HomeArticle

AI has reshaped management assumptions, and organizations are different.

穆胜2025-12-02 11:54
In the AI era, middle management is being streamlined, and organizations are becoming extremely flat.

Recently, large enterprises have a common direction in organizational adjustment - the bosses bypass the middle - management and directly take over various functional lines or business areas, re - entering the centralized power mode. Either they directly serve as department heads or establish committees, which essentially marginalize the department heads. Especially in highly competitive industries such as new energy vehicles, the Internet, and hard technology, typical examples are companies like Xiaomi and XPeng.

A basic organizational common sense is that centralization can achieve unified overall command, bringing low - cost and efficiency for simple tasks; while decentralization can achieve flexible local responses, bringing innovation and efficiency for complex tasks. Logically, in these highly competitive industries that require flexibility, they should not move towards centralization. However, I want to say that considering the major technological variable of AI, this form of centralization might be correct. Of course, unless the bosses suddenly have a "stubborn mind" and want to establish a feudal dynasty.

The Middle - Management Trap: The Birth of Bureaucracy

Let's start with why the middle - management exists. Understanding their origin can help us know why they may gradually disappear.

In the traditional pyramid - shaped organization (Hierarchy), senior management needs the middle - management to manage front - line employees. The main reason is that complex fields require division of labor. Once the division of labor becomes detailed, information becomes extremely complex. Once it exceeds a certain threshold, senior management cannot handle it.

Since senior management cannot handle a large amount of information, the middle - management is needed to perform "information filtering" and "execution supervision". In other words, senior management sets standards for information upload and decision execution, and then the middle - management acts as a bridge between the top and the bottom. As a result, senior management can see filtered information and ensure that commands are executed. This is the origin of authorization.

The more detailed the division of labor, the more middle - management is needed because each middle - manager's management span (i.e., ability) is limited. As the number of middle - managers increases, senior management is also limited by their management span. So, they have to authorize more and more. When authorization still cannot solve the problem, new levels have to be created to assist in management. Thus, the middle - management keeps expanding, giving rise to the MOM (Manager of Manager) level. In a broad sense, except for the top - level bosses and front - line workers, all "intermediate levels" in an organization are middle - management.

The question is, can the middle - management really improve organizational efficiency? The reality is that the middle - management is in an awkward position with a large scope for slacking off. In the traditional pyramid - shaped organization, they are very likely to become a "hindrance" rather than a "help".

Extremely speaking, if they only act as "PPT aggregators" in their positions, along with some impressive reporting routines such as underlying logic, strategies, links, and feelings... How can senior management discover their lack of ability and inaction?

As the organization grows in scale, there are more and more MOMs, and this "phony" atmosphere spreads rapidly. The reason is simple. If subordinates adopt this style, when superiors receive their reported materials, it's like getting a bad hand of cards, making it difficult to achieve results. Over time, superiors are also likely to degenerate. If superiors are like this and don't seriously check subordinates' work, why should subordinates work hard instead of slacking off? In fact, this is why the middle - management in many companies deteriorates collectively.

Most of the time, the slacking off of the middle - management is not the most terrifying thing. The most terrifying thing is their parochialism. When they prioritize maintaining their "territories", their professional abilities will rapidly decline. In this state, the middle - management can only act as information passers - on and cannot create professional value. Even when passing on information, they may hide things for personal gain. At this time, they become complete "obstacles".

The "big - company diseases" such as departmental barriers, insulation layers, and process bottlenecks emerge according to the above logic. This is what I call the "middle - management trap".

AI Tools: Transforming the Organization

In the past, bosses understood the above principles, but they had no choice but to fall into the "middle - management trap" because they had no other solutions for organizational design. So, they could only be "on guard against and make use of" the middle - management.

Repeatedly advocating entrepreneurship and professional ethics, and conducting cultural inspiration are all aimed at setting the behavioral bottom - line for the middle - management. Various cadre training programs and training classes named after bosses are all for cultivating some "capable confidants". Even so, among the cadres cultivated by bosses, there are always some laggards, pulling down the lower limit of the organization's efficiency. In most cases, this is not a problem of ability but of willingness. After all, human nature is hard to control.

However, with the arrival of AI, bosses may no longer need to solve this difficult problem. Two trends will lead to drastic changes in the organizational model, enabling bosses to rely less on the middle - management.

Firstly, senior management can implement broader and more in - depth control through AI tools.

Management scholar Graicunas' research through mathematical models shows that as the number of subordinates increases, the number of potential relationships that managers need to handle grows exponentially. Therefore, the span of control is generally considered to have a limit. After research by several management scholars and exploration by practitioners, the principle that "a person cannot directly manage more than seven people" has become an established rule in the era of industrial economy.

However, in the digital and intelligent era, enterprises have gone through a step - by - step evolution from standardization → dataization → digitization → digital intelligence. The relationships that managers need to handle have been greatly simplified, and the span of control they can manage naturally becomes larger and larger.

Imagine an ideal scenario. When the work of all departments is standardized and turned into data, senior management can naturally implement more efficient management through this data. In the past, they had to go to the front - line in person, wait for data reports passed on by the middle - management, and even conduct complex analyses to discover the actual operation of the business. But now, data is not uploaded in the form of reports but is real - time online hot data. Moreover, as long as the model is mature, AI can automatically analyze the objective operation of various departments, functions, and businesses. So, senior management only needs to make decisions based on the accurate information provided by AI. They can manage not only widely but also deeply.

Secondly, front - line employees can greatly improve their delivery levels through AI tools, naturally reducing the difficulty of supervision.

Take the "white - collar work" in middle - and back - office functional departments as an example. If an employee's text - output level (in the form of proposals, reports, systems, special analyses, etc.) suddenly improves significantly, rest assured that they must be using AI.

Previously, it was very difficult to train a professional with good text - output ability. Even if superiors taught them hand - in - hand, it still required the trainees to have the willingness and potential. So, the success rate of training was not high. Moreover, after being trained, the trainees may not think it is the credit of the company or their superiors. They are likely to look for better opportunities elsewhere and may leave with resentment.

Since there are such convenient and low - cost AI tools, why "endure unnecessary hardships" to train people and bear the blame of being a "heartless capitalist"? Let all white - collar workers use AI to ensure a basic level of text - output work. On this basis, those who want to add creativity can do so freely. Isn't this much easier? Do we still need the middle - management to summarize and review? Isn't it more convenient to directly summarize this high - quality delivery by AI and send it to senior management?

If a large number of middle - management positions disappear in a company, the pyramid structure will start to collapse, and the organization will naturally become extremely flat. In form, isn't this a platform - based organization?

The Disappearance of the Middle - Management: Who Will Remain?

In essence, the positioning of a considerable number of middle - managers is naturally "anti - AI". Due to their lack of creativity and only acting as "PPT aggregators", they are very afraid that their territories will be penetrated. So, they are committed to making work non - standardized to highlight the value of their "manual control".

In fact, they are not only anti - AI but also anti - digitalization. They have also opposed process - orientation (standardization) and enterprise - wide integrated data reports (management reports). The reasons are simple, such as "our field is very special and difficult to standardize", "we have a large historical debt and no data foundation", "our employees' abilities are insufficient and cannot handle standardized tools", "the solutions provided by consulting companies are not practical"... After contacting many companies, those middle - managers who are used to "slacking off" and "occupying territories" really have no new excuses, just the same old sayings.

The problem is that no one can stop the technological trend because the cost - performance and return - on - investment brought by new technologies are far ahead and cannot be rejected in the business world. In the early days of the Industrial Revolution, some people smashed the spinning jenny, but they could not stop the establishment of industrial automated production lines. During the Internet era, some people refused e - commerce, but they could not stop online shopping from becoming a trend. Today, if someone resists AI, can they succeed?

Considering the technological trend of AI, future organizations will definitely be extremely flat, and the middle - management will inevitably be greatly reduced. Even without this technological trend, in the current economic downturn, bosses who have repeatedly fallen into the "middle - management trap" are almost out of patience.

A large Internet company is a typical example of this trend. Under the pressure of improving labor efficiency, they laid off a large number of P8 positions and let P9s directly manage P7s. As a result, they found that the effect was surprisingly good. According to insiders, P9s come up with ideas, and P7s implement them. Many P8s were just information passers - on. Canceling this level naturally improved the organization's efficiency.

Similar cases have also occurred more than once in the client companies served by Musheng Consulting. Bosses originally hoped that their key employees would perform their duties and improve the efficiency of various functions or business lines. However, unexpectedly, these key employees established their own territories. So, the bosses took over forcefully, eliminated the obstacles, accelerated decision - making, and the organization's efficiency became higher.

The reality is that the slacking off of the middle - management may greatly exceed the bosses' expectations, especially in large or super - large enterprises. Some people may question how bosses could not know. Well, they know but don't know the full extent. They know there is such a situation but don't know how bad it is. There is always information asymmetry between superiors and subordinates. It is very easy for subordinates to pretend to be hard - working while slacking off and seeking personal gain. So, when the external economic environment brings pressure and AI provides the conditions, when bosses make a decisive move against the middle - management, they suddenly find so many redundant employees.

It seems that the reduction of the middle - management has become a foregone conclusion in this round of organizational evolution. Of course, I'm not saying that the middle - management will "disappear", but that they will be greatly "compressed". In other words, those middle - managers who only act as information passers - on will be squeezed out of the organization by AI. Overall, in the future organization, only two types of middle - managers can truly remain:

One is those with "top - level design thinking". They have wild "imagination", can build frameworks, and explore possibilities among different models. In whichever functional area they are, that area will continuously generate new ideas, new data, and decision - making indicators, thus creating a "computational advantage" over competitors. In other words, they make the company "smarter" in this area compared to competitors, making the company more like an "intelligent organization".

The other is those with "off - line scenario value". They have a down - to - earth "implementation ability", are good at implementing plans, and can make the organization's operations enter the right data track through interpersonal communication, creating professional results. They are the link between AI and human brains. Based on their understanding of business scenarios, they can bring their professional skills into play and let AI amplify their professional value. If an organization is compared to a living organism, the former type of people are the brain, and they are the nerves. They make the power of AI penetrate every corner of the organization.

This article is from the WeChat official account "Musheng's Office" (ID: hrm - yun), author: Mu Sheng. Republished by 36Kr with permission.