👋 Hey, Leo here. Antifragile Intelligence is where I explore how leaders think, decide, and build in uncertain environments shaped by technology and change.
Each edition is a reflection, a principle, or a field note from the work itself.
Hey Friends,
Welcome to the 161st edition of Antifragile Intelligence.
On June 22nd, Bülent Duagi and I will host the 3rd session in our AI x Business series, with Marius Ioniță, CEO of Alumil Romania.
This one is called: AI Starts at the Top, and I’m very excited for it.
Last week, I started the 5-edition series on “AI and the end of the Org chart” in which I explain why this might be a good time to rethink your structure.
#160: The Org Chart Is Quietly Dying
·
Jun 12
👋 Hey, Leo here. Antifragile Intelligence is where I explore how leaders think, decide, and build in uncertain environments shaped by technology and change.
In this edition, I will take you through three companies that are at the forefront of this change. They are examples of how companies can change and be more profitable than before.
One thing to remember: There is no single “flat.” There is a design choice.
The popular AI-and-orgs narrative, the one that shows up in every Forbes article and every executive newsletter, implies that the future is a single thing.
The org chart dies. Middle managers disappear. AI takes over coordination.
That is not quite right. There are several serious models of what comes next, and they look surprisingly different from each other.
I want to take you inside each one.
And I want you to notice, this is the important part, that two of these companies built their flat structures before AI existed.
What AI changes is not whether flattening is possible. AI changes who gets to do it.
And how fast.
Start in Qingdao, China, 2005. Haier, the Chinese appliance manufacturer competing with GE and Whirlpool in the global white-goods market, does something nobody else in the industry is willing to do. Their CEO, Zhang Ruimin, looks at his 80,000-person company and announces that the middle is going away.
He calls his new model Rendanheyi. Three Mandarin syllables. Ren means person. Dan means order, or value. Heyi means unity. Together: the unity of each employee with the customer’s value.
What Zhang did was take his 80,000 employees and reorganise them into roughly four thousand microenterprises.
Ten to fifteen people each.
Each microenterprise has its own P&L. The customer-facing ones contract directly with end customers, real people paying real money, not internal stakeholders. There is no middle management layer in between.
There are three flavours.
Transforming microenterprises, existing businesses being reinvented.
Incubating microenterprises, new ventures.
And node microenterprises, the shared platforms (HR, supply chain, manufacturing) that serve the other microenterprises as paying internal customers.
Look at that last category. The platforms didn’t disappear. They got priced.
HR doesn’t sit at corporate doling out services.
HR is a microenterprise that has to earn its keep by selling its services to other microenterprises at internal market prices.
That is the model. Twenty years in. Roughly 4,000 microenterprises today.
No traditional middle management.
A $30+ billion revenue company.
Zhang has told the story himself, most accessibly in McKinsey’s interview with him, Shattering the status quo, where he describes Rendanheyi as a network of entrepreneurs rather than a hierarchy of employees.
And here is something worth keeping in mind: Haier did this before AI existed.
The org form the McKinsey and Block camp is now selling is not a 2025 invention. It has two decades of evidence behind it.
Why couldn’t every other company copy Haier?
Because the infrastructure Zhang built, internal markets, transfer prices, and P&L accountability across 4,000 units, was expensive to build by hand and culturally specific to a Chinese state-adjacent industrial conglomerate with a 25-year-tenured founder.
What AI changes is not whether the form is achievable.
What AI changes is the cost of getting there. AI lowers the price of the internal coordination infrastructure that Zhang had to build the hard way.
Now jump forward two decades. San Francisco. March 2026. Jack Dorsey and his Sequoia investor, Roelof Botha, publish From Hierarchy to Intelligence on the Sequoia Capital blog.
Block has about 12,000 employees.
Square, Cash App, Afterpay, TIDAL.
They are remote-first, which is going to matter in a moment. And the essay proposes a structure with four building blocks instead of a traditional org chart.
Capabilities. Atomic financial primitives. Payments, lending, card issuance, banking, BNPL, payroll. Not products. Building blocks with reliability and compliance targets. No UIs.
The world model. Two sides.
The company world model: how the company understands its own operations, performance, and priorities. This replaces the information that used to flow through layers of management.
The customer world model: a per-customer, per-merchant representation built from transaction data.
Why does Block get to build this? Two preconditions.
First, they are remote-first, which means every decision, every discussion, every design, every plan is already machine-readable. Raw material for the model.
Second, money is the most truthful signal a customer ever produces. Block sees both sides of millions of transactions every day. The model has something rich enough to be worth building.
The intelligence layer. It composes capabilities into solutions for specific customers at specific moments. A restaurant’s cash flow is tightening ahead of a seasonal dip. The intelligence layer composes a short-term loan, adjusts the repayment schedule, and surfaces it to the merchant before they ask. No PM decided that. The capabilities existed. The intelligence layer recognised the moment.
Interfaces. Square, Cash App, the apps you see. Delivery surfaces, not value loci.
