This is the scarcest capability in the next decade.

Now, AI can generate ten sets of growth strategies in seconds, drastically shorten R&D cycles, and even participate in strategic simulation.

For modern enterprises, what is truly scarce is no longer whether they have access to AI tools, but who will define problems, who will select solutions, and who will bear the consequences of decisions.

Truly leading enterprises are restructuring their decision-making power: which issues can be handed over to machines, which must be finalized by humans, how to roll back when anomalies occur, and how organizations should re-divide responsibilities.

Wang Sai, Chairman of the Board of Directors of Notesman PPE (Politics, Economics, Philosophy) Academy and a renowned CEO consultant, puts forward an insight: "Strategies are infinite, while decisions are scarce." When solution generation becomes cheap, judgment, trade-offs, and accountability become even more valuable.

In today's article, Wang Sai explores a question that every entrepreneur cannot avoid: when AI becomes a new participant in decision-making, how can enterprises upgrade it from an auxiliary feature to an operating system that drives growth, organizational evolution, and strategic choices?

We hope today's content will bring you inspiration.

1. Business Decision-Making Power Is Undergoing Restructuring

1. AI Begins to Participate in Enterprise Decision-Making

In May this year, US-based company Coinbase did something that kept all Silicon Valley executives awake at night.

They laid off 700 employees, not because of poor performance or a shortage of funds - the company had abundant cash flow on its books.

The reason for the layoffs was: AI had replaced their jobs.

Coinbase built a system, a "think tank" composed of four AI roles.

When facing major strategic decisions, these four AIs first "debate" internally: the analyst conducts initial evaluations, the explorer searches internal knowledge bases for supporting evidence, and the synthesizer integrates all opinions into a proposal.

The most powerful role is the fourth one, called the "Opponent", which is specifically responsible for identifying flaws: "What risks have you overlooked in your solution?" "Is this assumption really valid?" "If the market changes, will this set of measures still work?"

Guess what happened? The issues identified by AI were ones that human executives could not come up with even after three days of meetings.

Even more astonishing is a prediction from Gartner: by 2027, half of all global business decisions will be enhanced or automated through AI.

So today I want to discuss with you: when AI starts making decisions for you, what core competitiveness remains for our enterprises and ourselves?

My answer is: Decision-making - not the ability to "make decisions", but the ability to "design who makes decisions".

In today's era of artificial intelligence, AI can provide you with an infinite supply of strategies and respond in seconds. For humans and human organizations, the ability to evaluate and select which strategy proposed by AI to adopt has become particularly critical, even life-or-death important.

2. The New Scarce Skill in the AI Era: The Ability to Ask Good Questions and Evaluate Solutions

In Silicon Valley, two of the most important new positions have emerged in the past three months: one is called the "Chief Question Officer" (some companies like DeepMind call it the "Chief Philosophy Officer"), and the other is the "Chief Evaluation Officer".

One is responsible for asking excellent questions beyond efficiency considerations, rooted in human-centric values; the other makes high-quality, high-value selections from the collaborative outputs of artificial intelligence.

This is the value of decision-making today, which I call "Strategies are infinite, while decisions are scarce".

3. Three Paths for AI to Transform Business

The transformation that AI is bringing to business can be viewed from three dimensions: Innovation, Operations, and Decision-Making.

The first two layers are "using AI to do the same things faster", while the third layer is "AI making entirely new things possible". Most enterprises have only achieved the first and second layers, but the third layer is where the real value lies.

First Layer: Artificial Intelligence + R&D

Andrej Karpathy (Notesman Note: Former core member of OpenAI's founding team, Senior Director at Tesla, who recently joined Anthropic), a legendary figure in the AI field, shared a detail at a Sequoia Capital summit in April this year.

He said that last December was a watershed moment - he suddenly realized that he no longer needed to modify code written by AI. He remarked, "I can't even remember the last time I corrected it."

In the past, writing software required you to type code line by line, telling the computer to do this first and then that.

