Dialog mit Mu Sheng: KI "dringt" in Unternehmen ein – wer wird am Ende bleiben?

Artificial Intelligence (AI) is penetrating companies at an unprecedented speed.

On the one hand, tech giants in Silicon Valley are constantly laying off employees and investing more budgets in large models, computing power, and AI infrastructure. On the other hand, more and more companies are examining their employees' AI usage and token consumption, and even linking AI application with performance evaluation.

However, a confusing reality is: Employees are using DeepSeek, Kimi, and Yuanbao, but the number of meetings doesn't decrease; Organizations are integrating large models, but the approval processes don't disappear; The capabilities of AI are constantly rising, but the overall efficiency of companies doesn't increase synchronously.

Where lies the problem? Is the organization truly reshaped by AI, or is it just a new technology layered on top of the old organization? Who will be replaced by AI first, ordinary employees or middle - level managers? Do tech giants like Tencent, Alibaba, and ByteDance already have the organizational form suitable for the AI era? And are the hot - discussed "one - man companies" the future mainstream form or just a technological utopia?

With these questions, the Xuebao Finance Society had an in - depth exchange with Dr. Mu Sheng, a well - known Chinese management expert and the founder of "Mu Sheng Consulting".

In his view, the biggest miscalculation of companies currently is to regard AI as a "panacea" that can penetrate the organization. Improving individual productivity doesn't automatically mean improving organizational productivity. The most important factors determining a company's competitiveness are, in order, organization, people, data, and models.

In a conversation with the Xuebao Finance Society, Mu Sheng outlined a future vision of an "agent organization": a few human elites, a large number of AI employees, a multi - centric collaborative network, and business units that infinitely form around customers.

In contrast to the view that AI will completely surpass humans, he firmly believes: Regardless of technological development, creativity, complex judgment ability, and empathy remain unique human capabilities.

These capabilities, which cannot be described by algorithms, form a "God's secret" that AI can hardly penetrate.

Below is the transcription of the conversation between Mu Sheng and the Xuebao Finance Society:

01

Xuebao Finance Society: AI not only has many impacts on business models and products, but more and more people are also starting to be interested in its impacts on organizations and employees. Now, "agent organizations" are being talked about everywhere. What exactly is it, and what are its characteristics?

Mu Sheng: An agent organization is a complex structure. I've simply summarized its characteristics into three points:

First: "Few humans, many AIs". In an agent organization, AI takes over all standardized tasks. The number of human employees is relatively small, and only the top elites and employees at the lowest level remain.

Second: "Multi - centric dynamic network". In this type of organization, there are a large number of AIs as nodes, and there is abundant cooperation between "human - AI" and "AI - AI". Thanks to the standardized communication protocols of API interfaces and motivating economic contracts, the cooperation runs smoothly.

Third: "Customer - centered and infinitely malleable". Within this organization, clusters of AIs and the infrastructure for AI generation are rarely visible from the outside. However, there are many "small business units" in the market that form around customers and use the resources of the entire company to realize products, services, and solutions.

02

Xuebao Finance Society: Currently, almost all companies are discussing AI. Some companies even make AI application an obligatory KPI and examine token consumption and the degree of AI replacement of work processes. However, why don't most companies really become stronger?

Mu Sheng: These companies have a superficial understanding of AI technology and organization. They believe that "replacing humans with AI" can improve productivity. An example illustrates this well: After Thomas Edison popularized electricity in the 1880s, entrepreneurs began to replace steam engines with electric motors. However, in the following nearly 30 years, there was no productivity boom.

Besides the backward electricity transmission technology and the high costs of equipment replacement, the bigger reason was that the factory layouts were centralized before. That is, a large steam engine and a "main shaft" ran through the factory hall to drive each machine. When companies replaced the steam engine with a powerful generator, the factory's operating model remained unchanged, and the efficiency improvement was hardly noticeable.

Later, the improvement of productivity lay in two points: First, the production method changed from "main - shaft drive" to "unit drive". Each machine got its own motor, which freed up the physical arrangement of the factory. Second, the process was redesigned, and "assembly - line production" was introduced. That is, the work processes were divided along the material flow and connected by conveyor belts, so that products could automatically flow to workers, which reduced transportation and waiting times.

What I want to say is that the fact that the technology doesn't achieve the expected effect is due to the delay of the organization. However, most company owners don't seem to understand this. They place too much hope in the technology and expect the technology to penetrate the organization.

In fact, this situation also appeared in the digital era. When digital technology emerged, many companies dreamed of a rapid digital transformation. How many companies have actually been successful?

03

Xuebao Finance Society: In many companies, employees are already using DeepSeek, Kimi, and Yuanbao for work. Why doesn't the number of meetings decrease, why do the approval processes not disappear, and why aren't the hierarchies reduced? Microsoft and OpenAI are heavily investing in AI, but why doesn't the organizational efficiency increase synchronously?

Mu Sheng: Essentially, business leaders don't want to change the organization. They regard AI as a panacea and hope that employees will use AI and the organization will be directly changed.

But individual productivity is not the same as organizational productivity. For example, programmers can write code faster with the help of AI, and human resources personnel can speed up candidate selection with intelligent recruiting agents. However, these results don't necessarily contribute to an improvement in customer experience and business value.

