The measurers: what Cloudflare's CEO told the rest of us this week

On Tuesday, Matthew Prince, the CEO of Cloudflare, published an op-ed in the Wall Street Journal titled "How I Choose Which Cloudflare Employees to Replace With AI." It is not an interview he gave reluctantly. It is not a leaked memo. It is a deliberate, named, signed argument, placed in the opinion section of a paper that every CHRO in enterprise tech reads with their morning coffee. The piece names a category of human work as obsolete, and the company has already acted on the claim: roughly 20% of Cloudflare's staff were cut earlier this year, in a quarter when the company posted record revenue.¹

I want to be careful here. The story is not that a profitable company laid people off; profitable companies have been doing that for forty years. The story is what Prince chose to say about it, where he chose to say it, and what his framework now licenses other executives to do.

The taxonomy, and what it carries. Prince's argument, borrowed from Peter Drucker and updated for 2026, sorts every employee into one of three buckets. Builders make the product, engineers, designers, product managers. Sellers bring in revenue, sales, marketing, customer success. Measurers count, verify, audit, and coordinate, finance, legal, internal audit, revenue recognition, middle management. AI, in Prince's telling, is best at measurement first, because measurement is "legible" in a way that building and selling are not.² So measurers go first. Cloudflare is hiring into the other two categories at the same time it is cutting this one.

There is a tidy logic to this. It is also, on closer reading, a redistribution of risk dressed up as a taxonomy.

Who measurers actually are. The "measurer" category sounds abstract until you write out the job titles: revenue recognition analyst, internal auditor, financial controller, compliance manager, mid-level operations lead. These are mid-career roles. They skew toward workers who came up without elite engineering credentials, and they have been, for two decades, the internal mobility ladder for people who entered tech without a CS degree. The category Prince has named as newly redundant is, in distributional terms, the layer of a tech company most populated by people who got there the long way.

Cloudflare had roughly 3,500 employees at the end of 2025. A 20% cut implies the company let go of something in the region of:

I do not know the demographic breakdown of those cuts, and Cloudflare has not published it. But the category Prince has publicly identified as the target tells you most of what you need to know about who paid. Builders and sellers were protected. The audit, finance and middle-management layer was not. This is a choice the firm made, and it is the choice Prince has now made citable.

The accountability gap. Here is where I find the op-ed hardest to swallow. Measurers do not only measure. They verify. Internal audit catches procurement fraud. Revenue recognition catches channel-stuffing. Compliance flags illegal hiring. The whole point of the function is that it is structurally separated from the people whose work it checks, because the people whose work it checks have an interest in the check coming out clean.

An AI system trained to measure revenue recognition will be trained, ultimately, on what Cloudflare's management considers correct revenue recognition. It will be tuned by the same people whose results it is meant to verify. The structural independence of the audit function, the thing that makes it useful when something goes wrong, is precisely what Prince's framework dissolves. He has not addressed this in the op-ed. I do not think he can, because the honest answer is that the independence is the cost, and the cost is what he is cutting.

The Association of Certified Fraud Examiners has been publishing the same finding for a decade: the single largest source of fraud detection is tips, mostly from employees, and the second is internal audit. Automated monitoring is third, and the gap is not close.³ This is not an argument that AI cannot help with audit. It is an argument that the people Prince is cutting were doing something the metric of "tasks completed per hour" does not capture, and that something matters most exactly when things go wrong.

Efficiency for the operator is precarity for the worker, and deferred risk for everyone else.

The same week, the opposite framing. On the same week Prince published, David Solomon, the CEO of Goldman Sachs, published an op-ed in the New York Times arguing that fears of AI-driven mass unemployment are "overblown."⁴ Two CEOs, both in the financial-services-adjacent corner of the economy, same week, opposite framings.

It would be easy to read this as a debate, and the financial press has been reading it that way. I don't think it is one. Solomon is making a macro argument: in aggregate, across the economy, over years, AI will not produce mass unemployment, because historically technology transitions have not. He may be right about the aggregate. Prince is making a firm-level argument: at his specific company, this specific quarter, he chose to replace specific people with AI, and would like other CEOs to feel comfortable doing the same. These claims are not actually in tension. You can believe Solomon's aggregate forecast and still notice that the path to that aggregate runs through several hundred actual people at Cloudflare losing their jobs in a quarter the company made more money than it ever has.

