Jianshanxing Technology

English

中文

English

中文
Sponsored by [KnowFun]: Transform your materials into personalized learning with AI.

AI Doesn't Create Value — People Do ​

We · October 2026

This is the third article in the "Core Abilities for the AI Era" series. The first two covered the amplifier effect of AI, and taste and discernment. In this one, we look at specific people and stories—with the same AI technology, why can the gap be so absurdly large?

Let's start with two stories.

The first story is about one person.

Pieter Levels, a Dutch developer, built three websites by himself: NomadList, RemoteOK, and Rebase. Combined, the three products generate over $390,000 in monthly revenue—roughly $4.7 million a year, about 34 million RMB.

How many employees does he have? Zero. Just one person.

His costs? Less than $200 a month in infrastructure and tools.

How does he do it? He uses AI to write code, AI for design, AI for data processing, and automation tools for operations. One person, with AI and automation, matches the output of a team of dozens.

"I don't have a team. I have scripts and AI agents that do the work for me."

— Pieter Levels

This isn't an isolated case. Base44, acquired by Wix for $80 million in 2025, was also a minimal product with almost no team. And there's Zine, a SaaS product whose founder built the MVP in 8 weeks using AI coding tools, and had paying customers within 4 months.

These people used AI and made a fortune.

One-Person Companies: A New Species Exploding Globally ​

These cases of "one person equaling a team" aren't outliers—they're a microcosm of a global entrepreneurial wave.

This model has a specific name: One-Person Company, or OPC.

"OPC仅凭一人、一台电脑、一套AI工具，便可独立完成产品研发、内容创作、市场拓展、客户服务等全业务链条，成为人工智能发展背景下极具活力的全新创业模式。"

— China Daily, July 2026

In English: with just one person, one computer, and a set of AI tools, you can independently run the entire business chain from product development to market expansion—an entirely new and vibrant entrepreneurial model emerging in the AI era.

How fierce is this trend? Let's look at some data.

United States: Solo entrepreneurs now number nearly 30 million, accounting for 82% of all small businesses, contributing $1.7 trillion to GDP annually—equivalent to the economic output of a mid-sized country. Data from MBO Partners shows that US independent workers reached 72.9 million in 2025, up from 38.2 million in 2020—nearly doubling in five years.

Data from equity platform Carta is even more telling:

"Roughly 36% of new startups founded on the platform in 2025 had a single founder, up from 31% in 2024. AI tools lowering the barrier to entrepreneurship is a core driver."

— Carta data, reported by Xinhua, 2026

About 36% of new startups founded on the platform in 2025 had a single founder, up from 31% in 2024. AI tools lowering the barrier to entrepreneurship is a core driver.

Europe: Solo self-employment and single-person companies have long been mature, but AI has given individuals the tools to break free from time constraints and scale their business for the first time. Stripe's economic research report shows:

"New business registrations have risen roughly 40% in Australia, 70% in Finland, and 80% in France since 2017, with meaningful acceleration in 2025 alone."

— Stripe Economics, The Age of the Solopreneur

China: This wave arrived quickly too. By the end of 2025, over 730,000 new one-person companies had been registered nationwide. Even more notable is the policy response—23 cities across China have introduced special support policies for OPCs:

  • Pudong, Shanghai: Newly registered one-person companies can receive up to 300,000 RMB in free computing power
  • Shenzhen: Launched an AI-driven OPC entrepreneurship ecosystem action plan, the first provincial-level AI solo entrepreneurship policy framework in the country
  • Sichuan: Introduced 16 key tasks to build a regional OPC innovation ecosystem
  • At the national level, the Ministry of Human Resources, State Taxation Administration, and State Administration for Market Regulation jointly issued a document to streamline the entire process of personal entrepreneurship—registration, business licensing, tax filing

Why did one-person companies suddenly blow up now? The answer is simple: AI.

Before, no matter how capable a person was, their time and energy were limited. You had to do product, operations, customer service, design, finance... one person simply couldn't handle it all. So starting a business required hiring people, hiring people created management costs, and management costs required more revenue, forming a "must get bigger" cycle.

AI broke that cycle. AI writes code, AI does design, AI processes data, AI handles customer service, AI generates content—work that used to require a team, now one person with a set of AI tools can handle it.

This isn't simply "entrepreneurship is easier now"—it's a redefinition of what "entrepreneurship" means.