Apparently, the efficiency of individual parts increases, but the company's structure and process remain unchanged. The emerging productivity gets stuck in the old nodes (approvals, reviews, meetings, etc.). This congestion not only keeps the overall efficiency unchanged but can even reduce the overall efficiency due to the accumulation of tasks. Let's take code development as an example again. If programmers write more code with AI tools, but the code - review process is still backward, it leads to an accumulation of code and extends the time until publication.

04

Xuebao Finance Society: Due to AI, some companies have carried out layoffs. Who will be replaced by AI first, ordinary employees or middle - level managers?

Mu Sheng: This is an interesting question. According to the usual assumption, business leaders want to replace ordinary employees with AI. However, the correct assumption is that middle - level managers should be replaced first. Middle - level managers are called MOM (manager of manager). Their main function is to mediate information between the upper and lower levels. This ability is the easiest to be replaced by AI. The capabilities of AI can fully cover the steps of information collection, comprehensive decision - making, command issuance, and result review. In fact, MOMs in Silicon Valley are the main targets of layoffs.

There is another reason: If MOMs are not replaced, it's difficult to change the organization. The departments where middle - level managers work regard them as their territory, and they have a thousand reasons to block AI. The history of the "resistance to the spinning jenny" could repeat itself.

Naturally, department heads with professional expertise are more difficult to replace. However, their work is less about directly leading people and more about training large models in their professional fields and planning the cooperation between humans and AI. They are more like experts than professional managers.

I want to point out that when everyone gives the same answer to a question, either the question is too simple or people are thinking too superficially.

05

Xuebao Finance Society: When AI starts to conduct analyses, coordination, and make decisions, what will be the most important work of humans? Does a company still need so many deputy general managers? Does it only need the CEO?

Mu Sheng: Deputy general managers are elites in different fields, and elites are difficult to replace. If we refer to the entire group of human employees, there are some types of work that AI cannot achieve.

First, original creativity. That is, to define a new business path and explore new customer experiences. We can elevate it a bit and say that it's about writing a "new worldview" in the business field. For example, a few years ago, no one would have thought that the tea - drinking sector could become an independent business path, and only a few people could understand how a cup of milk tea could bring a "little happiness" to customers.

Second, system architecture. That is, to develop business models, create new products, services, and solutions, including innovative methods in specific professional fields. Creativity must be implemented into a concrete work system through these measures.

Third, human touch. On the one hand, there are always parts that cannot be covered by digitalization. Human employees must act as informants and "map" reality into the digital world. On the other hand, customers are real people, and human employees must interact with them at critical points, understand their emotions, and ensure the customer experience.

Fourth, supervision and takeover. AI makes decisions based on algorithms, which can sometimes lead to unfair decisions. For example, layoffs and performance evaluations based on algorithms can be discriminatory, and human ethical judgments are needed. Moreover, the algorithm of AI is constantly evolving, and it cannot be guaranteed that the best result will be achieved. In critical situations, humans must intervene and take control.

06

Xuebao Finance Society: This brings me to a question that has been discussed for a long time but seemingly has no final answer: Do humans have capabilities that cannot be replaced by AI? One view is that AI has already completely surpassed humans.

Mu Sheng: I'm firmly convinced that the elites among human employees still have value in the AI era and are at the top of value creation. These people have three capabilities that cannot be replaced by AI:

First, creativity, the ability to break the "cognitive barrier". Creative people can connect one thing to another unrelated thing, like how Steve Jobs created the smartphone in his mind. Second, the ability to process complex information, which is more of an intuition, the ability to find the key answer from a large amount of data and information. Third, empathy, that is, the ability to understand the emotions of other people. Emotions are complex and difficult - to - code signals that contain a lot of implicit knowledge (Tacit knowledge). AI cannot process this information.

Essentially, these three capabilities cannot be described by algorithms. I also call these parts the "God's secret" that AI cannot penetrate.

07

Xuebao Finance Society: Let's focus on the organization again. The most important organizational invention in the industrial era was the department system. Will the AI era end the department system? Will the marketing departments, finance departments, human resources departments, and brand departments in today's companies disappear like the former typists and telegraph operators?

Mu Sheng: This concerns the description of the agent organization.

This organizational model formally looks like a "large platform + small business units". The small business units use the resources of the entire company and act like special units, while the large platform still includes functions such as procurement, production, marketing, finance, human resources, and legal affairs. As long as there are companies, these functions will objectively exist. For example, although an electric car doesn't have an internal combustion engine, it at least has a core drive unit.

But will these functions exist in the form of departments? I think so, because their work is quite complex and requires cooperation to achieve results. The products of these departments are large models, but these models don't come out of nowhere. The departments have to invest a lot of work to abstract the models. Even if the models are called by front - line departments through API interfaces, human supervision and control are still needed to ensure the application effect.

What I see is the trend that on the one hand, these functions output large models and continuously improve them, and on the other hand, by shifting BPs (Business Partners) to the front - line departments, the capabilities of these units are strengthened. And employees in the departments and the shifted BPs can exchange their positions.

Naturally, these departments don't have to be as bulky as before. A few expert elites are enough.

08 

Xuebao Finance Society: Is such an agent organization large or small? In the industrial era, companies became larger and larger. Will the AI era make companies smaller again, and will many companies return to the state of "small business owners"?