Now you describe your requirements in natural language, and AI generates them directly for you. You no longer need to write a bunch of scripts to install software - just throw your requirements to AI, which will read your machine environment, find paths on its own, and resolve dependencies automatically.

As a result, the very act of writing code has been redefined.

It is even more dramatic in the field of scientific discovery. The 2024 Nobel Prize in Chemistry was awarded to AlphaFold, whose AI model shortened the protein structure prediction cycle from decades to just a few days.

Google has developed an even more powerful new system that can independently propose scientific hypotheses and design verification solutions - it does not just assist scientists in experiments, but acts as a scientist itself.

The characteristics of this layer are: It does not happen frequently, but each occurrence brings enormous value.

In the past, you needed to maintain a large team for R&D. Now, AI can run through all relevant global papers, patents, and competitor data in a few hours and deliver a research report. For industries like pharmaceuticals, materials, and fashion, this means opportunities for overtaking on a curve.

However, the key insight is: Breakthroughs by AI in the R&D field do not put scientists out of work, but allow scientists to return to their core tasks.

In the past, a PhD student's prime time was spent running data and adjusting parameters. Now AI has taken over all these tedious and laborious tasks, freeing humans to think about more fundamental issues.

Second Layer: Artificial Intelligence + Operations

If the R&D layer is like an "atomic bomb", the operations layer is like a "machine gun" - it occurs frequently, brings moderate value each time, but accumulates to a very considerable total.

Zara's fast-fashion myth was once a classic case: it only took two weeks from design to shelf placement. Traditional fashion brands like Nike needed a year and a half from design to market, having to predict fashion trends and prepare inventory in advance.

Zara's approach was completely different: its store managers transmitted customer feedback back to headquarters in real time every day, such as "Many customers tried this floral dress but few bought it - why?" Designers, purchasers, and analysts sat in the same large office and made quick adjustments based on this data.

If products sold well, they increased orders; if they did not sell, they stopped production. Their inventory turnover rate was three times faster than peers, and their unsold rate was only half the industry average.

However, SHEIN, which is also in the fast-fashion industry, took a completely different approach: it relied on AI to accelerate the entire supply chain, capturing trending keywords from social media in real time, and using AI to generate design drafts to eliminate manual brainstorming;

It implemented small-batch fast response, with AI automatically triggering reorders within 24 hours based on click conversion data;

The SCM system dynamically assigned orders, synchronizing hit product traffic with production, shortening the traditional multi-month operations cycle to just a few days to achieve ultra-fast response. SHEIN's approach let AI directly lead humans in operations, with humans stepping back to execution positions and AI taking center stage.

This also tells us one thing: operational efficiency competition is not about process speed, but about the rotational speed of the decision-making loop.

"I predict what will be popular in half a year" has given way to "I already know what is selling and what is not selling right now".

Starbucks' "Digital Flywheel" is another example.

This coffee giant has nearly 30 million active mobile app users.

Its algorithm associates your consumption data - what coffee you bought at what time - with the weather, holidays, and even your location, to push "timely and location-appropriate" recommendations to you.

It pushes an espresso on your way to work in the morning, and a hot chocolate on rainy days.

H World Group's intelligent system allows you to complete a room renewal just by speaking a sentence.

The logic behind these examples is the same: AI does not replace humans, but turns the knowledge within the organization into capabilities for the entire team.

In this scenario, AI reshapes the high-frequency, repetitive operational links of the past, engraving operational best practices into the memory of artificial intelligence, letting AI drive the organization.

Third Layer: Artificial Intelligence + Decision-Making

No matter how powerful the first two layers are, their essence is still "using AI to do the same things". The third layer is completely different - AI brings a qualitative change to the very act of "decision-making".

What does a qualitative change mean? There are two stories.

The first is Netflix's $1 million high-stakes bet.

In 2006, Netflix was still a DVD rental company.

It did a very interesting thing: offered a $1 million reward, inviting top talents from around the world to improve its recommendation algorithm. Whoever could increase recommendation accuracy by 10% would take home the $1 million.

