An organization must always ensure that every individual has a proper place to belong.
When AI evolves from "assisting humans in completing work" to "undertaking work tasks directly", we tend to equate the speed of technological evolution directly with the speed of organizational change: as if accessing large models, flattening hierarchies, and deploying Agents can complete an organizational revolution.
But in reality, the impact of AI on organizations and people is not that fast in time, not uniform across scenarios, and not applicable to all business scenarios.
Some people are already rewriting workflows with Agents, while some companies have not even completed standardization and processization;
Tasks in some positions are being rapidly transformed and replaced, while some work still relies heavily on on-site experience and comprehensive judgment of humans;
The most urgent problem for some enterprises is AI transformation, while for others, the more severe challenges at the moment are still cutthroat industry competition and breaking through operational bottlenecks.
What enterprises are facing is no longer just a tool upgrade, but a rethinking of organizations and people.
AI can take over tasks one by one, but why can it not replace a complete human being?
When individual efficiency is rapidly improved, why have organizations not disappeared, but instead re-collaborated at a higher dimension?
In the latest issue of Hundun online course, Cong Longfeng, Founder of Cong Feng Consulting and expert in organizational management, does not rush to give a set of standard answers for AI-native organizations. Instead, he first leads us to find the coordinate: return to the technical path first to clarify the facts that have happened and the predictable trends; then return to the organizational principles to discuss the change and constancy of organizational management; finally return to human beings themselves to discuss how to prepare talents for the future AI era.
He will guide you to go beyond the superficial hype of "OPC", "eliminating middle managers" and "Agent collaboration", and re-see the things that truly determine how far an enterprise can go: goals, responsibilities, relationships, judgment, mental strength, and the desire of people to accomplish great things together.
AI Transformation Starts with Local Optimization
In fact, in a sense, this round of AI wave only has two critical moments.
The first one was November 2022, when ChatGPT was launched to the public. The second moment was February 2025, when Claude Code was released.
However, most of us perceive these changes with delay and unevenness.
For example, I did not feel the arrival of the AI era until the second half of 2023, and the popularity of Deepseek during the 2025 Spring Festival made more general public aware of AI.
But when I led a team to Silicon Valley in January 2025, engineers in Silicon Valley clearly told me that the speed of Chinese enterprises' catch-up has indeed exceeded their expectations, and DeepSeek has made another innovation with low cost. But at the same time, DeepSeek has not surpassed Silicon Valley in technology. It was also at this time that I first learned about vibe coding and Anthropic. However, most of our compatriots did not perceive this until March 2026 through the "lobster-raising" trend.
I even remember that in April and May 2024, many partners around me were still asking "what is AI", but after I went to Silicon Valley at that time, I found that engineers were already talking about Agents. This year, most of us have just begun to know what Agent means.
So there is no doubt that with the arrival of Claude Code this year, AI has begun to truly enter the workflow and impact organizations. But I think what we can already see at the moment is that many definite, repeatable, regular and sufficiently convergent tasks have been gradually taken over by AI. Previously, it could only reach 60 points, but now it has reached 80 or even 90 points.
For example, there is a phenomenon this year called "distilling colleagues". That is, import the working materials of colleagues, including Feishu messages, documents, emails, screenshots, etc., to create a skill that can replace their work. But on the other hand, we will find that some work is not easy to be replaced by AI in a short period of time.
First, the work is not standardized or repeatable, so it cannot be accurately defined and patterned. Second, the work requires more comprehensive judgment, rather than single-task, mechanical execution. Third, the work requires more interaction with specific people, rather than program-like input and output.
If your work meets these three conditions, you will find that AI cannot distill you, because the real core competence lies in you. So I say, if your work can be "distilled" so easily, does it just mean that your work is too unsubstantial?
Because only water can be distilled away. If your work is like protein or fat, it can only be scorched. We need to find the most valuable part in our work, which is the value that a person needs to continuously deepen in the AI era.
However, this "trend of AI replacing humans" has been established in a sense, so how fast will this trend develop? I don't think it will be that fast.
Because today's large language models are still built on generative pre-training and Transformer architecture. They learn patterns from a large amount of data, and then generate results based on context. The continuous improvement of capabilities does not mean that the nature of technology has been completely changed.
Current AI still has three defects: "unstable thinking, insufficient deep reasoning ability, and no support for continuous learning". It acts like a genius in some tasks, but makes stupid mistakes in others, which is called "zigzag intelligence".
This year I have seen some experts confidently discussing management in the AI era, as if AI can already replace humans, and all work will be done by Agents in the future. But if you look back at its technical path, our current situation is more like a phased progress achieved by a certain technical route in the process of AI development.
