Matthew Prince, the CEO of Cloudflare, cut more than twenty percent of his company in the same stretch it was growing over thirty percent a year, and then he did the rare thing of explaining the logic in the Wall Street Journal. He wrote that he could not find another case in US business history of a public company growing that fast while cutting that deep. The explanation rests on one distinction. Every employee is a builder, a seller, or a measurer. Builders make the product, sellers sell it, and measurers do the work in between: the auditing, the finance, the reporting, the layers of management whose job is to check on other people's work. AI can now measure a company continuously, at a level of detail no human team could hold, so measuring is the layer that thins.
The framework is clean and the direction is right. What matters is what a leadership team does with it, because the lesson most will copy was never built for a company their size.
Where the framework holds
Grant Prince the strong version of his case first, because it is largely right. Measurement is exactly the kind of work that gets cheaper and better as the models improve. It is structured and repetitive, it runs against data that already exists, and it rewards tireless attention more than judgment. Cloudflare moved internal audit from sampling a few risk areas each quarter to checking every risk continuously, closed its books faster, and caught more of its own mistakes along the way. The interns make the same point from the other end. The company hired more than a thousand of them out of nearly a million applicants and described them all as builders and sellers, the two categories it still wants more of.
There is a skeptical reading, and it deserves air. Marc Andreessen has argued that large companies were simply overstaffed, and that AI now gives them a convenient reason to make a cut they already wanted. Some of that is surely true. It does not weaken the part that matters here. Whether a company was bloated or lean, the work AI absorbs first is measurement, and measurement is the work most companies have the least visibility into.
Why the cut won't translate
At more than five thousand people, measurers are a department. They have titles, they sit on the org chart, and a CEO can point at the function and remove a fifth of it. A company of fifty to three hundred people does not have that department. The measuring still happens, often more of it per head, and it lives inside the people you would never want to lose. Your strongest engineer spends a third of the week reconciling numbers between two tools, your best salesperson updates the CRM instead of selling, and your operations lead rebuilds the same Monday report that nobody reads to the end. The work is real and constant, and it is spread across the exact builders and sellers Prince tells you to protect.
So the Cloudflare move, copied straight, has nothing to grip. There is no measurer floor to clear. Go looking for one and you start cutting into the people who actually build and sell, which is the opposite of what the framework was meant to do.
What Cloudflare built before it cut
There is a second reason the cut travels badly, and it sits in the order of operations. Cloudflare did not fire its measurers and hope AI would cover the gap. It had already moved the company onto continuous measurement, with every risk audited, the books closing faster, and the CEO able to see how the business was performing in close to real time. Only once the measuring ran without those people did removing them become safe. The layoff was the last step of a rebuild, not the first.
Most companies that copy this will copy the visible half. They will take the headcount cut, skip the long work of rewiring how measurement actually happens, and find that the work did not disappear. It moved back onto the people who are left. That is the road into the sixty-one percent: the organizations that put real money into AI and report no change they can measure in their profit, because they added the tools to the way they already worked and the work underneath never changed.
Where the margin actually sits
For a company that is not Cloudflare, the better reading of this story has little to do with cutting. The measurement buried inside your best people is capacity you already pay for and do not get back. When AI takes the reconciling and the report-building and the checking, that time comes back to your most capable people, and they spend it on the work only they can do: the building and the selling and the judgment that does not survive being handed to a model. Nobody has to lose a job for that to be the larger win, and in a smaller company it usually is, because the constraint there was always your best people's attention, and measurement is what quietly eats it.
Getting there is quieter, harder work than a layoff: looking at how a function actually runs, naming the measurement that no longer needs a human, and rebuilding the role around what is left. That is the same move the thirty-nine percent who get real results from AI already made. It does not photograph well and it does not fit in a headline, which is part of why the loud version of this story is the one about cuts.
The question worth sitting with
The question underneath all of this is quietly concrete. What share of your best people's week already goes to measurement, the reconciling and reporting and checking that AI is about to absorb? Most leadership teams cannot say, and that uncertainty is the first thing worth fixing. When the capability lands, and for much of this work it already has, the decision is easy to describe and hard to live with. You can treat the freed hours as a reason to cut, or as a chance to put your best people back on the work that made them your best people. Cloudflare could answer it one way because it had built the measurement to back the answer. The companies that pull ahead from here will mostly be smaller, will never run that exact play, and will win on the same question anyway, by finding the measurement hiding in plain sight and freeing it.