And why it isn't cruelty, but biology.
Pavel Řehák did it. Meta did it. Google did it. ProductBoard did it. 30% has become the number hanging in the air of every boardroom discussion about AI strategy, whether spoken out loud or not.
The narrative going around goes something like this: managers, scared or emboldened by AI, are cutting into living flesh before machines even get the chance to take over the work. Bloodlust disguised as vision. Cynicism dressed up as strategy.
There's a grain of truth in that. But this article will try to think through the other claim.
Because there's another explanation — deeper, less comfortable, and far more important. Mass layoffs aren't just a consequence of AI transformation. They're also its precondition.
The problem nobody names: the cost of coordination
For more than two and a half thousand years, one rule held true, whether in Roman legions, Napoleon's armies, or today's corporations: one commander can effectively manage roughly eight people. The basic tactical unit of a Roman legion, the contubernium, consisted of exactly eight soldiers. British military strategist Sir Ian Hamilton formalized this in 1922 into the concept of “span of control,” and French consultant V. A. Graicunas added mathematical proof a decade later: the number of relationships one person has to track doesn't grow linearly with each additional direct report — it grows geometrically.
That's why middle management came into being. Not as a layer of power, but as coordination infrastructure. A manager was essentially a human router: receiving information from below, filtering it, translating it, and sending it up — and vice versa. Every layer of hierarchy existed because one person simply can't maintain a meaningful working relationship with an unlimited number of people.
AI rewrites this equation — not because people can suddenly handle more relationships, but because a machine takes over the coordination, translation, and synchronization. Take a team of ten people. Between them, there are 45 mutual connections, everyone with everyone. Add ten more, and suddenly you have 190 connections. Add another twenty, and you have 780.
This isn't linear growth. It's quadratic. The formula N × (N−1) / 2 describes how many channels information has to cross before a company can decide and act. For two and a half thousand years, the answer to this problem was another person in the middle. Today, the answer is a tool that manages those channels on its own.
A large organization doesn't get stuck on strategy. It doesn't get stuck on vision. It gets stuck on the fact that, by the time a vision travels from the boardroom down to the fortieth layer of implementation, the world has already moved on. A competitor with ten people and no approval processes is long gone ahead.
Layoffs aren't about money. They're about the speed of the loop — from observation, through interpretation, to decision and action. And that loop gets longer and more expensive with every additional person in the company, because the system was built from the start on the assumption that people handle coordination. If AI handles it instead, part of that structure simply stops being necessary.
The bystander effect: why big companies don't act
The quadratic growth in connections explains why decisions are slow. But it doesn't explain why they sometimes don't happen at all.
For that, we need a different phenomenon: the bystander effect. First documented after the murder of Kitty Genovese in 1964, when dozens of neighbors watched an attack lasting more than half an hour from their windows, and no one called the police. Everyone assumed someone else would do it. The more people who saw the situation, the more responsibility diffused, until it disappeared entirely.
The same thing happens in every large company facing an inevitable AI transformation.
The problem is visible to everyone. Everyone sees it, everyone feels it, everyone talks about it over coffee. But in an organization of five hundred people, everyone simultaneously assumes that responsibility for the decision belongs to someone else — the right people, with the right agenda, in the right meeting. Responsibility dilutes exactly in proportion to how many people are present. In a boardroom of twenty people, everyone still feels it on their own shoulders. In an organization of five hundred, it's diluted to the point of becoming invisible.
That's why transformation at corporations so rarely starts from the inside. It isn't triggered by collective will — it's triggered either by external pressure (a crisis, a market downturn, an investor), or by a small group that manages to seize responsibility before the group dilutes it back away.
So layoffs don't just solve the coordination math. They also solve the coordination psychology. A company of three hundred people, after a strategic cut, isn't just smaller. It's a company where everyone once again feels like the first responder, not a bystander.
The founder effect: why evolution hates big herds
In evolutionary biology, there's a phenomenon called the founder effect. It happens when a small group splits off from a larger population — through migration, catastrophe, or isolation. The gene pool narrows dramatically. And that's exactly the moment evolution speeds up.
Why? Because any new variant, mutation, deviation, or innovation spreads through the whole group exponentially faster. In a population of fifty individuals, one carrier of a new trait quickly influences everyone. In a population of five thousand, that same impulse gets diluted into the noise of the average.
That's why the fossil record contains practically no transitional species — that famous snake with tiny vestigial legs. Evolution in a small population happens fast, and that moment is too brief to leave a trace. Transitional forms don't last long, because small groups either evolve or go extinct. The middle ground doesn't hold.
Now picture a company after a 30% layoff. A smaller population emerges, and within it, if the cut was done right, “carriers of the new genetic information”: people who operate AI-first, who think in terms of automation, who don't need five meetings to make one decision.
In an organization of a thousand people, these people are statistical noise. In a company of three hundred, they can rewrite the culture of the entire organization before the company grows again. You have to write the new culture into the DNA before you add people back — otherwise you end up back in a large, interbreeding population, and the founder effect gets lost.
The transitional form is the most dangerous position
The snake with legs was doomed to extinction. It wasn't as fast as a mammal, nor slithery enough to be a proper snake. It was inefficient in both worlds.
A company trying to be AI-first without a structural cut is exactly this creature. Too slow for purely AI-native players. Too expensive for legacy players, who at least know what they are. A hybrid creature stuck in a transitional phase, from which it either leaps quickly, or dies within it.
Managers who wait for AI to prove its value before starting to streamline are working with the logic of consequence. Real transformation requires the logic of cause. Layoffs first, then AI. Not the other way around.
What's different: nature versus the boardroom
And here's where the most important difference between evolution and strategy comes in.
In nature, no one decides the founder effect. Whoever happened to be there survives. A small group stranded on an island after a flood isn't selected for its fitness — it's selected by chance.
A company has this advantage. A company has to use this advantage.
Cutting headcount by 30% without a strategy is a natural disaster. The survivors are whoever didn't happen to be on vacation, whoever sat on the right project, whoever was in their manager's good graces. That's not a founder effect — that's a lottery.
A cut carried out with strategy is a deliberate assembly of a new founding population — a selection of “DNA carriers,” of the new skill you want to spread. Fitness for the AI era can't be measured by historical performance; what matters is the ability to function in an environment that doesn't fully exist yet.
Nature selects randomly. You don't have to.
What this means in practice
A company entering AI transformation needs to answer three questions before it reaches for a list of names.
1. Where do coordination costs arise? Which roles primarily translate, approve, and synchronize, instead of producing? That's where the quadratic cost is highest, and that's where the cut is needed.
2. Who carries the new capability you want to spread further? Don't look for whoever performed best under the old model — look for whoever is capable of operating in the next one. These people need to survive, and they need room for the new culture to take root.
3. How fast do you want to grow back? The founder effect only works while the population stays small. A company that lays people off and immediately hires new ones will overwrite the new culture with the old one. A new culture needs time before it becomes dominant.