Kai-Fu Lee's latest interview: These two types of people will be the most valuable in the future

43 years ago, when a young man applied for a PhD program at Carnegie Mellon University, he wrote this sentence in his application letter: AI is the last mile for humanity to recognize and understand itself.

He has kept that letter all along. Today, this sentence has a new interpretation: AI does not define meaning for humans, but allows people to re-prove their own value.

On November 4, Kai-Fu Lee appeared on the cover of *Entrepreneur* magazine. As the founder of 01.AI and Chairman of Sinovation Ventures, he has experienced the full cycle of AI development from laboratory to industrial application.

This time, what he discussed is no longer whether the model can be built, but a more pointed question: why has the business performance not improved even after the company adopted AI?

His answer is: what is lacking is not tools, but people, and two specific types of people.

I. Not just being able to use AI, but redefining how people work

Kai-Fu Lee has written three books, *AI Superpowers*, *AI 2041: Ten Visions for Our Future*, and now this one *AI Future Has Arrived*. The word "Has Arrived" in the title carries great weight.

Artificial intelligence is transforming from a technical possibility into a reality for business and society. There are three obvious trends: Models can independently complete increasingly complex tasks, technology iteration is getting faster and faster, while the cost of intelligence is decreasing.

Combined, these three factors have turned intelligence from a scarce resource into an affordable commodity.

Kai-Fu Lee believes that the depth of this transformation may exceed that of the Industrial Revolution.

The Industrial Revolution mainly liberated physical labor, while AI changes cognitive labor, reshaping the business operation mode, organizational form and competition rules of an enterprise.

But in reality, most CEOs still stick to old mindsets. The most common misjudgment Kai-Fu Lee has observed is treating AI as an efficiency tool.

True AI transformation requires redesigning the entire enterprise, just like how electricity restructured factories back in the day.

He divides the path forward for enterprises into three stages:

Digitalization means making information online and data interconnected;

Intelligentization means using data to analyze, predict and optimize;

When it comes to AI-enabled operation, AI is no longer an auxiliary tool, but an intelligent agent that can understand goals, participate in decision-making and execute tasks, which redefines the way enterprises create value.

He takes manufacturing enterprises as an example: in the digitalization stage, they master order, inventory and production progress; in the intelligentization stage, they predict demand and optimize production scheduling; when entering the AI-enabled operation stage, intelligent agents can actively identify delivery risks, propose solutions and push for implementation.

Kai-Fu Lee specially reminds that enterprises do not have to wait until digitalization is fully completed to adopt AI. The real challenge is to make AI understand the enterprise.

To deploy models in enterprises, an "enterprise map" is also needed to connect data, business relationships and processes, so that AI can work based on real business context. On this matter, he draws a line for entrepreneurs: they do not need to become algorithm experts, but must be the first person in charge of AI transformation.

II. Money spent, but no change in financial statements

No matter how good the tool is, it must ultimately be reflected in the accounts. Kai-Fu Lee's point of view is very straightforward: enterprise AI transformation must ultimately pass the test of financial statements. The amount of money spent is not an achievement, but the business value created is.

He proposed a very practical test called the quarterly earnings call test. Suppose the CFO asks on the earnings call: which financial indicator has this AI investment actually improved? Can you give a convincing answer?

Many companies cannot give a valid answer to this question.

McKinsey surveyed 1719 respondents across 97 markets, and 80% of the respondents said AI has improved their productivity. But at the enterprise level, only 37% of organizations reported that AI has made a positive contribution to earnings before interest and tax. Only 6% of "high-performance enterprises" can attribute at least 5% of their profit growth to AI.

So where should the money be invested?

Kai-Fu Lee said that at the current stage, the most worthy investment is not necessarily the scenario with the most complex technology, but the core business scenarios that occur frequently, have clear operational losses, accessible data and verifiable results.

He is optimistic about two directions: one is the full-process collaboration of sales and delivery, and the other is intelligent decision-making for production and supply chain.

