Reading the Room

Something is happening to the org chart. You can see it in the language before you see it in the structure. Nine functions collapsing into three. A hundred-person team planning for twenty-five. The specialist giving way to the generalist who runs a fleet of agents instead of a team of people. The title you trained your whole career for, dissolving into something broader.

Foundation Capital recently mapped this shift in a piece called "The Great Reorg", and the mapping is good. The claim is that this is new, and that the new shape is better. Flatter. Leaner. Fewer handoffs, less coordination, more speed. And the claim is mostly right. When the cost of a task falls to near zero, you don't just do it faster. You get to ask whether it needed to be a task, whether the role around it needed to be a role, whether the box on the chart was ever describing anything real.

That last question is the one worth staying with. Because the boxes were never quite real.

Section I · Recognition

The Title Is Only the Floor

I keep having the same feeling reading about the collapse. Not disagreement. Recognition.

Here is the thing the org chart never told you. The title describes the floor of a role, not the whole of it. What a person actually does on a given team is an amalgam. The job as written, plus whatever the rest of the team left for them to carry, plus whatever their particular gift happened to be, plus whatever they saw was possible beyond the role and took on anyway.

The first three accrue to you. The last one you reach for. A title tells you what someone is responsible for; it says nothing about what they can see past it. The people we call 10x aren't just better inside the lines, they redraw them, taking on what the role never contemplated because they can see it's possible and worth doing. That reach is the least legible thing on any chart, and usually the most valuable.

My last CFO was a great dealmaker. He read a room better than most people at the table. None of that is in the CPA. The certificate promises controls, filings, the clean close. The floor. The part that mattered most on our team was nowhere in the credential. It had attached itself to him because he was good at it and because no one else was going to do it. On another team, with another CFO, the same title sits over a controller who would be lost in a negotiation, and the dealmaking has quietly migrated to someone else entirely. Same box. Same title. A completely different machine underneath.

So when we talk about collapsing the CFO function, the question is never really about the CFO. The question is what had actually collected under that title, on this team. And that is not written down anywhere. It lives in the room. It is also why organizations are so hard to read from the outside. The amalgam was never in the deck.

Section II · The Unread Room

What the Collapse Can't See

This is a process problem, and it is not a new one. Understand what a role actually contributes to the flow of work before you decide what happens when you remove it. That is the oldest question in business process engineering, and we have answered it badly many times before. I spent years in that world, back when the actors were deterministic and the orchestration had to be modeled by hand. The lesson from that era was not that automation makes you fast.

Automating a step you do not understand just lets you make the mistake faster, and at scale.

And the amalgam is exactly the part we do not understand, because it was never documented, only lived. Three things follow from that, and they are the three places the reorg will quietly break.

The first is the independent no. Separate functions were never just a coordination mechanism. They were an adversarial one. The handoff between the deal team and legal, between engineering and product, between finance and the business unit. Those boundaries were also review boundaries, a set of structural objections sitting in different reporting lines with different incentives. But where the real check lived on any given team was part of the amalgam, not the chart. Collapse the boxes without reading the room, and you delete the coordination cost and the adversary in the same stroke. The builder becomes the checker, and self-review is the weakest review there is.

The second is friction. It is fashionable to treat organizational friction as pure waste, and much of it was. But some of that friction was the time it took to read the room, the slow accumulation of knowing who actually carried what. The old constraints forced the reading. Removing a role used to require a case, a reorg, political capital. Now it requires deleting a function and pointing an agent at it. The discipline is the same as it ever was. The forcing function is gone.

The third is speed, and speed is the dangerous one, because it never has to argue for itself. Cost-cutting at least announces a tradeoff. Speed presents as pure virtue. Nobody stands up in the room and asks to go slower so the amalgam can be read. And the cost of the unread room is deferred and invisible, right up until it isn't.

Three observations, one root. The reorg's outcome depends on seeing the real distribution of work, and everything about this moment, the tooling, the vanished friction, the speed, conspires against the seeing.

Section III · The Case for Collapse

The Better Argument for Collapse

So is this an argument against collapsing roles? No. It might be the best argument for it.

