The goal of a company’s organizational structure is the routing of information.
A few weeks ago, Jack Dorsey (CEO of Block, the company that laid off more than 4,000 people a few months ago) published an essay that caught my attention. I’ll leave the link below because it’s worth reading.
It’s not that I think the content is brilliant, I have my doubts about the underlying reasoning and whether it’s partly a justification for his earlier decision to slim down the company — but I did find the reflection interesting. I don’t agree with his analysis, but I do think it deserves analysis..
The essay is called From Hierarchy to Intelligence, and it starts with the Romans. Specifically, with the Roman legion as the first large-scale information-routing system: eight soldiers, one decanus; ten decani, one centurion; six centurions, one cohort; ten cohorts, one legion (I still remembered that from my student days). The pyramid wasn’t a power structure, it was a Roman engineering solution to the problem of coordinating thousands of people when information travels slowly and people, let’s say, have varying levels of understanding.
His essay focuses on this idea: if hierarchy exists to route information, and AI can route information better than any human... then why do we need hierarchy at all? A very optimistic conclusion, considering we spend half our day telling ChatGPT “you didn’t understand me” or “let’s start over,” while Claude responds with messages like “I don’t have access to…” or “I can’t do XXX,” or you simply run out of tokens. That’s why I say that, just like with his February announcement, you can sense a motive or deeper agenda behind this article.
First, the layoff of 4,000 people.
Because you can’t talk about the essay without talking about what happened just before it: in February, 40% of Block’s workforce (more than 4,000 people) left the company involuntarily. Jack announced it publicly in a post on X, saying it wasn’t a cost-cutting measure but a permanent restructuring because the company no longer needed that layer of middle management.
The reaction was predictable: some applauded the courage to say the uncomfortable truth, while others called it textbook AI-washing, using the AI narrative to justify what was really a correction to the massive overhiring of the COVID years.
And honestly, there’s probably some truth in both views. Does AI allow companies to operate with fewer middle managers? Probably yes, in some contexts. Did Block grow from 3,900 employees in 2019 to more than 12,000 in 2022 and now suddenly “discover” it only needs 6,000? That also says a lot.
But I think it’s a mistake to stop there, in the debate over whether Dorsey is a visionary or an opportunist, because the underlying question is far more interesting.
What is hierarchy in a company actually for?
That’s what hooked me about the essay, because at Bjumper we’ve spent years thinking about company structure, and we’re only medium-sized. In fact, we constantly rethink even the titles or roles we put in email signatures. And one concept stayed with me: the purpose of a company’s organization is the routing of information.
I really liked the historical timeline he worked through. I had no idea about many of the examples he mentioned. It’s fascinating to learn that corporate hierarchy didn’t come from management philosophy — it came from the Prussian army, passed through the American railroads of the 1850s, and arrived in modern companies through Frederick Taylor and McCallum’s organizational chart.
All that structure, layers, reporting lines, middle managers, steering committees, review meetings, isn’t fundamentally about organizing power (a lesson many people should learn), it’s a solution to the problem of coordinating people when you have no other way for information to flow.
Seen that way, hierarchy is basically a human network protocol. And like any protocol, it can be replaced if a better one appears — although, spoiler alert, consultancies and large corporations have spent decades trying new formulas, and none have really solved the core problem.
In fact, we’ve spent years talking about flattening organizations, agile structures, squads, holacracy... and several other concepts I discovered while researching this article. Yet no experiment has truly survived at scale: Spotify eventually moved back toward a more conventional structure after launching its tribe-and-squad model; Zappos also tried it with its circles (goodbye bosses!!) and ended up with major talent loss, where employees without managers became trapped in bureaucracy... reality just never quite clicks into place. And the funny thing is that while writing this article and researching the topic, I realized where my boss’s ideas over the last 20 years have come from… because I can honestly say I’ve personally tried all of these models. Luis really does love circles and tribes!!
It’s true, though, that all those models still had the same underlying issue: you need a mechanism for information to flow, for priorities to align, for someone to know what’s happening across the organization. And without hierarchy, we still don’t really know how to do that… at least not today.
What Dorsey proposes is different in one key way: it’s not “let’s remove hierarchy and trust culture.” It’s “let’s build a system that does what hierarchy does, but better.” A model of the company that knows in real time what’s being built, where the bottlenecks are, what works and what doesn’t — without information needing to climb five layers of reporting before reaching whoever has to make a decision.
What concerns me, however, is what the essay leaves unsaid:
Hierarchy doesn’t just route information, it also contains conflict, manages tensions between departments, supports people going through difficult moments, and handles ambiguous decisions where there simply isn’t enough data to know who’s right or what decision to make. All that human and chaotic side of organizations won’t disappear just because we have a good data model. And frankly, I don’t think AI is ready to manage that. I honestly doubt it is right now, no matter how much fun I have like a little kid experimenting with Claude!
I always come back to the same point.
Maybe it’s professional deformation, or more simply, I only see Data Centers wherever I go, read, or listen… that’s just how it is. I work in infrastructure (it’s like a good cult!), so when I read this essay, I inevitably think about data center operations: how decisions are made when something fails, how you manage the knowledge of a technician who has spent 15 years learning the peculiarities of the data center they operate, how teams from different departments coordinate during a critical incident..
Can AI help with all of that? Absolutely. Can it replace the entire structure? That’s where I have more doubts, and I suspect I’ll keep having them for a long time.
In critical environments, the cost of mistakes is extremely high. Trust in a system isn’t built overnight, it’s built slowly, through organized data, processes that are actually followed, and transitions that allow you to maintain control while the system gradually earns the trust of operations. Exactly the opposite of what it means to lay off 4,000 people from one day to the next, with all due respect to Jack… (who, by the way, now seems to be in “quiet rehiring” mode).
My small personal reflection
Either way, I am convinced that this will change the way companies are organized, and also the way Data Centers are operated. And I don’t think the real focus is information itself. Going back to the concept I mentioned earlier, in information routing the focus is on processes. But this can’t be done abruptly, and I have the feeling the solution won’t come from big tech companies. Technology will absolutely be a major part of the equation, of course, but perhaps the real solution will come much more from the humanities and from understanding human behavior. And that still feels very far from AI.
So yes, we read Dorsey, we feel slightly uncomfortable, and then we go back to the same old truth: processes first, data second, technology afterward. As painful as that sequence may be, especially since everyone wants to start with the hardest part.
What do you think? Is this a real transformation, or are we looking at yet another corporate essay that will perform well on LinkedIn and last only as long as the hype cycle tech companies keep feeding us? (sorry for the closed-ended question 🫣)
Block layoffs statement, February: https://x.com/jack/status/2027129697092731343