III. The dividend of the AI era does not lie in smarter technology

Kai-Fu Lee has a very pointed judgment: AI will not automatically create competitive barriers.

When everyone can access powerful models, what is truly scarce is customer insight, industry knowledge, and the ability to turn technology into business results.

He compares the previous waves of technological development: the Internet reduced the cost of information dissemination and connection, the mobile Internet lowered the threshold for services to reach users, and AI reduces the cost of intelligence and professional capabilities.

In the past, when an enterprise grew larger, it usually meant more employees; in the future, there is an opportunity to serve more customers with a smaller team.

Behind this lies a more fundamental question. Kai-Fu Lee said that the first thing enterprises need to answer in the AI era is not what I can do with AI, but: When intelligence is no longer scarce, why do customers still need me?

In the past, enterprises relied on talent scale, information advantages and professional skills for survival. Now, these things that were once difficult to replicate are becoming widely available. The starting point of AI transformation is to find the advantage that is most worth amplifying, which may be unique craftsmanship, in-depth understanding of customer needs, or years of accumulated industry experience and trust.

He also pointed out an easily overlooked fact: the judgment of excellent salespeople on customer needs and the experience of senior engineers in handling anomalies were previously locked in individual minds. In the future, enterprises can use AI to precipitate them into reusable and continuously optimized organizational capabilities.

This step seems to be a technical task, but in fact it is an organizational task. To turn the skills of one person into the skills of a whole team, the obstacle to overcome is not the algorithm, but that someone is willing to hand over their most valuable experience, and someone is willing to change the original way of working.

But the premise is that you dare to change the organizational model, dare to eliminate inefficient processes, and dare to give more authorization to people who create greater value.

Because AI will amplify the real advantages of enterprises, and also expose the original inefficiencies. The data flywheel is not equal to a moat, and he has reminded this point more than once.

IV. The 2 types of people most lacking in companies

After all the previous explanations, the core point falls on people.

In the interview, Kai-Fu Lee set a measurement standard for people: a person's value in the AI era mainly depends on whether he can make judgments, collaborate with others, and take responsibility for results.

Because after AI takes over execution, thinking, delivery and other tasks one by one, the remaining part is exactly what machines cannot replace humans: judgment, trust, and the sense of responsibility after making decisions.

Following this standard, the two types of people most lacking in companies are as follows.

The first type: people who can command AI and dare to take responsibility for results.

Kai-Fu Lee repeatedly mentioned a term: DRI, which stands for Directly Responsible Individual. As AI undertakes more and more execution work, each task must have a clear directly responsible individual, who sets goals, allocates resources, commands AI, verifies results, and finally takes responsibility for the outcome.

He put it firmly: AI can assist decision-making, but someone must be responsible for the final result. Power can be delegated, but responsibility cannot be outsourced.

What does this kind of person look like?

Kai-Fu Lee gave a portrait: a compound talent who understands business, is good at leveraging AI, and has a sense of ownership and decision-making ability. In the future, an excellent DRI can lead multiple AI intelligent agents to complete the work that previously required a whole team to collaborate on.

In other words, the traditional organization where everyone only does their assigned work, has a superior to rely on, and reports problems upwards will be increasingly unsustainable. After AI takes over the execution work, what is required in return is that every link has a leader who dares to sign off on decisions.

He also shared a personal story: 01.AI uses its self-developed "Boss AI" to analyze business conditions and identify risks. The first time his own AI pointed out a problem with his work, he was not used to it. But he sticks to one point: truly valuable AI should not only cater to the boss, but dare to point out the boss's mistakes.

What needs to be built is not a culture where AI criticizes executives, but a culture where facts are above positions, people dare to question, and errors are continuously corrected.

The second type: people who can build trust and judge what is worth doing.

Kai-Fu Lee summed up this point with a sharper sentence: When intelligence is gradually commoditized, what is truly scarce for humans is the wisdom to choose the right problems.

He gave an example of young people: 01.AI allows young people to act as DRIs directly, but senior employees have a better understanding of the business. How to deal with this?