Then based on these known realities, how does AI affect management? How to realize the transformation of organizational construction in the AI era?
My judgment is: we should not be overly optimistic in the short term, because technology is far from reaching its peak. But the medium-term trend is very clear, and the long-term final outcome is still uncertain. Therefore, it is more suitable to carry out local optimization at present, rather than global advancement.
Many companies are destined to act blindly in 2026, because it is very likely that they will have to start all over again by this time next year. So my reminder for building AI organizations is to complete "standardization, processization, dataization, and knowledgeization" first, and then move on to "intelligentization". If you skip the first four stages and directly talk about intelligentization, what AI takes over may just be a mess.
At the same time, I also want to remind that the applicable industries of AI and the applicable stages of enterprises need to be accurately defined. Because a company usually goes through four periods: pattern exploration period, scale expansion period, mature operation period and business transformation period.
Current AI is most suitable for the mature operation period. Because the business model has been verified, a large number of definite, repeatable and convergent work can be handed over to AI for continuous optimization. But in the exploration period, transformation period, and even expansion period, a lot of work still needs to be done by humans. Therefore, not every company should copy the same AI transformation plan.
Then from a longer-term perspective, how should enterprises position their own pace on the AI track?
After visiting Silicon Valley again in July this year, I confirmed some judgments and facts. My experience has three points in general:
First, AI is not a project, AI is a journey. If you take AI as a journey, your mentality will be much more peaceful.
Second, I think the first phase of the large model war has ended, and it will enter a period of platform waiting for breakthrough. It is still unclear where the algorithm of the next stage will be. But the public domain data has been exhausted and has begun to deteriorate. There has been a consensus on embodied robots that Chinese enterprises are better at making hardware, while American enterprises are better at developing algorithms.
Third, when the next intelligent breakthrough really comes, all companies will become Harness companies that call for the integration of a large number of business processes. At that time, the chapter of organizational management needs to be rewritten.
Organizations Still Exist After OPC
From the perspective of current technology, one point is very clear: AI can eliminate tasks, but cannot eliminate humans in the short term. Let AI do what belongs to AI, and let humans do things that only humans can do.
Then where is the future of organizations? We must first return to the basic concepts, basic principles and basic methods of organizational management, and talk about what will not change in organizational management.
In this field, there is a theoretical master that cannot be avoided, named Barnard. The first question he asked when studying organizations was: why do humans have to be organized?
The importance of this question in the current AI era is that if human society is not organized, OPC will be realized —— human society will only have two layers: individuals and society, and no organizations are needed in between.
But Barnard said that every formal organization arises from informal organizations. The occurrence of informal cooperation depends on two prerequisites:
The first is that every individual has free will, motivation and purpose, and wants to pursue a larger goal; the second is that when every individual pursues a larger goal, sooner or later he will encounter the limitations of his own ability, resources and experience, and then instinctively want to find another person to partner with.
For example, a driver is driving down a street and finds the road blocked by a huge stone, which he cannot move by himself, so he will instinctively wait for a second driver to pass by, and then they move the stone away together.
So according to Barnard's premise, if one day humans no longer want to pursue larger goals, they can choose not to be organized. Once he wants to pursue a larger goal that he cannot complete by himself, he will find more people to partner with, which is the origin of organizations.
Back to the present, everyone still needs teams, but AI allows everyone to collaborate at a higher dimension. In the past, maybe everyone could only get 60 points, but with AI, everyone can reach 80 points. So in this sense, I can say with certainty that OPC is mostly a transitional phenomenon. After the overall level is improved, everyone will carry out team collaboration at a higher dimension.
But Barnard also put forward a conclusion: the first problem of organizational management is the inconsistency between personal goals and organizational goals. Your goal is not equal to my goal, your pursuit is not equal to my aspiration.
Take the previous case of the two drivers again. One may be going to a meeting, and the other is going to pick up his child from school. Their long-term goals are different, but at the moment of cooperation, their common goal is to move the stone away.
For another example, the entire team of Anthropic split off from OpenAI. To put it bluntly, the two companies once had old grudges. So even in the most cutting-edge AI organizations, human grievances, interests and emotions cannot be eliminated by technology.
Then we can understand the three elements of organization proposed by Barnard: common goals, willingness to collaborate, and information communication, which constitute the foundation for the existence and operation of formal organizations. These three conditions are still not outdated in the AI era.
What AI can most easily transform is the efficiency logic, but it cannot automatically generate the willingness to contribute, cannot make personal goals naturally consistent, and cannot ensure that people are willing to tell the truth.