And inside this structure, Dorsey and Botha argue, there are only three human roles.
Individual contributors who own a layer of the system.
DRIs, Directly Responsible Individuals, who own a cross-cutting problem for, say, 90 days, with full authority to pull resources from any team.
Player-coaches who combine building with developing people. No permanent middle management layer.
Block’s model is, in a sense, a Haier with AI inside it.
The microenterprise infrastructure becomes the world model.
The internal market becomes the intelligence layer.
What Zhang built by hand, Block is building in code.
Third company. Nvidia. As of mid-2026, the most valuable company on earth.
I am going to give you three facts about how Jensen Huang runs Nvidia.
One. Jensen has sixty direct reports. The Fortune 500 median is between eight and ten. Entrepreneur covered this in detail after the org-data service TeamChart published Nvidia’s chart, and people started counting.
Two. He holds no 1:1 meetings. None. Entrepreneur quotes him saying,
“I don’t have one-on-ones with them because it’s impossible. We present a problem, and all of us attack it.”
In a separate exchange with Stripe’s Patrick Collison that did the rounds last year, he was even sharper:
“I give you feedback right there in front of everybody. I don’t do one-on-ones, and almost everything that I say, I say in front of everybody.”
Three. He says he is not the boss. The mission is the boss.
Nvidia has dramatically less middle management than its peers. Not because of AI. Because of a single founder’s operating philosophy. Three things make it work.
The top-five thoughts cadence. Every direct report sends a short list of their top-five priorities and observations on a regular schedule. Jensen absorbs the cross-pattern across all sixty lists. The signal isn’t any individual list, but the correlation.
The default-to-broadcast habit. Decisions get made with the whole leadership team in the room. Status is shared in the open. The cost of routing information point-to-point is replaced by the cost of having a slightly bigger room.
The public reasoning. Jensen thinks out loud rather than handing down conclusions. The team learns the reasoning, so they can extend it when he is not there.
What Jensen does manually, Block is building in code.
What Block is building in code, Haier built with internal markets.
Three different paths. One destination.
Nvidia doesn’t have fewer middle managers because of AI. It has fewer middle managers because Jensen Huang decided that the work middle managers were doing was wrong work.
AI is now letting other companies make that same decision without needing Jensen.
That is the move. That is what changes in 2026.
The Block essay opens with the Roman contubernium, eight soldiers sharing a tent and a mule under a decanus, and tracks two thousand years of organisational coordination through the Prussian General Staff and the American railroads to today’s matrix.
Their point, and I think it is the right one: for two millennia, the decision to abolish middle layers was reserved for an extreme founder with extreme authority in an extreme context. Zhang Ruimin. Jensen Huang. A handful of others. The rest of us watched and admired and never thought we could copy them.
We can now. Not because we got smarter. Because the infrastructure that they had to build by force of personality is now buildable in software.
The question is no longer can we flatten?
The question is, do we want to, and if so, which architecture do we build?
You have three real choices.
You can build a Haier, an internal market with priced platforms.
You can build a Block, a world model with an intelligence layer.
You can build an Nvidia, a culture of broadcast and public reasoning.
Or you can do nothing and let your competitors pick.
Next week, I want to break the dominant narrative completely with the counter-evidence almost nobody is quoting.
Because the right architecture for your company depends on a much more precise diagnosis than the press is providing, and the diagnosis has to start with the case against flattening, not the case for it.
Stay with me.
Cursor CEO’s 10-minute talk about the next era of AI coding. (Btw, Cursor was just bought by SpaceX)
Finished reading one of the best SF books I have read (not many books, as it was never my preferred type of literature, but I’m getting there). Red Rising, by Pierce Brown . I recommend it a lot, I’m already halfway through the second book in the 6-part series.
AI Automation Society is the best AI Skool community, in my opinion. Great for people who want to automate processes in their life or business and need some structure to begin with.
Thank you for reading.
If this resonated, forward it to someone who might benefit from it.
Stay antifragile.
Leo
If you enjoyed this piece and want to support my writing, please consider sharing it and becoming a paid subscriber.
Thanks for reading Antifragile Intelligence! This post is public, so feel free to share it.
This is Edition B of a five-part series on how AI is reshaping organisations. Next week: the counter-evidence almost nobody is quoting, and why the dominant flattening narrative is wrong in important ways.
Dorsey, J. & Botha, R. (March 2026). From Hierarchy to Intelligence. Sequoia Capital. https://sequoiacap.com/article/from-hierarchy-to-intelligence/
McKinsey & Company. Shattering the status quo: A conversation with Haier’s Zhang Ruimin.
Strategic HRM in Digital Transformation: Haier’s “Rendanheyi” Model. International Journal of Current Science Research and Review.
Nvidia CEO Jensen Huang Never Has One-on-One Meetings With His 60 Direct Reports. Entrepreneur.
Three Management Secrets from Jensen Huang. Tremendous.
TeamChart. NVIDIA Org Chart & Executive Leadership Structure.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.