Li Yanhong is working on the "intelligent organization", while Yang Yuanqing is a big fan of the "AI-native organization"?
At the Baidu World Conference held yesterday (November 12, 2026), Robin Li released a number of Baidu's thoughts on "Intelligent Organization", and also invited Yang Yuanqing from Lenovo to launch an in-depth dialogue on this theme.
From the outside perspective, Baidu is an Internet enterprise while Lenovo is a traditional enterprise, but both are sprinting at full speed in the direction of "enterprise AI transformation". After practicing at the forefront of the trend, the two leaders of the enterprises do have unique insights into AI, and they both agree that — AI brings not only the improvement of individual efficiency, but also a profound organizational revolution.
In terms of how to promote organizational reform in the AI era, the two enterprises have their own discourse systems and have been continuously speaking out actively.
We noticed that Robin Li mentioned "AI Native Organization" at the 2025 Baidu World Conference, making him one of the first entrepreneurs in China to adopt this concept. However, at this year's Baidu World Conference, he began to refer to this type of organization as "Intelligent Organization" (which is also the widely used concept of "Agent Organization").
Lenovo, on the other hand, jointly released the AI Native Organization Transformation Guide with the Business School of Renmin University of China in September 2026, and even proposed a five-level maturity model for this type of organization, becoming an active practitioner and advocate of the "AI Native Organization" concept in China. Obviously, they chose a conceptual expression different from that of Baidu.
After observing a large number of practical cases at home and abroad, Mu Sheng Consulting found a more interesting phenomenon — at present, Chinese enterprises prefer to use the concept of "AI Native Organization", while benchmark enterprises abroad prefer to use the concept of "Agent Organization".
We finally found that the choice of which concept to adopt is by no means a dispute over the right to speak, but a real difference in practical orientation.
01 Agent Organization
The concept of "Agent Organization" does not have a unique "first proposer", because it actually points to meanings in two different fields, with completely different origin times and proposers.
As early as December 2004, Bryan Horling and Victor Lesser from the University of Massachusetts Amherst first proposed the concept of "Multi-Agent Organization Paradigm" in a paper. It should be noted that the "Agent" in this paper is a software entity based on rules or symbolic logic, which is different from the current agents using large language models, but the two share the underlying idea of "collaboration of autonomous entities". Their many thoughts on agent organization are of enlightening significance.
After more than ten years, "Agent Organization" has gradually become a relatively mature professional term in the field of computer science, referring to a multi-agent system composed of elements such as agents, tasks and rules. As a standard term, this concept was recognized by the China National Committee for Terms in Sciences and Technologies, and was included in the Computer Science and Technology Terms (3rd Edition) and officially announced in 2018.
Interestingly, this concept did not become popular after being proposed twice. The reason is that the AI technology at that time was not yet popularized, and the agents were not mature enough to play an important role in the organization. The real starting point for the popularity of this concept was 2025.
In September 2025, McKinsey & Company released a report titled Agent Organization: The Next Paradigm in the AI Era, which formally proposed and systematically elaborated this concept. In this report, McKinsey defines it as a new paradigm: Humans work side by side with virtual and physical AI agents on a large scale in a way of nearly zero marginal cost. Their statement of this concept is more from the perspective of organization, but it also triggered thinking from various perspectives under this concept.
In October 2025, the Furu Wei team from Microsoft Research published a paper, which directly defined "Agent Organization" as a new operation paradigm of AI from the technical perspective of large language models: Agents solve complex problems through collaboration and concurrency, thus surpassing individual intelligence. At the same time, scholars at home and abroad have also put forward different views from business perspectives such as process reengineering, organizational behavior and product management.
In the first half of 2026, Dr. Mu Sheng systematically presented his and his team's views on agent organization through articles, interviews and other forms. From the perspective of organizational reform, he gave the judgment that "Agent Organization = Platform-based Organization + AI Technology". In August 2026, he published the work Agent Organization, which systematically put forward the core concept of "Agent Organization" for the first time, built a complete theoretical system for organizational evolution in the AI era, and further precipitated and systematized the concept.
He believes that, Agent Organization is an organizational mode that is customer-centric, takes AI as the core productivity, and is based on a comprehensive platform (Platform) composed of elements such as computing power, algorithms (such as open-source or closed-source large models), data, tools (such as low-code development platforms) and rules (such as human-machine collaboration charter) provided by the enterprise. It takes market signals as the first driving force, and a small number of elite human employees create and manage a large number of Functional Agents to carry out networked, autonomous, approximately standardized human-machine collaboration, so as to efficiently meet the needs of existing market customers and continuously expand the future market boundary.