Organization is a complex problem, and it is impossible to form natural synergy, because it has three facets at the same time. It is an economic organization that follows the logic of efficiency; it is also a social organization with emotions and relationships; it is also a small society composed of different positions, where games exist.
So AI-native organizations are destined not to be applicable to all enterprises. Preparing organizations for the AI era cannot be achieved overnight. I think that in the face of this new thing, we really need a very cautious attitude, and we need to do exploration, investigation, abstraction and even imagination in a down-to-earth manner.
When it comes to organizational issues, there are three benchmark enterprises that cannot be bypassed —— Huawei, Alibaba and ByteDance. Why are their organizational forms so different? Because organizational design is not only the organization of people, but first of all the organization of work and business.
The business interdependence of Huawei, Alibaba and ByteDance is different. Complex corps operations rely more on processes, while businesses with stronger individual attributes are more likely to emphasize Context not Control and super individuals.
Generally speaking, a real organization always grows between the objective constraints of business and the subjective mental cognition of the founder. The founder's mind will also affect the organization for a long time. But the past textbooks on organizational design mostly ignored the subjectivity of the founder. So I say that an important starting point of entrepreneurial organizational design is to "amplify the advantages of the founder".
Organizations are being Rewritten in the AI Era
On the other hand, based on the unchanging organizational principles, our organizational aesthetics has indeed begun to change.
The 7 founders of Anthropic have exactly the same equity, and the 60 executives of NVIDIA have exactly the same compensation, forming a community with a shared future and a common cause at a higher dimension.
So I think future start-ups will likely rely more and more on collective wisdom and close collaboration, rather than lone heroes or people who just follow orders. An important reason why an organization can amplify personal advantages is the psychological safety of the team.
So for the new organizational paradigm in the AI era, I have listed three inequalities at present:
The first is that talents are more important than positions.
The discipline of organizational management has always been pursuing the replaceability of people —— writing JD, building competency models, and checking whether this person has done similar work before. But the more standardized and clearly defined the position is, the easier it is to be reconstructed by AI. So experience is not that important. What is more important is whether people's general quality, underlying quality and life energy are sufficiently full.
Last year, a founder in Chengdu asked me how to view people with good underlying qualities, and I put forward three "ones": first-class standards, quick to understand, and one step ahead in thinking. Later, a friend who is an investor added another point: strong enough small sample abstraction ability —— being able to discover universal rules after experiencing two things shows that this person is really smart.
The second is that teams are more important than organizations.
Organizations will definitely become smaller and continue to return to the core. When the team is small enough, many problems that used to be solved by systems and hierarchies can return to direct management, and the standard process system will return to the natural sense of responsibility between people.
It is not OPC that changes the world, but small teams. OpenAI, Anthropic and DeepSeek all relied on high-density core teams at the key breakthrough stage. Future small teams need three roles: good player, excellent individual; good team leader, people who can lead the team; great organizer, real organizer.
Therefore, when medium and large enterprises transform, they need to first find the core teams, special forces and commando teams that can really take action; without teams, organizational-level transformation can hardly happen.
The third is that form is more important than structure.
Structure emphasizes stability, stereotype and boundary, while form is flowing. In the past, we pursued the invariance of organizations, making organizations fixed in terms of structure, system and institution; but now, we find that making organizations fluid to generate changes and innovations is more important.
Therefore, the core construction principle of future organizations is centered on information decision-making, rather than driven by execution efficiency. The general direction of future organizations will be as clear as a crystal ball, which everyone can see when they look up. And it needs to be sufficiently diverse, inclusive and dynamic to generate new things.
Moreover, I have felt more and more in the past two years that one thing is particularly important. That is, when the top leader of the company thinks he is more important than the company, the organization can no longer grow. When the top leader is both strong and sincere, and still has a sense of dissatisfaction and modesty, the company will continue to grow.
From this, I thought of a possible minimal organizational model in the future: CEO+CMO/CPO+CTO, or CEO+CMO+CPO/CTO.
In fact, there are only two most attractive types of organizations in the business world: brand/marketing-oriented organizations and technology/innovation-oriented organizations.
The core role most needed by brand/marketing-oriented organizations is like the False 9 in football tactics —— the combination of CMO and CPO, who understands both the market and products, and connects the entire offensive line like Messi.
For technology/innovation-oriented organizations, such as current AI organizations or companies making embodied robots, the core role most needed is like the Deep-lying Playmaker in football tactics —— the combination of CTO and CPO, who understands technology and is familiar with products, and activates the whole game like Rodri.