Cloudflare's Layoffs Exposed the Broker Layer

Matthew Prince gave executives a brutal new vocabulary: builders, sellers, and measurers.

I think the more useful model is sharper:

producers, consumers, and brokers. 🧭

The broker layer is where the floor just moved.

Cloudflare said on May 7 that it would cut more than 1,100 roles while redesigning the company for the agentic AI era. In the same announcement, Cloudflare said internal AI usage had jumped more than 600% in three months and that employees across engineering, HR, finance, and marketing were running thousands of AI agent sessions per day (Cloudflare).

Then Prince published the louder version of the argument: builders and sellers are safe, measurers are exposed. Fortune summarized the WSJ op-ed as a claim that the majority of cuts hit "measurers" such as middle management, finance, legal, internal audit, and revenue recognition (Fortune).

The obvious reaction is to argue whether Prince is right, cynical, or both.

The more useful reaction is to ask a colder question:

What part of your job creates new value, and what part only moves information around?

🔴 The dangerous role is the broker

Most companies pretend jobs map neatly to titles.

They don't.

A single person can be all three roles in one day:

• Producer: creates something new: product, code, strategy, customer insight, a decision, a plan people can act on.

• Consumer: uses information to make a judgment: buys, prioritizes, approves, rejects, gives feedback, commits budget.

• Broker: moves information between producers and consumers: summarizes, forwards, routes, reconciles, schedules, tracks, checks status.

The broker is useful when they add judgment.

The broker role becomes exposed when it turns into a human API gateway.

If your job is mostly turning Slack into slides, slides into meetings, meetings into trackers, and trackers back into Slack, the risk is not that you are untalented.

The risk is that the workflow was designed around moving information instead of creating new value.

That is the uncomfortable part of the Cloudflare story.

Prince called the exposed group "measurers." Fortune also noted the skeptical counterargument: some companies may be using AI as a cleaner narrative for headcount reductions that also reflect overhiring, bureaucracy, or margin pressure (Fortune).

That skepticism matters.

Cloudflare's own Q1 release shows the messy reality: $639.8M in revenue, up 34% year over year, plus a $62.0M GAAP operating loss and $140M-$150M in expected restructuring charges (Cloudflare Q1 release).

So yes, "AI did it" can become convenient executive theater.

But the theater does not make the underlying shift fake.

And none of this should flatten the human cost.

Layoffs are brutal. Careers get interrupted. Families absorb the shock. People who did exactly what the company rewarded for years can suddenly be told the system changed.

That deserves empathy.

But empathy should not make us lie about the work pattern.

📌 Measuring is valuable. Brokering without judgment is not.

This is where Prince's model needs an upgrade.

Measurement is not automatically low-value.

In serious companies, measurement is how you avoid lying to yourself. It is how you know whether users are succeeding, whether revenue is real, whether risk is acceptable, whether your product is actually getting better.

Bad measurement is dashboard cosplay.

Good measurement changes decisions.

The same is true for brokers.

A weak broker says: "Here is the latest update."

A strong broker says: "Here is the decision we need, here are the tradeoffs, here is the risk, here is my recommendation, and here is who owns the next step."

One is routing.

The other is judgment.

AI is very good at routing.

It is getting very good at summarizing, reconciling, drafting, comparing, extracting, scheduling, and checking. LinkedIn's editor-picked reactions around Prince's op-ed picked up the same theme: the compressed layer is coordination overhead: reporting, approvals, version reconciliation, and operational waiting time (LinkedIn News).

That should terrify anyone whose value is measured by how many internal messages pass through them.

📉 We have seen this movie before

The floor trader is the cleanest historical example.

For decades, a lot of financial trading depended on humans standing in a pit, shouting prices, reading hand signals, and physically brokering buyers and sellers.

Then the market became electronic.

Not overnight. Not perfectly. Not everywhere.

But directionally, the broker layer collapsed.

CME Group said in 2015 that open outcry futures trading had fallen to just 1% of its total futures volume before it moved to close most futures pits in Chicago and New York (CME Group).

The SEC approved the NYSE's hybrid market in 2006, allowing the exchange to combine its traditional floor auction with enhanced automated trading (SEC).

That is the real analogy.

Electronic trading did not remove the need for markets.

It removed a large class of people whose job was to stand between intent and execution.

The traders who survived were not the ones who were best at yelling.

They were the ones who moved up the value chain: risk, strategy, market structure, algorithms, relationships, judgment.

That is what AI is about to do to corporate information work.

If your role is being used as the human trading pit for internal knowledge, the screen is coming for the workflow.

That distinction matters.

People deserve dignity.

Workflows still deserve scrutiny.

🎯 The new career question

The old career question was:

"Am I important to the org chart?"

The new one is:

"Would the work lose meaning if I disappeared, or would the system just route around me?"

That is harsh, but useful.

If you are a producer, AI can make you faster.

If you are a consumer, AI can help you evaluate more options with less latency.

If you are a broker, AI forces you to prove that you add something beyond transfer.

That "something" is usually one of five things:

• judgment under uncertainty

• accountability for the outcome

• trust with a customer or team

• taste about what good looks like

• context the system cannot infer from documents

No one is safe because their title sounds strategic.

No one is doomed because their title sounds operational.

The dividing line is behavior.

A finance partner who changes a product decision is producing.

A product manager who only forwards customer feedback is brokering.

A chief of staff who resolves ambiguity is producing.

A VP who turns every decision into a pre-read is brokering with a larger calendar.

AI will expose the difference.

🧨 The broker layer was already too big

The dirty secret is that many companies built huge broker layers because software made coordination cheap enough to tolerate and growth made headcount easy enough to hide.

More dashboards.

More syncs.

More status docs.

More people asking other people for updates that already existed somewhere else.

That worked when the cost of internal latency was invisible.

AI makes the cost visible.

It also gives founders and executives a very tempting button: compress the broker layer, call it an operating-model shift, and move the savings toward people who make or sell.

Sometimes that will be smart.

Sometimes it will be lazy.

The difference is whether the company preserves accountability while cutting coordination.

Removing human brokers without redesigning decision rights just creates automated confusion at higher speed.

🧭 What I would tell teams

Do not ask, "Will AI replace my role?"

Ask:

• What do I produce that did not exist before?

• Who consumes my work, and what decision does it improve?

• Where am I merely brokering information?

• Which broker loops can I automate myself before someone else automates me out of them?

• What judgment do I add that a model cannot responsibly own?

The best people will not wait for a reorg to answer those questions.

They will convert broker work into producer work.

They will turn "I coordinated the update" into "I made the decision easier."

They will turn "I built the dashboard" into "I changed the roadmap."

They will turn "I summarized the customer call" into "I found the product gap and drove the fix."

That is the move.

AI does not punish people for touching information.

It punishes people who touch information and leave it in the same state.

The next org chart will not be built around who controls the flow of information.

It will be built around who creates value after the information arrives.