In May 2026, Cloudflare's CEO Matthew Prince published a letter that was unlike any other public CEO communication that year. The company had just laid off more than 20% of its workforce. It had record revenue growth, strong cash flow, a growing customer base. And yet 20% gone. Prince's explanation: "The vast majority of those we laid off were measurers."

Not underperformers. Not redundant roles left over from an acquisition. Measurers. Middle managers across the organisation, operations functions that existed to coordinate, finance roles that reported on performance, marketing teams that tracked what was working. People who existed, in his framing, primarily to measure and relay what they found.

This is happening at more companies than are saying so publicly. The question is not whether it will happen at yours. The question is whether you understand why and what it means for how you run the company going forward.

What Measurement Work Actually Is

Measurement work is not strategy and it is not execution. It is the layer in between.

It is the weekly status report. The dashboard someone maintains because the CEO cannot read the raw data. The operations manager who coordinates between engineering and support because neither team has a shared system. The finance analyst who packages numbers for the executive team. The compliance officer who checks whether things are being done correctly. The project manager who tracks whether things are on schedule.

None of this work is useless. It exists because without it, information does not reach the people who need to act on it. The problem is not the people doing it. The problem is that most of this information relay no longer requires a human.

What measurers do
Collect, format, route

Gather information from different parts of the organisation. Format it so decision-makers can understand it. Route it to the right people. Follow up when something looks wrong. This is the work of middle management, operations, and much of finance.

What agents do now
Monitor, flag, assemble

Monitor every business risk continuously, not quarterly. Close the books faster. Flag anomalies in real time. Assemble the relevant context for a decision before the human even knows the decision is needed. No latency. No formatting required. No follow-up email needed.

What does not change
Judgment and relationship

Judgment about what matters. Relationships with customers, employees, and partners. Accountability for outcomes. The decisions that require a human to own the result. Builders and sellers. None of that is going anywhere.

The Phrase That Clarifies This

A comment on Prince's announcement put it precisely: "They were translators. People who existed to relay context between layers that could not talk directly. Agents do not just measure faster, they eliminate the need for human middleware entirely. The org chart does not flatten. The middle just disappears."

Human middleware. This is the clearest description of what is changing. Middleware is a software concept: it is the layer that sits between two systems and translates between them. In companies, humans have been playing this role for decades. The meeting is middleware. The status report is middleware. The operations function is often almost entirely middleware. People who exist to make sure one part of the company knows what another part is doing.

Agents eliminate the need for that translation layer. Not because they are smarter than the people doing it. Because they can be in all places simultaneously, work continuously without fatigue, and route information to whoever needs it without the latency of a human relay.

What This Means for Your Organisation

The companies moving fastest on this are asking two questions about every role: what does this person spend most of their time doing, and how much of that is information relay versus judgment and relationship?

The information relay portion is what agents are replacing. The judgment and relationship portion is what is accelerating. When agents handle the measurement and relay, humans have more time and better context for the decisions that actually require them.

Cloudflare's Prince described what this looks like from the CEO perspective: "As CEO, I've never had better tools to measure exactly how the business is performing, including identifying our rising stars." Not because he has more staff. Because he has agents doing continuous measurement that used to require teams.

Organisation before agent measurement

Reports. Meetings. Summaries.

CEO visibility
Gets weekly reports, formatted by analysts. Signal arrives after a round of human packaging.
Risk detection
Risks identified quarterly by internal audit. Problems surface when someone notices them.
Coordination
Happens in meetings between function heads. Requires a human to schedule, attend, and follow up.
Performance data
Visible in monthly close process. Packaged manually before it reaches anyone who can act on it.
Senior leader experience
Insulated from raw operational data by layers of summarisation. Gets the conclusion, not the signal.
Organisation with agent measurement

Live signal. Direct access. Real time.

CEO visibility
Queries live company state directly at any time. No analyst layer between question and answer.
Risk detection
Every business risk audited continuously. Anomalies surface in real time, before they compound.
Coordination
Happens through shared agent systems, not meetings. Context moves without a human to carry it.
Performance data
Visible in real time without manual packaging. The close is a formality, not a discovery process.
Senior leader experience
Direct access to the signal, not the summary. Leaders engage with reality, not a representation of it.

The Part That Is Easy to Get Wrong

The misread: this is about reducing headcount to cut costs.

The right read: this is about redirecting human effort toward work that compounds rather than work that relays.

Companies that treat this as a cost-cutting exercise will cut the measurement layer and still have the underlying coordination problem. They will have fewer people producing status reports and the same information gaps as before, just with worse visibility.

Companies that treat it as a redesign question will replace the measurement layer with agents, maintain or increase the number of builders and sellers, and find that the company makes faster and better decisions because the information gets to the right people without the relay cost.

"The interns we hired are extremely qualified and AI-native. They're all builders or sellers, and we expect that the majority will get full-time offers." — Matthew Prince, CEO, Cloudflare

The company is not smaller. It is rebalanced. Fewer measurers. More builders and sellers. Agents handling the relay in between.

The First Question to Ask

The diagnostic question is: in your company, what percentage of each function's time is information relay versus judgment, relationship, and direct execution?

That percentage is approximately the portion that agents can handle. In most companies it is higher than expected. Finance teams often spend 40 to 60% of their time on reporting rather than analysis. Operations functions are often almost entirely relay. Middle management layers in mature companies can spend the majority of their time coordinating rather than deciding.

The First Build picks one of those relay functions, maps how the information actually flows, and replaces the relay with an agent. Not a better tool for the human doing the relay. An agent that does the relay itself. Continuously, without latency, with better accuracy than a manual process.

Which functions in your company are relay functions?

Bring it to a free Diagnostic. 30–45 minutes, one conversation. It maps where information relay is happening, what the relay is costing you, and what the first agent replacement looks like.

Book the Diagnostic →
Sources
1Matthew Prince (CEO, Cloudflare), on AI replacing "measurers" and restructuring middle management. Via @wallstengine on X, May 2026.
2Tom Blomfield (YC General Partner), How to Build a Self-Improving Company with AI, YC Root Access, May 2026. On reimagining company structure when AI handles information flow.
John Tan
John Tan

Fractional AI & Product Founder at nativefirst.ai. Ex-CEO, Depict (Y Combinator). Embeds on-site with scaling founders and CEOs to ship Level-3 agents and AI workflows in production.