Before, the standard for starting a business was: find co-founders, build a team, rent an office, raise funding, expand.

Now there's a new possibility: one person, one computer, one set of AI tools, one niche market.

"Small and beautiful" is no longer synonymous with "can't scale up"—it's a deliberate choice.

The second story is about many large companies.

MIT's NANDA Lab published a State of AI in Business report in 2025, and one number in it is striking:

"Roughly 95% of enterprise generative AI pilots produce no measurable P&L impact."

— MIT NANDA, State of AI in Business, 2025

About 95% of enterprise generative AI pilot projects produce no measurable profit and loss impact.

95%. What does that mean? Out of 20 AI projects, 19 are just spinning wheels.

Gartner's data is more specific. Their 2025 survey shows:

"73% of AI pilots never reach production. The average company wastes $21.5 million on failed AI initiatives."

— Gartner, AI Project Success Rates, 2025

73% of AI pilots never make it to production. The average company wastes $21.5 million on failed AI initiatives.

IBM's 2025 CEO study reached a similar conclusion:

"Only 25% of AI initiatives have delivered expected ROI, and just 16% have scaled enterprise-wide."

— IBM Institute for Business Value, 2025 CEO Study

Only 25% of AI initiatives delivered expected ROI, and only 16% scaled across the enterprise.

RAND Corporation's 2025 analysis is even more pessimistic:

"80.3% of AI projects fail to deliver their intended business value."

— RAND Corporation, 2025

Over 80% of AI projects fail to deliver their intended business value.

These companies used AI and burned a fortune.

On one side, one person making tens of millions a year with AI. On the other side, large companies spending tens of millions of dollars and getting nothing back.

With the same AI technology, why is the gap so big?

Is the model not strong enough? No—everyone's using more or less the same models.

Are the tools not good enough? No—good tools are available to everyone.

Is it not enough money? Definitely not—large companies have budgets thousands of times larger than individuals.

So where's the gap?

Gap One: Whether You've Found the Right Problem ​

Look at the successful individuals and small teams—they all share one thing in common: they're very clear about what problem they're solving.

Pieter Levels' NomadList solves the problem of "where should digital nomads live"—a very specific, very narrow problem. He didn't build an "AI lifestyle platform." He built a "city ranking for digital nomads." The problem is narrow, the solution is clear, and AI helps him maximize efficiency in data processing and website maintenance.

What about Base44? It solves "how small business owners can quickly build websites"—equally specific. It didn't build an "AI万能 website builder." It focused on one scenario and perfected the experience.

Midjourney is even more extreme. When David Holz founded Midjourney, there were already many AI drawing tools on the market. But he didn't say "I want to make the best AI drawing"—he said:

"We're a research lab focused on new mediums and tools for empowering imagination."

— David Holz, Midjourney founder

Sounds vague, but Midjourney's product is actually extremely focused—it does one thing, "text to image," and it runs entirely within Discord, without even its own app. 11 employees, $500M ARR—because they solved one problem to perfection.

Now Look at Failed Enterprise Projects ​

Where do failed enterprise AI projects usually go wrong?

Not because the technology is bad, but because the problem wasn't properly identified.

MIT's report summarizes the main reasons for failure:

"The failure pattern is rarely technical. It is operational: weak success criteria, fragile data foundations, missing MLOps, and no feedback loop after launch."

— MIT NANDA, State of AI in Business, 2025

Failure is rarely technical. It's more operational: weak success criteria, fragile data foundations, missing MLOps, and no feedback loop after launch.

Notice what's ranked first: weak success criteria.

What does "weak success criteria" mean? It means you don't even know "what success looks like" before you start implementing AI.

For example, many companies say "we need to implement AI customer service"—that's not a problem, that's a technology choice. The real problem might be:

  • "Customer service response is too slow, and users are complaining a lot"
  • "Customer service costs are too high, accounting for 15% of revenue"
  • "There are too many repetitive questions, and the service team is overwhelmed"

If you don't clarify these problems and just jump into AI customer service, the outcome is likely:

  • AI customer service launches, but answer quality is poor
  • Users get angrier, and even more escalate to human agents
  • Costs don't come down, and user satisfaction drops
  • The project fails, money wasted

Technology is a hammer, but you first need to know where the nails are. Many companies buy the hammer first, then go looking for nails—searching and searching but not finding any, and eventually the hammer just gathers dust.

BCG's research confirms this:

"Leading firms focus on 3.5 AI use cases versus 6.1 for lagging firms, yet generate 2.1 times more ROI."