It is not to let young people replace senior employees, but to turn personal experience into organizational wisdom. Business experts define problems, pass on experience and verify results, while young employees explore AI capabilities and build intelligent agents. Senior employees understand business and young employees understand AI, which is only a complementary state in the early stage of transformation, not the end point.

What really matters is not age, but who can combine business experience with AI capabilities and take responsibility for results.

At this point, Kai-Fu Lee made his point clear. The people in the enterprise can actually be divided into two categories.

One category is worthy of key training, including people who can command AI, and people who are good at managing customers, resolving conflicts and building trust. The other category is people who only know how to sign documents, only mechanically repeat execution, and are unwilling to take the initiative to change.

But he expressed his point in a restrained way: enterprises should first provide these people with tools, training and transformation opportunities, and arrange a decent exit for those who really cannot adapt.

He added a more humanized sentence: Managers should not ask employees to hand over their business experience while preparing to lay them off.

As for middle managers, he also gave three paths: those who can coordinate business and take responsibility for results can become DRIs; those who are good at cultivating talents can become coach-type managers; those who are familiar with business processes can become designers of AI work processes.

Enterprises cannot promise that all positions will exist forever, but they must let employees see a clear growth path.

V. The first place that CEOs should adjust is not technology

When the right people are in place, the rest is the responsibility of the top leader.

Kai-Fu Lee breaks down AI transformation into three levels of a problem: strategic adjustment, organizational transformation, and business implementation. The first question CEOs need to answer is how AI is changing the rules of competition, and where to start restructuring the business operation model.

This is not a simple technical upgrade, but a productivity revolution. With the continuous decline of intelligence cost, the original cost structure, competitive barriers and business model of enterprises may all be redefined.

He gave an example: a manufacturing enterprise is under profit pressure, and the CEO needs to judge whether the problem lies in inventory, delivery, production efficiency or customer churn. Choose a scenario that can truly improve profits, deploy AI in it, and adjust the organizational division of labor and responsibility mechanism at the same time.

So his suggestion is only one action: Choose a high-value scenario that affects business results, personally find the first batch of DRIs, and give full authorization.

The most important job for CEOs in the AI era is not to manage AI projects, but to redesign how enterprises create value.

He also redefined an excellent CEO: not necessarily the person who masters the most information, but the person who is most willing to accept challenges, constantly revise judgments, and take responsibility for the final results. To do this well, they need to complete three things: set the direction, define the boundary, and take responsibility.

Even for meetings, he drew a line: meetings for progress tracking and information transmission can be handed over to AI; meetings involving strategic trade-offs, major decisions and complex problem discussions are worth investing more time in.

To judge which enterprise can continue to lead, he looks at three capabilities: strategic determination, organizational learning capability, and the capability of rapid trial and error and large-scale implementation.

The winners in the future will not necessarily be the first to use AI, but the enterprises that can continuously turn technological dividends into organizational capabilities and business results.

Technological progress is an opportunity for everyone, and the speed of organizational evolution determines how far an enterprise can go.

VI. What can be learned is not the trend, but the specific actions

After explaining all the principles, Kai-Fu Lee gave actionable solutions that can be directly referenced.

He pointed out the most common mistake enterprises make now: letting powerful AI do trivial and irrelevant work.

A truly valuable pilot should start from core business problems, enter with a small incision, and verify with large value. There is only one principle: deduce AI applications from business results, not find usage scenarios from technical capabilities.

He gave a "first action" for companies of different sizes respectively.

For a company with 300 people, prioritize improving sales conversion. Let AI participate in inquiry analysis, product matching and quotation preparation, with sales personnel reviewing key commitments, focusing on response time and order conversion rate.

For an enterprise with 3000 people, focus on production and delivery. Connect order, inventory, production scheduling and quality information, let AI identify delivery risks in advance, and check whether on-time delivery rate, production cost and quality have improved.

For an enterprise with 30000 people, carry out cross-departmental business collaboration.