Think about your last week. How many meetings existed primarily to move information from one group to another? A status update. A handoff. A briefing. A review where one team reported what they had done to another team that needed to know. Count them. They are probably most of your calendar.

Now think about why those meetings exist. Not why each individual one is on the calendar, but why the category exists at all. The answer is the same across almost all of them: information needed a human to carry it. The engineering team could not automatically share context with the product team, so someone scheduled a sync. The CEO could not read the raw operational data, so someone packaged it into a report. The legal team needed to understand what sales had agreed, so someone held a handoff.

Every meeting, every layer of management, every operations function that primarily coordinates rather than builds: these are all solutions to the same underlying problem. Information needed people to move it. That constraint is lifting. And when it lifts, everything built around it becomes optional.

Why Hierarchy Exists

The org chart was invented in 1855 by the New York and Erie Railroad. It was one of the first organisations that needed to coordinate thousands of people across distances too large for direct communication. The solution: nested hierarchies with named individuals at every node, passing orders down and sending information back up.

This was not a choice between good and bad structures. It was the only available technology for the coordination problem. Information had to travel through humans because humans were the only available relay mechanism.

The same logic applies to every layer of management ever added since. Middle managers do not exist because they are inefficient. They exist because someone had to translate between the strategic layer and the execution layer. That translation requires proximity, which requires people.

The information relay chain

How a risk reaches the CEO today

  1. A customer complaint arrives in the support inbox.
  2. A support agent handles it and logs it in the CRM.
  3. A support manager reviews logs weekly and writes a summary.
  4. The VP of Customer Success packages the summary into a QBR.
  5. The CFO includes the QBR highlights in the board report.
  6. The board asks about it; the CEO follows up.

What Changes When Information Moves Itself

Here is the same scenario with agents.

A customer complaint arrives. An agent reads it, cross-references the customer's account history, identifies it as a pattern seen in 14 other tickets this month, flags it as an emerging product issue, drafts a resolution for the support team, and sends a summary to the product team's shared system. All within minutes.

The CEO queries the live company state Monday morning: three emerging customer pain points, ranked by frequency and revenue at risk, with draft responses already prepared. Total latency: minutes, not weeks. No meetings. No packaging. No relay.

This is not speculation. Cloudflare CEO Matthew Prince described exactly this: "Internal audit previously picked a handful of business risk areas to scrutinize each quarter. Now we're moving to a system in which every business risk is audited continuously." Not faster audits. Continuous audits. Qualitatively different.

"Internal audit previously picked a handful of business risk areas to scrutinize each quarter. Now we're moving to a system in which every business risk is audited continuously." — Matthew Prince, CEO, Cloudflare

Information relay through humans

Packaged. Periodic. Latent.

What the CEO sees
Packaged reports, not raw signals.
Risk identification
Quarterly, after packaging by someone further from the data.
Escalation
Depends on whether the right human noticed and acted.
Decision latency
Weeks between signal and CEO awareness.
Coordination cost
Significant headcount dedicated to moving information, not creating it.
Information relay through agents

Live. Continuous. Direct.

What the CEO sees
Live company state, queryable at any time.
Risk identification
Continuous, automated, without human packaging.
Escalation
Automatic, based on defined thresholds, routed to the right person.
Decision latency
Hours or minutes from signal to relevant party.
Coordination cost
Setup and maintenance of the agent layer, not ongoing headcount.

What Remains When the Relay Is Gone

This is the important question. If agents handle information relay, what do the humans do?

Three categories remain irreducibly human.

The Cloudflare CEO's observation points at exactly this: AI is not coming for builders or sellers, but it is coming for measurers. Builders create. Sellers connect. Measurers relay. The relay is what is changing.

The Uncomfortable Audit

Take any function in your company. Map what each person in it does. Sort the activities into three buckets:

Bucket 1
Creating

Making something new, deciding something, building something. This is the irreplaceable bucket.

Bucket 2
Connecting

Relationships, selling, partnering, managing humans through change. Also largely irreplaceable.

Bucket 3
Relaying

Moving information, measuring, reporting, coordinating, packaging context for others. This is the automatable bucket.

The relay bucket is the portion that agents can handle. In most organisations, it is the largest bucket in operations, finance, compliance, and middle management layers. It is not the largest bucket in engineering, sales, or senior leadership.

This is not an exercise in identifying who to let go. It is an exercise in identifying where to deploy agents first, so that the people currently in the relay bucket can move into the creating and connecting buckets.

Building the First Relay Agent

The relay that is most automatable is almost always the one that meets three criteria.

The First Build maps one of these relays, replaces it with an agent, and measures the result. Not in theory. In production, on live data, with real latency reduction visible within two weeks.

The Install builds the relay replacement function by function across the whole company. Each sprint shortens one information journey that currently requires a meeting or a management layer.

Which relay in your company should go first?

Bring it to a free Diagnostic. 30–45 minutes, one conversation. It identifies the highest-volume, clearest-rules relay in your company and maps what it takes to replace it with an agent.

Book the Diagnostic →
Sources
1Matthew Prince (CEO, Cloudflare), on continuous AI measurement replacing periodic human audit. 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 companies as sets of intelligence loops rather than human relay chains.
3@ivanhzhao, Steam, Steel, and Infinite Minds, X, Dec 2025. On the 1855 origin of the org chart as a relay technology and what replaces it.
4Raphaƫl Dabadie (YC P26), Field Work Is the New Moat, X, May 2026.
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.