The contribution of Dr. Mu Sheng lies in developing "Agent Organization" from a vague vision into a management theoretical framework with clear characteristics and internal logic. It not only emphasizes that organizational reconstruction must precede technology application, but also integrates in-depth understanding of AI technology into organizational design, which provides important theoretical reference and practical guidance for people in the fields of "management" and "technology" to understand organizational transformation in the AI era. In other words, he and his team provide a complete set of management grammar from "what it is" to "how to transform", extending the concept of "Agent Organization" from the "concept level" to the "methodology".
02 AI Native Organization
At the same time, we also noticed that many enterprises have chosen another even more popular concept — AI Native Organization.
Objectively speaking, the concept of "AI Native Organization" does not have a single "inventor". It is more like a product that was jointly promoted and rapidly evolved by the business community, consulting industry, media and academia between 2025 and 2026. This concept originated in Silicon Valley, and quickly became popular in China once introduced, experiencing an extremely explosive process.
On November 6, 2025, Lin Wei, a columnist of FT Chinese, cited the radical statement of Silicon Valley investor Jeremiah Owyang in his article titled AI-Native: The New Enterprise Paradigm Being Defined in Silicon Valley: "The operation logic of AI-Native enterprises is: look for AI first, if you can't find it, build AI yourself, and recruit people at last."
In fact, the type of enterprises he described are baby companies established during the period of AI technology maturity from 2024 to 2025. These emerging enterprises are completely built based on the "AI-first" concept. In the white paper cooperated by Jeremiah Owyang and other institutions, he even described "AI Native" as an organizational culture, that is, "AI as the reason for existence". Driven by Lin Wei's article, Chinese entrepreneurs quickly accepted this trendy concept.
On November 13, 2025, at the Baidu World Conference of that year, Robin Li also used this concept. He put forward the direction of "internalizing AI capabilities and building a native AI organization", emphasizing the internalization of AI capabilities into a "native" capability of the enterprise, making it a productivity rather than a pure cost.
In April 2026, Zhou Yunjie, Chairman of the Board of Directors of Haier Group, formally put forward the Chinese name of "AI Native Organization" at an internal meeting, which was a key node for this specific expression to be clearly "put forward" in the business community. In the same April, Professor Chen Chunhua and Li Zhiyong co-authored the book AI Native Organization, and the concept began to be systematized and spread beyond the industry. In September of the same year, the Business School of Renmin University of China and Lenovo jointly released the AI Native Organization Transformation Guide, proposing the L1-L5 maturity model, which systematized and instrumentalized the existing concepts.
Of course, mainstream foreign research institutions have also adopted this concept. In June 2026, Hyunjin Kim and Rembrand Koning from Harvard Business School completed their working paper AI-Native Firms. Based on the data of start-ups supported by Y Combinator and venture capital, they strictly defined the category of "AI Native Enterprise (not AI Native Organization, but can be regarded as a variant of the same concept)", and empirically tested their organizational characteristics: the number of personnel is about 25% less than that of non-AI native companies in the same industry, the hierarchy is flatter (about half a rank), and the per capita value creation is higher.
Summarizing the different viewpoints above, in short, the concept of "AI Native Organization" can be summarized — it is an organizational form built based on AI. In this kind of organization, AI is the production base rather than an auxiliary tool. Once it is shut down, the core business will be directly interrupted. Figuratively speaking, it is more like the organization is doing "AI +" rather than "+AI". Naturally, considering that AI has become another protagonist besides human beings, the organization's processes, departments, posts, meeting coordination forms will be redesigned, and even mechanisms such as assessment and incentive will undergo fundamental changes. In short, this concept emphasizes an ultimate ideal state where AI and humans fully collaborate.
03 Different Choices of Chinese and Foreign Enterprises
In the discourse system of domestic enterprises, the voice and adoption rate of "AI Native Organization" are indeed significantly higher than that of "Agent Organization", but this is more of a strategic conceptual choice rather than a strict distinction between the essential differences between the two. To sum up, the popularity of AI Native Organization has the following major reasons:
First, the concept is more "safe" and avoids the association of "machines replacing humans".
The literal meaning of "Agent Organization" easily directs the focus to "organizations composed of agents". In the current public opinion environment, this will naturally trigger anxiety and resistance to "machines replacing humans". "AI Native", by contrast, is a more neutral narrative of technological evolution. It does not presuppose whether the main body of the organization is silicon-based or carbon-based, but emphasizes "redesigning everything in the way of AI", and people can still play an important role in such organizations. This leaves hope for "people", and there is less resistance in internal communication and external publicity.
Second, the narrative is more grand, which fits the strategic positioning of the "top leader project".
"AI Native Organization" is packaged as the ultimate vision related to the reshaping of enterprise genes, which gives a sense of magnificence to a large number of traditional enterprises. Just imagine, if a large traditional factory can "transgenically" transform itself into an "AI Native" state in the AI era, what a remarkable achievement it is. Bosses like this kind of ultimate and grand narrative, while "Agent Organization" sounds too down-to-earth, not "strategic enough", and its expression does not have such strong appeal.