— BCG, 2025

Leading companies focus on just 3.5 AI use cases, while lagging companies pursue 6.1—yet leaders generate 2.1 times more ROI.

Doing less earns more—because fewer but better beats more but worse, by a lot.

Gap Two: How Deeply You Understand the Problem ​

Finding the right problem isn't enough—you also need to understand it deeply enough.

Creator Case: Same AI Drawing Tools, Why Do Some Go Viral and Others Stay Mediocre? ​

You've probably scrolled past those AI-generated travel short videos—camera sweeping across Tokyo's Shibuya crossing, neon reflections on rain-soaked pavement, full cinematic quality. Some accounts get tens of millions of views in 72 hours from a single video.

The AI tools they use aren't any different from what you use—Midjourney for images, Runway for video. The tools are public, the models are public, everyone can use them.

But why can some people create viral hits with tens of millions of views, while others make videos nobody watches?

The difference is in understanding "what makes good content."

Creators of those viral videos don't just type a sentence and let AI generate. They know what kind of shots feel cinematic, what kind of rhythm grabs attention, what kind of imagery triggers emotional resonance. Their prompts say things like:

"Cinematic wide shot, Tokyo scramble crossing, low angle, wet pavement, neon reflections, hyper-realistic, 35mm lens"

These details—camera position, angle, visual language, lighting and texture—AI didn't teach them. They already knew them. AI just helps them quickly realize the ideas in their heads.

The deeper you understand content, the more AI can help. People who have no feel for content just use whatever AI gives them, and the result is mediocrity.

Indie Developer Case: The Better You Understand Product, the More Useful AI Is ​

Take the indie developers building products alone.

They use AI to write code, but it's not "let AI build me a product"—they know what the product should do, who the users are, what the core features are, how the technical architecture should work. AI just helps them write code faster.

Zine's founder wrote on his blog:

"I built Zine in 8 weeks with AI. But the 8 weeks was coding time. The thinking, the positioning, the understanding of the user — that took years."

— Zine founder, 2025

AI can help you write code, but it can't help you figure out what product to build. Product sense, user insights, business judgment—these are things AI can't replace.

Enterprise Anti-Case: Implementing AI Without Understanding Business ​

What about on the enterprise side? In many companies' AI projects, the tech team pushes while the business team watches from the sidelines.

The tech team understands AI but not the specific business. The business team understands the business but not AI. The two sides can't communicate, and the result is:

  • The tech team builds an AI system that's "technically impressive"
  • The business team says "this isn't what I need"
  • The project can't move forward and eventually fizzles out

RAND's report says:

"Enterprises are acquiring the technology but not the organizational conditions that convert technology into returns."

— RAND Corporation, 2025

Enterprises are acquiring the technology but not the organizational conditions that convert technology into returns.

What are "organizational conditions"? Put simply—do people who understand business know how to use AI, and do people using AI understand the business?

If business understanding and AI capability exist in separate silos, even the strongest technology can't produce value.

Gap Three: Using AI to Solve Problems vs. Using AI for AI's Sake ​

There's another fundamental difference between successful people and unsuccessful ones:

  • Successful people: I have a problem, let's see if AI can help solve it
  • Unsuccessful people: I have AI, let's see what I can do with it

These two mindsets seem similar, but the results are worlds apart.

The Positive: Starting from the Problem ​

When Pieter Levels built NomadList, AI hadn't blown up yet. He started by manually collecting data and maintaining the website, doing it for many years. When AI came along, he handed off repetitive work to AI—data organization, content generation, customer service replies—and focused himself on product and strategy.

He didn't build a product in order to use AI. He was building a product and discovered AI could help.

That's an important distinction. For people starting from the problem, AI is a tool. For people starting from AI, AI is the goal.

The Negative: AI for AI's Sake ​

On the enterprise side, "AI for AI's sake" is extremely common:

  • The boss says "every other company is doing AI, we need to too," then buys a solution from a vendor
  • The tech team wants to learn new technology, finds a project to practice on, and once it's done nobody cares
  • The marketing team wants to ride the hype, holds an "AI launch event," and after the announcement nothing follows

Gartner calls this phenomenon "solution in search of a problem"—holding a solution and looking for a problem.

You already bought the hammer, so you can't just not use it, right? So you go looking for nails everywhere—and what you find are fake problems, small problems, problems not worth solving with AI. The result: lots of money spent, nothing to show for it.

