If you’re using AI to write your weekly reports, you’re already making a catastrophic mistake. Not because the reports are bad—but because you’re using the most powerful organizational tool in history to build a faster horse.
I spent a Thursday afternoon in a tea house in Shekou, Shenzhen, with a group of old colleagues. We were all building AI startups. We talked about the weather, about our kids, about the existential dread of watching our industry reshape itself in real time. And then someone said something that made the whole table go quiet:
“The real reason companies are cutting middle management isn’t cost savings. It’s that AI has made the information router obsolete.”
Later that night, I found the article that had been circulating in Silicon Valley. It wasn’t a PowerPoint deck from McKinsey. It wasn’t a marketing stunt from an AI startup. It was a joint document from Jack Dorsey’s Block and Sequoia Capital’s Roelof Botha. And it wasn’t asking whether AI could help you save money. It was asking something far more dangerous:
If AI can understand your entire company, coordinate your resources, and find opportunities faster than your managers—why do you still have managers?
This is not a theoretical question. Block has been living this experiment. They’ve been cutting layers, compressing management, and watching the old hierarchy burn. And the article they published doesn’t call it “restructuring.” It calls it something else: From Hierarchy to Intelligence.
Let me show you why this changes everything.
The 2,000-Year-Old Hack You Call “Management”
The article starts not with ChatGPT, but with the Roman legions. Two thousand years ago, the Romans faced a problem that every CEO still faces today: how do you get thousands of people to move in the same direction when communication is slow and unreliable?
Their answer was hierarchy. Eight men to a squad, one leader. Ten squads to a century. Multiple centuries to a legion. Every layer collected information, compressed it, and passed it up. Every layer received orders and passed them down.
This wasn’t a power structure. It was an information system.
“Management hierarchy isn’t a law of business. It’s a hack for human cognitive limitations.”
That sentence stopped me cold. Because it means that every performance review, every weekly standup, every all-hands deck, every reporting line—they’re all just workarounds for the fact that humans can’t process information fast enough to manage directly.
And now, AI can.
What Actually Happens When AI Handles Information Flow
Most companies are using AI the wrong way. They give everyone a copilot—sales writes emails, legal reviews contracts, HR screens resumes. This is like giving a spreadsheet to someone who’s been using a calculator. It’s better, but it’s not different.
Block and Sequoia are suggesting something else entirely. They’re arguing that AI should replace the function that hierarchy used to serve: collecting, compressing, interpreting, and routing information.
Imagine a system that reads every meeting transcript, every code commit, every customer support ticket, every sales call, every financial report. It knows who’s blocked, what’s duplicated, where the risk is building, and what opportunity is being missed. It doesn’t need a manager to tell it what’s happening. It already knows.
This is what they call the “Company World Model.”
And then there’s the “Customer World Model.” For Block, this is especially powerful. They own both sides of the transaction—Cash App on the consumer side, Square on the merchant side. They see the money flow. And as the article says, “Money is the most honest signal in the world.”
Customers lie in surveys. They ignore ads. They add things to carts and never buy. But when they spend, save, borrow, or repay—that’s the truth.
This is the real AI moat. Not the model. Not the API. The proprietary, continuous, high-quality signal that your business generates every single day.
What Happens to Middle Management?
This is the part that makes everyone nervous. The article doesn’t dance around it. It says: if the Company World Model handles information flow and alignment, then permanent middle management layers become unnecessary.
But that doesn’t mean everyone gets fired. It means the role changes.
The future organization has three kinds of people:
1. Individual Contributors (ICs). Deep experts who build and operate the capabilities, models, and interfaces. In the past, a manager gave them context. In the future, the World Model gives them context.
2. Directly Responsible Individuals (DRIs). Temporary leaders who own a specific outcome. A DRI might spend 90 days fixing a merchant churn problem, pulling resources from the model team, the lending team, and the interface team. When the problem is solved, the role dissolves.
3. Player-Coaches. People who both do the work and develop others. They’re not managers in the traditional sense—they don’t spend their days in alignment meetings and status updates. They’re close to the work, close to the people, close to the real world.
“The most dangerous place to be in an AI-native company is the information router. The safest place is the edge—where reality meets the system.”
AI can handle information. It can’t handle intuition, trust, ethics, judgment, or the nuanced feeling of a conversation that’s going sideways. That’s what humans do.
Your Company Is Either a Dinosaur or a Brain
Here’s the question that the article leaves echoing in your head:
What does your company understand that no one else does? And is that understanding getting deeper every day?
If the answer is no, AI is just a cost-cutting tool for you. You’ll fire people, improve margins, look good for a few quarters. And then you’ll get absorbed by a system that’s smarter, faster, and more connected than you.
If the answer is yes, AI doesn’t just enhance your company. It reveals what your company actually is.
For Block, it’s the “economic graph”—the real-time flow of money between consumers and merchants. For a manufacturer, it might be the causal relationship between equipment, supply chain, quality variance, and customer orders. For a retailer, it’s the dynamic model of inventory, membership, weather, and local buying habits. For an educator, it’s the understanding of how learning paths, content granularity, and feedback behavior interact.
AI competition isn’t about products. It’s not about talent. It’s about organizational intelligence. And the companies that will win are not the ones with the best chatbots. They’re the ones that have figured out how to turn their entire business into a learning system.
Three Questions You Need to Ask Right Now
Don’t ask your team to “adopt AI.” Ask yourself these three questions first:
1. Where is your company’s information bottleneck? Not the technical bottleneck. The organizational one. Who is spending 80% of their time routing information instead of creating value? That’s the first layer that should be replaced.
2. What signal do you own that no one else does? Your transaction data. Your customer support logs. Your sensor data. Your code history. If you’re not structuring this data and feeding it into a model that learns from it, you’re sitting on a gold mine that you’re not mining.
3. Are your managers information routers or judgment makers? If they’re the former, they’re going to be replaced—not by a chatbot, but by a system that does their job better. If they’re the latter, they should be transformed into player-coaches, and fast.
For two thousand years, the organization has been a tree. Roots at the bottom, trunk in the middle, canopy at the top. Information crawled up, commands crawled down.
AI-native companies don’t look like trees anymore. They look like brains. They have perception, memory, models, reasoning, and action loops. The people inside them don’t sit in boxes on an org chart. They sit at the edge, where the system meets reality.
Block and Sequoia just published a roadmap. But it’s not a roadmap for how to use AI. It’s a roadmap for how to reinvent your company before it’s too late.
Are you still building a tree?
FAQ
Q: Will AI really eliminate all middle management, or is this just hype?
A: Not all middle management—but the ones who primarily route information are toast. The managers who survive will be Player-Coaches: people who both do the work and develop talent. If your job is 80% meetings and status updates, start building a skill that AI can't replace.
Q: My company is not a tech giant like Block. Can this actually apply to us?
A: Yes, and it's more urgent for you. The 'moat' isn't a fancy AI model—it's your proprietary data. A manufacturer with sensor data, a retailer with transaction history, a clinic with patient outcomes: if you structure that signal and feed it into a learning system, you build a 'world model' that competitors can't copy. Start small, but start now.
Q: Isn't this just a fancy way to justify layoffs?
A: That's the cynical take, and it's partially true. But the article argues that the real problem isn't cutting costs—it's that the old hierarchy was a hack for human cognitive limits. If you keep the hierarchy and just add AI tools, you're building a faster horse. The companies that will win are the ones that redesign for the brain, not the tree.