The aggregate is a comfort to the people who will eventually be reabsorbed. It is no comfort at all to a 47-year-old compliance manager in Austin trying to find a comparable role in the same city, at the same salary, in a sector that has just publicly declared her function obsolete.

The retraining story, and why it is mostly fiction. Prince's framework gestures, as these arguments always do, at the idea that displaced measurers can move into builder or seller roles. Cloudflare is hiring, after all. The implicit story is one of fluid internal mobility: out of audit, into product.

This is not what the BLS occupational mobility data shows. Workers displaced from administrative, compliance and middle-management roles transition into technical roles at very low rates within a two-year window.⁵ The pathway from "revenue recognition analyst, 12 years' experience" to "software engineer, hired externally" does not really exist at the scale the displacement implies. The pathway to "revenue recognition analyst at a different company that has not yet automated the function" exists, but it is a shrinking chair in a game of musical chairs that Prince has just told every other player how to play.

The template effect. This, in the end, is the part of the op-ed I keep coming back to. Prince did not have to publish this. Cloudflare could have done the layoffs with a standard "evolving our cost structure" press release, and the story would have lasted a news cycle. Instead, the CEO chose to write 1,200 words in the Wall Street Journal naming the category of worker he replaced, explaining the reasoning, and citing Drucker.

I read this as the deliberate establishment of a precedent. Until this week, "middle management headcount" was a line item that public-company CEOs gestured at carefully on earnings calls, because cutting it implied something had gone wrong, or that the company had been carrying fat. Prince has now provided a vocabulary that converts those cuts from a sign of distress into a sign of strategic clarity. Every CHRO in enterprise tech now has a citable framework, in a respectable venue, by a respected CEO at a profitable company, for treating the audit and compliance layer as a defensible cut.

This is how norms move. Not through legislation, not through some announcement of a new era, but through one op-ed at a time, in venues that the next 200 executives read. The Q3 earnings calls this year will be full of "measurer" language. I will be surprised if they are not.

What is actually being redistributed. I want to name what I think Prince's framework actually does, underneath the Drucker references and the cool taxonomic prose. It redistributes risk. Cloudflare keeps the upside of the AI measurement tools, the cost savings, the productivity numbers it can put in front of shareholders. It transfers the downside, the missed fraud, the misclassified revenue, the unflagged compliance failure, to a future quarter, to shareholders, to regulators, to whichever auditor eventually has to find what the AI system was not configured to find. The workers it cut absorb the immediate cost. Everyone else absorbs the deferred one.

This is a known move. Financial firms have been doing it with risk management since the 1990s, shrinking the independent function, automating what remains, and absorbing the eventual blowup as a cost of doing business. It is not a coincidence that the framework Prince has chosen maps cleanly onto this template. It is a coincidence, possibly, that he published the op-ed in the same week the Financial Stability Board has been warning about exactly this pattern in AI deployment in financial services.⁶

What I think is going on. I do not think Prince is lying about AI's capabilities. The systems he is using really are better at large-scale measurement than they were two years ago, and they will continue to improve. I do not think Cloudflare's layoffs were uniquely cynical, by the standards of a market in which Microsoft, Meta, Google and Salesforce have all been doing variants of the same thing for eighteen months. And I do not think middle management, as a category, is sacred. Some of it was always overhead.

What I think is going on is that an executive class has discovered that AI gives them a public, defensible framework for a redistribution they wanted to make anyway, from labour to capital, from the middle of the workforce to the top, from people who carried the firm's risk to systems that pretend it has disappeared. Prince's contribution is to have said this out loud, with citations, in the Wall Street Journal. He has done the rest of the executive class a favour. He has not done the measurers one.

What to watch. Four things, over the next four quarters. First, whether Cloudflare's own compliance record holds, and whether any material misstatement, missed audit flag, or regulatory action emerges from a function it has automated. Second, whether Prince's vocabulary appears, by name or by paraphrase, in other companies' Q3 and Q4 earnings calls. Third, whether the workers cut in this wave find comparable employment within twelve months, and where. Fourth, whether the Financial Stability Board, the SEC, or any audit regulator takes any interest in what is happening to the independence of internal audit functions when they are absorbed into AI systems tuned by the management they are meant to check.

I expect the first three to be answered before the fourth even starts. That, too, is part of the redistribution.

Footnotes

¶