The titles were always fiction. The boxes never described the real allocation of work. They were an inherited approximation everyone quietly worked around, and a lot of what we call coordination cost was just the tax of maintaining boxes that didn't match reality. If the collapse means we finally stop reasoning about titles and start reasoning about the work as it's actually distributed, the result is a better organization, not just a cheaper one.

The process was always the truth. The title was the tax.

But that only holds if someone can see the actual distribution. And here the org chart is no help, because the amalgam was never written down. It formed in the room.

Brian Crook in the server room during the Cerebra years

Cerebra, 2008

I learned this building startup teams, where you never get to hire to a template. You go to war with the skills you have. Every early team is a misfit island, backgrounds that shouldn't fit together, and somehow the mismatch is the depth. The best ones I've been part of were exactly that, a cacophony of experience no job description would have assembled on purpose. You don't hire for cultural fit on a team like that. Fit reproduces what you have. You hire for cultural add and desired impact, the strength you're missing, the perspective nobody at the table carries yet.

And as the team grows, you architect it the way you'd architect a platform. Decide where you're going. Find the gaps between here and there. Hire strength into the gaps. Not titles into boxes. Big companies tend to do the opposite as they scale. Template the role, rinse, repeat. It's efficient, and it's how the gap between the title and the work gets baked in. The template describes the floor of the role. The team was always built from the whole of the people.

I've been the beneficiary of this myself. I've spent a career as the product-and-technology fusion, never the academic, never the CIO, able to see the problem and build the solution without a handoff in between. That merge is now showing up on the slideware as the future of the org chart. It's real. But it wasn't imposed on me by a diagram. It emerged from what I could actually do, on teams that were composed around actual capability. That's the difference between a collapse that works and one that just deletes the person who did the work and keeps the box.

Section IV · Agents Are Hires

Hiring, When the Hire Is a Machine

Which brings us to the question underneath all of this. Can you compose a team this way when some of the members are agents?

I think yes, because it's the same act. An agent joining a team is a hire. You're not adopting a technology. You're adding a set of strengths to a specific table, and the discipline is the one hiring always required. Know the gap you're filling. Understand what this addition is actually good at, not what its title claims. And hold to the oldest rule in the book. Hire slow, fire fast. Take real care deciding what an agent should carry. Then validate ruthlessly what it actually does well, and drop what it doesn't. Quickly, without sentiment, the way you'd act on a hire that isn't working.

There's one place the analogy breaks, and it's the place to pay attention to. A struggling human hire announces themselves. They hesitate. The team routes around them. The room tells you, ambiently and for free, that something's wrong. An agent gives you none of that.

It fails confidently, at full speed, with no tell.

So "fire fast" isn't a posture you can hold by instinct anymore. You have to build the legibility yourself. Before an agent joins the team, decide how you'll know it's failing. That's not an observability feature. That's the precondition for the hiring discipline working at all. The signal the room used to give you for free, you now have to architect.

And this is where the reorg's hardest question comes back around. Foundation Capital names this one directly; they call it the one-generation problem, the risk that if agents do all the entry-level work, the pool of people who can judge that work runs a single generation deep and then dries up. The people who can read a room, who can see what actually collected under a title, who know a gap when they're staring at one, learned it by doing the work, on teams, over years. If the agents absorb the early reps, the next generation may never sit in enough rooms to learn to read one. The judgment this whole transition depends on is the thing it's most at risk of no longer producing. They named the risk. What neither they nor I can yet say is how you rebuild a capacity that only forms in the room.

The Practice

I don't have a prescription for that last problem. I'm not sure anyone does yet.

What I have is a practice, and it's the same one it's always been. Understand the work before you touch the structure. Compose for strength, not for the box. Hire slow. Fire fast. Build the signal you'll need before you need it. None of this is new. It's just newly urgent, because the friction that used to force the discipline is gone, and the speed that rewards skipping it is everywhere.

A shelf of team-building and strategy books, including Good to Great, Radical Candor, StrengthsFinder, Crossing the Chasm, and The Four Steps to the Epiphany

Some of the thinking behind the practice

The teams I'll build from here, human, agent, and whatever mix comes next, will be built this way. Quietly, around what people and systems can actually do.