Third, the framework is more "compatible", allowing enterprises to advance and retreat freely and make self-explanations in organizational reform.
"AI Native" provides a stepped maturity framework from L1 to L5. Enterprises can claim that they are "marching towards AI Native" without immediately answering the sharp question of "whether the decision-making power has been handed over to AI". In contrast, "Agent Organization" in academic discussions often directly points to the substantive reconstruction of human-machine rights and responsibilities, with higher implementation thresholds and accountability pressure. By choosing "AI Native Organization", enterprises take the initiative to define and explain their own progress; by choosing "Agent Organization", they are more likely to be examined by the outside world with strict organizational reform standards. In fact, this is not only the choice of enterprise bosses, but also the choice of organizational reform executors (such as CHO). When it comes to the level of Agent Organization, they face higher risk of "being held accountable". Choosing the concept of "AI Native Organization" is more like a "conspiracy" between the boss and the reform executors.
Fourth, leading institutions make the right to speak more concentrated, forming a reinforcing cycle of communication.
The term "AI Native Organization" has been jointly endorsed by leading academic and industrial institutions such as the Business School of Renmin University of China, Lenovo and Haier, and continuously exported through forms such as Transformation Guide, award selection and summit forums. However, "Agent Organization" lacks the same level of organized promotion power in China, and it is not until now that enterprises like Baidu have joined the communication. The first-mover advantage of the right to speak makes later comers more inclined to use the widely known label of "AI Native" when describing their own practices.
But interestingly, unlike the domestic business community's preference for the narrative of "AI Native Organization", "Agent Organization/Enterprise" is the more dominant concept in current academic research and business practice abroad.
Abroad, the narrative of Agent Organization is completed by mainstream consulting companies and research institutions. McKinsey, Deloitte, Bain, PwC and Gartner all adopt the core framework of Agent Organization to discuss how Agent Organization can reshape the operation mode and revenue structure. Even if "AI Native Organization/Enterprise" is used in research, it is only used as an empirical analysis category, and there has not been a concentrated upsurge of discussion like that in China in rigorous academic literature. On the business practice side, enterprises generally accept such a narrative framework, and follow a number of methodologies to try to reach the state of Agent Organization. They are more accustomed to splitting the concept into deliverable and measurable modules. It can be said that abroad, since consulting, research and practice all emphasize the "operationalization" of the discourse system, Agent Organization is the absolute dominant framework.
04 Don't Shout Slogans, Do More Practical Work
It is worth noting that although domestic enterprises use "AI Native Organization" as a strategic banner or communication strategy, the executive layer still widely uses "Agent Organization" methods such as agents, process reengineering, organizational reshaping, three-platform reconstruction and incentive innovation. Domestic enterprises are faced with internal mobilization, policy narratives, capital market and employee anxiety, so they need a more neutral and grand label, but they also urgently need landing methods.
We cannot deny that domestic enterprises will gradually deepen their practices while using "AI Native Organization" for grand narratives. However, this narrative method may mislead practitioners. After all, the traction of the concept determines where the final practice will go. Several years ago, almost all large domestic enterprises (non-baby enterprises) were keen on grand concepts such as "organizational innovation", "Internet +" and related concepts, but in the end, all of them came to nothing. Returning to the field of human-machine collaboration, if the enterprise itself is not "AI Native", but puts forward the ultimate state of "AI Native Organization", and only takes the usage and coverage of AI agents as the measurement standard, it will inevitably lead to an embarrassing situation of spoiling things through excessive eagerness for quick success, rushing ahead recklessly, and replacing management with assessment. Rather than that, it is better to make clear what to change at each step and take every step steadily, which is exactly the reason why Mu Sheng Consulting chooses to narrate with "Agent Organization".
We should not ignore that the driving mechanism of domestic enterprises also plays a key role. The transformation of domestic enterprises is significantly affected by policies and capital. The "15th Five-Year Plan" elevates AI to a medium and long-term strategy, and the special action of the State Council has set clear penetration targets. This top-down promotion can quickly mobilize resources, but it is also easy to spawn projects oriented to "making statements" and "demonstration", making these experimental fields further bypass the deep-water area of organizational reform, and eventually leading to the situation that people can only stand by and sigh at the state of "AI Native Organization". For central SOEs and local SOEs, this trap is particularly obvious.
A view from Ms. Lou Jun, Partner of Mu Sheng Consulting and Dean of Mu Sheng Consulting Research Institute, is worth alerting — At present, the AI organizational practices of domestic enterprises are generally stuck in the "local reconstruction at the process and post level", and there are no publicly credible cases of "fundamental reform at the organizational model level". Those enterprises that claim to be "AI Native Organizations" actually open up several "special zones" within the traditional organization, and try out new human-machine division of labor and decision-making methods in the special zones, while outside the special zones, the old organization operates as usual. Using the end-state label of "AI Native Organization" to name these local experiments is essentially a conceptual "overstatement".