Self-Assessment: Are You Using AI to Solve Problems, or Using AI for AI's Sake? ​

Ask yourself a few questions:

  1. The last time you used AI, did you have a problem first, or did you first think "let's see what AI can do"?
  2. The things you use AI for—would you still do them without AI? (Or are you only doing them because AI exists?)
  3. Can you clearly explain "how much better things are because of AI"? (How much time saved? How much quality improved? How much money earned?)
  4. When you encounter a problem, is your first reaction "how do I solve this" or "can AI solve this"?
  5. Have you ever had the experience of "making things more complicated just to use AI"?

Four States ​

StateCharacteristicsOutcome
FollowerEveryone's using AI, so I should tooUsing it, but don't know what for
Tool UserUse AI for specific needs to boost efficiencyEffective, but limited impact
IntegratorDeeply integrate AI into workflowsBoth efficiency and quality improve significantly
DefinerStarting from problems, use AI to redefine solutionsOrder-of-magnitude improvements

Most people are at the "Tool User" level—they know AI can help, but only use it in scattered scenarios. What really拉开 the gap is the "Integrator" and "Definer" levels—not using AI sporadically, but starting from problems and using AI to redesign entire processes.

How to Upgrade from "Using AI" to "Using AI to Solve Real Problems" ​

After all these cases and data points—how to actually apply this?

Three specific practice methods:

1. Before Starting, Write a "Problem Statement" ​

Next time you use AI for something, don't rush to type a prompt. Spend two minutes writing a "problem statement" that includes:

  • What is the specific problem I'm solving? (Not "write copy" but "write email subject lines that improve open rates")
  • Who is the audience for this problem?
  • After solving it, what counts as success? (Open rate rising from 20% to 30%? Conversion rate up 5%?)
  • What's the current state? How big is the gap?

Write it out clearly, then let AI get to work.

This method sounds simple, but most people never do it—they just ask AI whatever comes to mind, end up with a bunch of useless stuff, and then blame AI for being bad.

The process of writing the problem clearly is the process of clarifying your thinking. Often, once you've written the problem clearly, the answer emerges on its own.

2. Use the "Five Whys" to Find Root Causes ​

When you encounter a problem, don't immediately think about solving it with AI. First ask "why" five times to dig down to the real problem.

For example:

  • "I want to start an AI short video account" → Why?
  • "Because AI videos are hot and can grow followers" → Why do you want to grow followers?
  • "Because followers lead to monetization" → How do you plan to monetize?
  • "Ads or courses" → What product or service do you have to sell?
  • "...haven't thought about it yet"

Oh, so the real problem isn't "how to make AI videos"—it's "what business do I want to build?" Without figuring out the business, even if your videos go viral, they won't convert to revenue.

The Five Whys helps you distinguish "symptoms" from "root causes." Solving the root cause is actually solving the problem.

3. Practice Subtraction: First Ask If It Can Be Solved Without AI ​

This is counterintuitive but extremely effective.

Every time you want to use AI, first ask yourself: If there were no AI, how would this problem be solved?

If it can be solved without AI and the cost is about the same—then you probably don't need AI at all.

If it's hard to solve without AI, or the cost would be very high—that's where AI truly adds value.

Why do this exercise? Because many people use AI not because AI solves problems, but because "AI is hot" or "everyone's using it." Practicing subtraction first helps you eliminate fake needs.

Good uses of AI are "can't do without it," not "nice to have."

Final Thought: AI Is an Amplifier, But You Need a Signal First ​

Returning to the question at the beginning: with the same AI, why is the gap so big?

Because AI is an amplifier.

An amplifier can't produce a signal on its own—it can only amplify the signal you feed in. If what you feed in is clear problems, deep understanding, and明确 goals, AI will amplify them tenfold, a hundredfold. If what you feed in is vague requirements, shallow understanding, and a bandwagon mindset, what AI amplifies is chaos and waste.

The indie developer who equals an entire team—AI didn't make them great. They already had deep understanding of product, users, and business. AI just amplified the effect of that understanding.

The companies burning money and getting nothing back—it's not because AI is bad. They hadn't thought through what problems to solve. AI just amplified the cost of that confusion.

So don't be in a hurry to learn AI tricks. First think your problems through clearly.

When the problem is clear, AI is a magic weapon. When the problem isn't clear, AI is a money-burning machine.

References: