McKinsey's Latest Report: The 3 Most Valuable Types of People in the Next 5 Years

In the past year, the assistance of AI to individuals has been remarkably evident.

Coding has become faster, proposal development has accelerated, and data retrieval has become swifter. An individual's output density has been multiplied several times over.

However, at the organizational level, we always feel held back. A proposal can be crafted in ten minutes, but it takes three to four days to implement.

Actually, what holds us back is not the inadequacy of tools, but an unacknowledged hidden cost in traditional organizations: Alignment Tax.

Training, meetings, approvals, and reports do not create value. They merely bridge the gaps between people.

When the carrier of alignment shifts from "people" to "codes" and then to "AI Agents", hierarchies disappear, power dynamics change, and organizations shift from "managing heads" to "managing capabilities".

In this article today, we present several specific solutions to restructure the existing organizational logic. We hope it will be helpful to you.

I. The Hidden Cost of Traditional Organizations: Alignment

The structure of traditional organizations consists of hierarchies, power centers, processes, and sequential collaborations. This structure has a hidden cost that has never been properly recognized, known as Alignment Cost.

People in an organization vary in their skills, understanding, and stances. To ensure that everyone moves in the same direction, the organization needs to carry out a large number of "management actions".

For example, training is used to transfer knowledge from those with stronger skills to those with weaker ones;

Meetings are held to align the goals and priorities of different departments;

Approvals are required to ensure that each decision - making point is confirmed by a sufficient number of people; reports are submitted to keep higher - level management informed about what lower - level employees are doing;

360 - degree peer evaluations are conducted to allow colleagues to recognize each other's contributions.

These actions do not create value. They bridge the gaps between people. They are the "Alignment Tax" of traditional organizations.

McKinsey mentioned in its newly released report "HR Monitor 2026: A turning point for the people function": Only 11% of organizations conduct strategic workforce planning for more than three years. 85% stay at operational headcount planning. The anchor point of organizational thinking is "how many people" rather than "what capabilities".

46% of organizations plan according to specific roles/positions, and only 22% adopt a skills - oriented approach. They still rely on the "job" mindset. 23% of employees lack the skills required for their current positions, and 22% doubt their ability to remain competent in five years. The speed of skills migration far exceeds the organization's response speed.

These data point to the same conclusion: The thinking inertia of traditional organizations still lies in "managing heads" rather than "managing capabilities" and "managing alignment efficiency".

This thinking results in the biggest hidden cost in traditional organizations being alignment cost and management friction, not labor cost. These costs are invisible, but they consume most of the organization's energy. Experiences like "a proposal takes ten minutes to develop but three to four days to implement" are manifestations of such energy consumption.

Once the alignment cost decreases, a chain reaction will be triggered.

II. What Happens to an Organization When Alignment Cost Decreases?

A decrease in alignment cost is not an isolated event. It triggers a complete causal chain: A decrease in alignment cost leads to the disappearance of hierarchies, a shift in the source of power, and a change in organizational form.

1. Three Progressive Levels of Alignment Cost

Alignment cost is going through three progressive levels. Each level is an order of magnitude lower than the previous one. Each time the level drops, the organizational form changes.

Level One: Person - to - Person Alignment, Extremely High Cost

When a person learns a good method, they need to train others. Training requires time, venues, instructors, and post - training assessments. And most of the content is forgotten after three months.

When a department sets a new goal, it needs to hold meetings for alignment. Preparing for the meeting, attending it, organizing meeting minutes, and following up on implementation result in a very low proportion of actual behavior change.

When a decision needs approval, each node in the approval chain needs to understand, judge, and sign. Each node is waiting.

The cost is extremely high because the carrier of alignment is "people". There are losses, delays, and misunderstandings in the transfer between people. When three people discuss a proposal in a meeting, it's highly likely that they will have different understandings of the proposal after the meeting. People are not precise transfer media.

Level Two: Person - to - Thing - to - Person Alignment, Extremely Low Cost

DeepSeek's alignment method is sharing, not training. When a person creates a good prompt or skill, they can directly share it with the team. Others can use it immediately without training, meetings, or waiting for approval.

The same goes for papers. DeepSeek frequently opens sources and publishes papers. Liang Wenfeng said, "For technicians, being followed is a very rewarding experience."

This may sound a bit idealistic, but it's actually very practical. When your achievements are followed by global peers, you won't want to leave the environment that enables you to produce good results.

Knowledge sharing is the most effective talent retention strategy.

The cost is extremely low because the alignment carrier has changed from "people" to "codes, texts, prompts". Codes can be copied infinitely, texts can be spread instantaneously, and prompts can be used immediately. Once written, they can be reused infinitely. The transfer loss between people approaches zero.

This is the core insight of the alignment method in AI - native organizations. Alignment is "capability alignment". I encode my capabilities into a shareable format, and you can just use it.

There's no need to hold meetings to discuss "how to understand this proposal" because the proposal itself is an executable code or prompt, and the understanding is already built into the format.

Level Three: Person - to - AI - to - Person Alignment, Cost Approaching Zero

McKinsey proposed the Agentic HR operation model in its report: AI Agents autonomously plan, make decisions, and execute multi - step workflows. Agents take over the sequential coordination function, not just "assist".

Previously, five nodes in the approval chain could now be handled by one Agent in a few seconds for information verification, rule comparison, anomaly marking, and conclusion output.

Previously, three meetings were required for cross - departmental alignment. Now, Agents can directly retrieve the data and progress of each team and generate an orchestration plan.

The alignment carrier has been upgraded from "shareable texts" to "self - executing Agents". People don't even need to "obtain" others' skills. Agents will automatically orchestrate.

There is a sentence in the McKinsey report: "Organizational flattening has been predicted many times in the past, but this time it's different: AI Agents replace specific coordination functions, not just free up management time."

In the past, flattening was predicted because "management theory says hierarchies are bad". This time, it's because "AI Agents take over the reason for the existence of hierarchies, which is the coordination function". It's a technology - driven logic switch, not theory - driven.

When alignment no longer requires "person - to - person" transfer, the reason for the existence of hierarchies disappears.

Hierarchies exist because the person - to - person alignment cost is extremely high. Someone needs to stand at a higher level to summarize information, make directional decisions, and coordinate progress. This person is the manager.

The core function of a manager is "coordination", not "creation". When Agents take over the coordination function, managers no longer need to be in hierarchical positions. The coordination function has a new executor, and the hierarchical position is replaced.

2. Disappearance of Hierarchies: Shift in the Source of Power

After hierarchies disappear, power doesn't disappear, but the source of power shifts.

In traditional organizations, power comes from position. When you are in a certain position, you have the power to arrange schedules, conduct initial performance evaluations, and allocate resources. This is identity - based power. Power is tied to the position, not to the output.

When a person changes positions, the power follows. A person with extremely high output but not in the right position has nothing to do with the power.

Tencent is conducting an experiment to shift from identity - based power to result - based power.

The WeChat Business Group has abolished the team - leader responsibility system and implemented the project - based system. Power exists as long as the project exists and disappears when the project ends. Power has shifted from being "tied to the position" to being "tied to the output".

Managers at the director level and above are evaluated every six months. The principle of "no merit means fault" has been written into the criteria. Managers can no longer act as "message - passers", simply translating information from above to below and reporting from below to above, acting as an information transfer station in the middle.

Henry Mintzberg's research confirms this: The core value of truly effective managers lies in mastering key information, connecting key resources, and making effective judgments in uncertainty.

Hierarchies give titles, but these three things need to be actually done.

The Hunyuan team has abolished the mid - year evaluation because AI R & D cannot be managed in the same way as mature businesses. It requires tolerance for trial - and - error and long - term investment.

The five - level performance evaluation system has been changed to a three - level system to reduce ambiguity and concentrate resources on those with the highest output. The difference in year - end bonuses between Outstanding and Underperformance can be up to ten times.

The 360 - degree peer evaluation has been reduced from twice a year to once. Promotion authority has been decentralized, and formal reviews have been replaced by project performance results.

These changes at Tencent point in the same direction: The organization no longer pays for "presence", but only for "output". In the past, a large amount of management power was a by - product of organizational expansion, not a proof of individual ability.

DeepSeek takes it a step further, pushing result - based power forward to capability - based power.

Liang Wenfeng doesn't pre - assign tasks. Roles emerge naturally from idea collisions. Whoever is most passionate and has the best ideas about a problem takes the lead.

Result - based power still has the "project" as an anchor. Liang Wenfeng doesn't even pre - set projects. Power comes from "understanding and passion for the problem".

The prerequisite is a high density of talents. High enough that each person has their own drive, and high enough that "evaluating people together" can determine compatibility.

Liang Wenfeng's only criterion for selecting people is passion and curiosity. Passion brings self - drive, resilience, and long - termism, eliminating the need for KPIs.

Elon Musk is also removing hierarchies, but in a more radical way. He cuts position - based power and replaces it with strong - man governance. One person (Elon Musk himself) becomes the decision - making center of the entire organization. The hierarchies are cut, but the power is not dispersed; it is concentrated in one person. The coordination function still relies on people, either Musk himself or a few people he trusts.

Musk has freed up management time, but the coordination function has not changed its executor.

So, Musk can serve as a control group. Because "removing hierarchies" and "being AI - native" are two different things. Hierarchies can be removed through strong - man governance, but that's not AI - native.

An organization can be non - hierarchical but still centralized. Musk's organization is in this form.

3. Shift in the Source of Power: Change in Organizational Form

As power shifts from position to output to capability, the organizational form changes accordingly.

Organizations with identity - based power adopt a hierarchical system. The power center drives alignment through training and meetings;

Organizations with result - based power adopt a project - based system. Power dissipates with the end of the project. Alignment is achieved through goals and KPIs;

Organizations with capability - based power are in a self - emerging state. Roles emerge naturally, and alignment is achieved through skill and prompt sharing.

This causal chain shows one thing: AI - native organizations are not an upgraded version of "flattening" or a radical version of "removing middle - management". The collaborative infrastructure has shifted from "people" to "AI", representing a completely different organizational logic.

High alignment cost requires hierarchies to force alignment, resulting in identity - based power;

Low alignment cost eliminates the reason for the existence of hierarchies, and power shifts from position to output and capability;

When alignment cost approaches zero, Agents take over coordination, hierarchies naturally dissolve, and the organization enters a natural state.

III. Why Do Most Companies Get Stuck in "AI Augmentation"?

1. From "Augmentation" to "Reinvention"

McKinsey summarizes organizational transformation into two routes:

One is Augmentation: Embed AI into existing processes to gain incremental efficiency without changing the organizational logic. This corresponds to "traditional organizations plus AI tools", and 80% of organizations stay at this stage.

The other is Reinvention: Redesign workflows, role divisions, and decision - making mechanisms from first principles. Rewrite the organizational logic. This corresponds to "AI - native organizations", and approximately 0% of organizations have reached this stage.