The Great Flattening: I Became a Manager Right as AI Started Deleting the Job

Gartner has a prediction floating around that should make every newly-minted manager set down their coffee: through 2026, 20% of organizations will use AI to flatten their org structure, eliminating more than half of current middle management positions.

I find this personally hilarious, because I just became a middle manager.

After 25 years as an IC and a Principal Tech Lead, I made the jump to Engineering Manager — and the ink was barely dry before the industry's biggest analyst firm started forecasting that my entire layer of the org chart is a rounding error waiting to happen. Welcome to management. Mind the trapdoor.

The Logic Isn't Stupid

Here's the logic Gartner is running. A huge chunk of what middle management actually does on a Tuesday is information plumbing: collecting status, aggregating it, reformatting it, and squirting it up or down the hierarchy. Scheduling. Reporting. Performance monitoring. Routing. The "span of control" — how many people one manager can oversee — was historically capped by how much of that plumbing a human brain could hold.

AI doesn't get tired of plumbing. So if a manager's span of control was eight because that's how many status updates a person could track, and AI can now track eighty, you don't need ten managers. You need one manager and a dashboard that writes itself.

That's the pitch: flatten the pyramid, widen the spans, automate the reporting, and let the survivors "focus on scalable, value-added activities." It's the kind of sentence that sounds great in a board deck and lands very differently when you're the layer being scaled away.

Engineers, Don't Cheer Yet

If you're an engineer reading this thinking "good, fewer managers, more shipping" — slow down. You're about to inherit the part of the job nobody automated.

Think about what middle management isn't: it isn't the status reports. It's the air cover when a project slips. It's the person who absorbs the political blast radius so you can keep your headphones on. It's the mentor who tells a junior dev their PR is fine and their panic is normal. It's the human who notices you've gone quiet for three weeks and asks if you're okay before you become a resignation letter.

AI can aggregate a standup. It cannot take a bullet for you in a steering committee. The "great flattening" tends to delete the cheap, visible parts of management (reporting) while quietly assuming the expensive, invisible parts (judgment, cover, growth, trust) will just… happen. They won't. They'll land on whoever's left, and that person will be running a span of twenty with no buffer.

It's Not Flattening. It's Barbelling.

I think "the great flattening" is real but mislabeled. It's not flattening — it's barbelling. The middle doesn't evaporate; it splits.

The administrative manager — the one whose value was reformatting information — is genuinely in trouble. That role was always a workaround for the fact that information didn't move well through an org. Fix the information layer and you don't need the workaround. Fair.

But the leadership function doesn't compress, it concentrates. The managers who survive won't be the ones with the tidiest Jira boards; they'll be the ones doing the stuff that has no API: developing people, making bets under ambiguity, holding a team together when the strategy changes for the fourth time in a quarter. AI eats the management tasks and leaves the management judgment — which means the bar for being a manager just went up, not away.

Here's my actual worry, and it's the one Gartner buries in the footnotes: flatten the org and you sever the career ladder. Where do senior engineers practice judgment before they're handed forty reports? The middle management layer was also the training ground for senior leadership. Delete the on-ramp and in five years you'll have a generation of execs who never learned to lead a team of eight, now leading eighty. That's not efficiency. That's a deferred outage with a really long incident timeline.

I took the management job because I wanted to multiply my impact through people, not despite them. If the role is getting hollowed out, I'd rather be the one deciding which half to keep than the one optimizing my own status reports until the dashboard makes me redundant. Pick the half AI can't do. Get very good at it. The plumbing was never the point.

Sources: By 2026, 20% of organizations will use AI to reduce 50% of middle management roles | Gartner · AI Is Eliminating Middle Management. Are Orgs Ready? | People Managing People · The Death of Middle Management Is Greatly Exaggerated | Suds & Som · The State of Engineering Management in 2026 | Jellyfish