Block cut 4,000 people in February 2026 — 40% of its workforce. Jack Dorsey said it plainly: “A significantly smaller team, using the tools we’re building, can do more and do it better.” Block’s stock jumped 24% the day of the announcement.

The popular read is that AI is coming for coders. That is the wrong read. Coders are fine. The layer between coders and decisions is not.

What’s actually getting cut

Look at where the cuts are landing.

Block’s restructuring targeted coordination and management overhead. ASML, laying off 1,700 people in the same period, said its cuts were concentrated in leadership and coordination roles (project leads, program managers, scrum masters) while it planned to add engineering headcount. Meta built its new applied AI division with a 50-to-1 employee-to-manager ratio, double the 25-to-1 that has traditionally been treated as the outer limit before things break.

Of the 78,557 tech workers laid off in Q1 2026, nearly half (47.9%) were attributed to AI-driven restructuring. The roles disappearing fastest are the ones whose primary job is coordination: sprint planning, status reporting, dependency tracking, resource allocation. AI handles those now.

The engineering compensation data points the other way. Senior engineers with AI fluency are seeing median pay up 12–18% year-on-year at surviving companies. Junior starting offers fell 8–15% from the 2024 peak. The market is splitting, not contracting.

Why coordination overhead got bloated in the first place

Early-stage startups rarely have this problem. Five engineers, one of them runs standups. No layers.

Scaling creates the layers. Junior hires need direction, so you hire mid-levels. Mid-levels generate coordination overhead, so you hire managers. Managers need someone tracking cross-team dependencies, so you hire program managers. By the time you’re at 50 people, a meaningful slice of your headcount exists primarily to manage the complexity created by the headcount below it.

AI breaks the chain at the bottom. When a junior developer using Cursor and Claude produces output that previously required a mid-level engineer, you stop needing the management pyramid that existed to support junior developers. The DX Q1 2026 impact report found engineering managers are now shipping 4x as much code as six months prior. That is not a productivity stat. It is a structural signal. The manager layer is collapsing into the engineering layer.

The megamanager problem nobody is talking about

There is a real cost to removing middle management faster than the organisation adapts.

Fortune’s April 2026 analysis found that 75% of HR leaders believe managers are already overwhelmed by expanding responsibilities. The average American manager now oversees 12 direct reports, nearly double the figure from when Gallup started measuring in 2013. Global employee engagement has dropped to 21%, near a 15-year low, with managers showing some of the sharpest falls.

One in three HR leaders reports that rapid restructuring stripped their organisation of institutional knowledge they did not realise they had. Dorsey’s own rollout at Block was not clean: staff laid off in February were being re-hired within weeks after senior engineers pushed back on team dismantlement.

Getting this right means replacing the coordination function, not just removing the people who did it. Automated tooling. Better-defined interfaces between teams. Engineers capable of self-directing. That is a different kind of organisational work than a layoff.

What this means for early-stage hiring

Most early-stage founders are still hiring from the old model. VP of Engineering to “build process.” A few mid-level engineers to “get work done.” Linear headcount growth as the product scales.

That model does not work anymore.

If your senior engineers are 4x more productive with AI tools, you need fewer of them. If coordination overhead is being automated, you do not need the people whose job was coordination. What you need is engineers who can pick up a problem, scope it, ship it, and move on without much management overhead.

The calculus shifts toward senior, self-directed engineers. Not ten juniors and three mid-levels overseen by a VP. Maybe four senior engineers and a founder who can still read a diff.

Gartner found in 2024 that one in five businesses planned to use AI to streamline organisational layers. In 2026, those plans are executing. Startups have one genuine advantage here: you have not yet built the legacy structure you would need to unwind.

The question to ask before your next hire

Before posting a job description, ask: what coordination work does this role exist to do, and is that work being automated?

If the answer is mostly yes, you are about to hire overhead. If the answer is no, and this person will write code, design systems, or make product decisions, that hire makes sense.

Get this wrong and you will build a management pyramid, then spend eighteen months unwinding it while your leaner competitors ship circles around you.

If you’re thinking through your engineering org for the next twelve months and want a second opinion, talk to us.

Sources