supportive conversation, fix what I can control, and point people to real help, not to diagnose, not to counsel, not to absorb a whole team's emotional load. The times I drifted into being everyone's therapist, it didn't feel heroic. It felt like burnout coming: an amateur version of a job real professionals train years for, while my actual work slipped.

Second, and simpler: emotional support on its own isn't a defensible role. If being warm and available is my whole value, I've described a great colleague, not a job the company protects when budgets get tight. "Human connection survives" is true, but it's not a strategy. It's the floor, not the building.

So the version of me that lasts isn't the one that leans hardest into empathy. It's the one that keeps the human part and stays close to the actual work, because the human part only has leverage when it's tied to something the business can see.

Which way to grow

This is the part I'm acting on right now. Managers aren't disappearing, but the job is changing enough that standing still is a choice, and a bad one. The thing getting automated is the middle, the translation layer. To stop being the middle, push toward one of the two ends AI isn't good at: the technical or the customer.

More technical. Not "learn to code" from scratch, since I came up through engineering, but what people now call AI fluency, not tool fluency. Tool fluency is using AI to draft my docs. AI fluency is knowing which workflows to redesign, which risks need a human in the loop, what's actually hard versus what only looks hard, and being able to call BS when a plan or estimate doesn't add up. I don't need to out-engineer my engineers. I need to be technical enough that my opinion counts in a technical room. In practice, that's meant getting my own hands back on the agent tooling instead of hearing about it secondhand.

More customer-focused. The other way out of the middle is toward the people we build for. Be the person who knows the customer better than anyone: what they do, where they get stuck, which "obvious" feature would flop and which boring one they'd pay for. That knowledge turns a manager from a relay into a source of direction. It's as hard to automate as the technical end, for the same reason: it lives in messy, specific reality that isn't in the tracker.

You don't have to go all the way to either end, and the managers I respect most are moving toward both. I'm trying to. The one thing I can't do, that none of us can do, is sit in the middle running the conduit and assume warmth will save me. The middle is the part the machines are good at.

So, is management dead?

No. And I'd say that with more conviction now than a year ago, because I'm still doing the job. But the status-update, information-relay, never-touches-the-work version is dying, and AI is what made that obvious. What's left is fewer roles that ask more, not less: keep the human judgment a team really needs, drop the coordination a model can handle, and anchor to the technical work or the customer so your judgment has somewhere solid to stand.

For me it's simple. My job isn't going away, but the comfortable middle of it is, and I can feel which parts of my week fall on which side. So I'm picking an end and moving: back toward the technical work, closer to the customer, and keeping the human part that was always the point. The only version of this role that doesn't survive is the one that stays exactly where it is. I'd rather not be that one.