Three years later, a team achieved this. But Netflix did not stop there - it discovered that the true value of the recommendation algorithm went far beyond helping users find content more easily, as it reshaped Netflix's content decision-making logic.

In the past, Hollywood made movies relying on the intuition of directors and producers: "I think this theme will be a hit."

Netflix used data to speak: it analyzed user viewing behavior and found that audiences who loved the original UK series *House of Cards* also enjoyed movies starring Kevin Spacey and works directed by David Fincher.

So Netflix invested $100 million to get these three people to produce the US version of *House of Cards*.

This work became Netflix's first hit original series, marking Netflix's transformation from a "content distributor" to a "content producer".

Netflix used its recommendation algorithm to change what users watched on one hand, and what it produced on the other. This is the qualitative change in decision-making.

The second story is the strategic intuition behind Amazon Prime.

In 2005, Amazon was just an e-commerce platform selling books. An engineer proposed an idea to Jeff Bezos: launch a membership system where users pay a fixed annual fee to enjoy unlimited free two-day shipping.

Many executives opposed it, because the logistics cost per order was more than a dozen dollars, and if members abused the service, the company would go bankrupt.

Bezos overruled the objections and said: "If customers love Prime, demand will rise. When demand rises, we can build more logistics centers to spread out the costs."

As a result, Prime membership grew from zero to 200 million. More importantly, the membership system turned uncertain consumption into definite user relationships.

The most brilliant part of Bezos' approach is that he did not stop at this "certainty".

The behavioral data of Prime members continuously expanded the boundaries of new businesses: Prime Video, Prime Music, cloud computing. Amazon later became the world's largest cloud service provider, with its foundation rooted in the data insights from Prime membership.

What do these two stories have to do with AI?

The Netflix story tells us: AI can turn intuition-based decision-making into data-driven decision-making.

In the past, only a genius CEO like Bezos could make strategic judgments at the Prime level. Now any enterprise can use AI to analyze user behavior, predict demand, and simulate strategies.

AI has a dual impact on decision-making: how decisions are made has been changed, and the very definition of "decision-making" itself has also been changed.

On the surface, Prime is a membership product, but in reality it is a strategic decision that changed the nature of the relationship between Amazon and its users - from "you buy, I sell" to "I understand you, I serve you".

2. Why Has Decision-Making Become the Most Valuable Resource?

1. AI Enables Infinite Supply of Strategies, Solutions, and Options

As I mentioned earlier, my judgment is: In the AI era, "strategies" are infinite, while "decisions" are scarce.

What does "strategies are infinite" mean? AI can generate an unlimited number of strategies, solutions, and options. Give AI a business problem, and it can provide you with ten sets of solutions, each of which can be split into ten variants.

McKinsey's research last year showed that nearly 90% of enterprises are already using AI somewhere, with AI-generated solutions costing only 1% of human-generated solutions, and the time required shortened from several months to a few hours.

2. Final Decision-Making, Value Trade-Offs, and Accountability Remain Scarce

But what does "decisions are scarce" mean?

Final confirmation, judgment, and risk-taking must be done by humans.

AI can tell you "based on data analysis, Solution A has higher expected returns", but it cannot say for you "I choose A, and I will take responsibility if something goes wrong".

There is a deeper question here: after AI makes "strategies" infinitely cheap, will the value of "decisions" rise or fall?

My view is that it will rise, and rise sharply.

When everyone can use AI to generate solutions that score 80 points, the people who can distinguish between 80-point and 95-point solutions become the scarcest resources.

When all competitors can access market data in real time, "identifying opportunities" no longer creates differentiation. The real differentiation lies in how you define problems.

Herbert Simon, the Nobel Prize-winning economist, proposed the theory of "bounded rationality" in 1957. He stated that human decision-making is constrained by information acquisition and cognitive processing capabilities, and can only pursue "satisfactory solutions" rather than "optimal solutions".

Seventy years later, AI's information processing capabilities far exceed those of humans, and it can find "optimal solutions" in many scenarios that humans would never be able to spot.

However, there is still an unshakable cornerstone in Simon's theory: Value judgment.