Summary of Key Points
This article discusses the impact of AI on traditional corporate structures: Not only does AI make skills such as programming more accessible (non-technical personnel can also write code), but it also points to a deeper issue—the hierarchical structure that has supported companies for two thousand years may be disrupted by AI’s information routing capabilities. Jack Dorsey, founder of Block/Twitter, proposes using “AI agents” to replace traditional hierarchies for information coordination and to redefine corporate architecture. At the same time, traditional business barriers (such as per-person pricing, functional barriers, and high technology migration costs) are collapsing, with the new barrier being a “deep understanding of the real world.” Companies are also experiencing a K-shaped division—efficient AI-driven companies versus inefficient traditional ones—and middle managers face more severe identity crises than programmers.
Detailed Analysis
1. Why do companies need hierarchies? A tradition that has lasted for two thousand years
Have you ever wondered why there are managers, directors, and vice presidents in companies? This wasn’t something that was “conceived” deliberately; it was a necessity that emerged over time. Two thousand years ago, Roman armies needed to coordinate the actions of 5,000 soldiers. Communication relied on shouting, and one person could manage at most 3 to 8 subordinates (this is called the “span of control”), so they had to be grouped in layers: 8 soldiers → 1 squad leader, 10 squad leaders → 1 centurion, all the way up to the army commander. This structure was essentially a set of “information transmission rules”—information from below was aggregated upwards, and commands were passed downwards.
The concept of middle management also originated from this need: After Prussia’s defeat by Napoleon, they established a “staff department” that handled information and coordinated troops without going to the front line. This was not because middle managers were particularly effective, but because it wasn’t possible to expect every leader to be a genius. Later, American railroads used this logic to manage their operations (to prevent train collisions), Taylor’s scientific management improved division of labor, Spotify’s team models, and Zappos’ experiments with management-free structures… However, no matter how things changed, as organizations grew larger, hierarchies were necessary because there were no other tools to replace human information transmission.
In short, hierarchies were not the “optimal solution”; they were just the “only workable one” over the past two thousand years.
2. Could AI replace hierarchies with “agents”? What is Dorsey’s radical idea?
Jack Dorsey’s article raises the question: If the core function of hierarchies is to “transmit information and coordinate resources,” then do they still need to exist if AI can do these tasks?
He designed a new architecture for Block:
- Functionality Layer: Basic services such as payments and loans (like Lego blocks).
- World Model Layer: Internal (what’s happening within the company, whether resources are sufficient) and external (e.g., changes in a restaurant’s cash flow).
- Intelligence Layer: Automatically combines functions based on models—for example, if a restaurant’s cash flow is about to dry up, AI would automatically arrange a short-term loan without the need for a product manager to write a request.
- Interface Layer: The app that users see (e.g., the Cash App).
In this architecture, people only need to play three roles:
1. Experts in basic functions.
2. People responsible for solving cross-departmental problems (replaced every 90 days).
3. “Player-coaches” who both do tasks and lead teams—there are no permanent middle managers, as AI has taken over information coordination.
Dorsey’s core idea is that in the past, it was “people with intelligence transmitting information through hierarchies”; now, “systems have intelligence, and people make decisions at the front line.” In the future, companies will be like octopuses—the central body is the AI agent, and the tentacles are the people.
3. The old barriers are collapsing! What are the new barriers in the AI era?
What did traditional companies use to protect themselves from competitors? For example, SaaS companies charged per employee, offered more features (that competitors didn’t have), and made technology migration difficult (high costs). But with AI, these no longer matter:
- Per-person pricing: 20 people using AI can handle the work of 60 people—why should customers pay for extra seats?
- Feature richness: You can take a screenshot of a competitor’s interface, clone it in minutes, and even customize it.
- Technology migration costs: AI can automatically transfer entire databases without human supervision.
So what are the new barriers? Dorsey says: “A deep understanding of the real world, which you continuously deepen.” For example, Block’s “economic graph” uses transaction data from millions of merchants and consumers to understand each person’s real economic behavior (money doesn’t lie; consumption records are more reliable than surveys). The better the model, the more transactions, the more accurate the data, and the better the model—this is compound growth.
In simple terms: In the past, companies competed based on what they built (features, systems); now, they compete based on what they understand (the laws of the real world).
4. Are companies dividing into two types? How scary is the K-shaped division and self-devouring cycle?
Companies are splitting into two camps:
- Model-first companies: Fewer employees, higher efficiency, using AI for more tasks, and extremely low costs.
- Traditional companies: Many hierarchies, slow decision-making, with profits squeezed by the former.
The more terrifying aspect is the “self-devouring cycle”: Traditional companies have lower profits → lay off employees → the laid-off employees learn about AI → use AI to start their own companies that attack their former employers → the original company’s profits drop even further → more layoffs. Sometimes, even without layoffs, ambitious employees leave (staying in a company resistant to AI means wasting their future) and use AI to compete with their former employer.
This cycle is accelerating: It used to take years for a competitor to emerge and attack; now, it might happen in just months.
5. It’s not just programmers who are at risk! Middle managers face an identity crisis
Many think that AI only takes jobs away from programmers, but middle managers are actually in even more danger. If your job involves “transmitting information, coordinating resources, and aligning priorities,” and AI can do these tasks faster and more accurately, what’s your value?
AI is also dismantling two things:
- Skill barriers: Skills like programming and design, which require years of learning, are being leveled out by tools (non-technical personnel can write code).
- Organizational hierarchies: Information transmission no longer relies on layers of people; systems handle it directly.
In the future, valuable people will not be those who “control key points in processes” (like department managers), but those who connect AI systems with real-world needs. The stronger the systems, the more people need to go to the front line to address real problems.
This is an identity crisis for many middle managers: The management frameworks they’ve built over decades may suddenly become useless.
Final Summary
AI is not just a simple cost-cutting tool; it is disrupting the corporate structures that have existed for two thousand years. Hierarchies will gradually disappear, and the new organizational form will be “AI agents + front-line personnel.” Companies and individuals that survive will rely on a “deep understanding of the real world.”