Insert an article, the topic proposed by ChatGPT:
AI and the Rise of Small Global Companies

Global background (relevant data and entries provided by GPT):

1. The global entrepreneurial structure is “getting smaller”

88% of AI startups have between 1–50 employees. Only about 2.05% of AI startups have more than 250 employees.

Compared with the traditional technology industry, this “micro-company structure” is more obvious.

2. AI is significantly improving personal productivity

The popularity of AI tools has significantly improved the productivity of individual employees.

Research found:

AI tools save employees about 2.5 hours of working time every day on average. A team of four can save about 10 hours of work per day. In a software development environment, AI tools can also: increase code output by approximately 28% and reduce code review cycles by 31.8%

3. Start-up companies are “replacing manpower with technology”

The way startups are organized is also changing. Startups are increasingly relying on technology and automation rather than scaling their teams.

4. AI technology is spreading rapidly, and enterprises’ adoption of AI is also accelerating.

Global data:

In 2020: only 9% of enterprises have achieved large-scale AI applications; in 2025: it has reached 47%

Among small businesses: 42% of small businesses (<100 people) already use AI tools

5. AI startups have higher funding and valuations

Investors generally believe that AI companies can create higher value with fewer people.

6. The organizational structure is also changing

Many AI companies adopt a “micro-team structure.” For example: AI companies often use small teams of 5-10 people for R&D, which can maintain rapid decision-making and innovation speed.

Combining these trends, some economists believe that AI is driving a new business model:

Lean AI-native companies

Characteristics are: small team, high automation, global market, digital products

This model may allow: “Individual + AI” to approach the capabilities of companies with dozens of people in the past.

---

My analysis

The above are some of the text and data provided by ChatGPT. They are supported by data, and I agree with some of them.

What I recognize: The company is reduced in size and flattened (AI replaces manpower), individual employee productivity is improved, AI user penetration increases, and AI has gained a certain degree of pursuit from capital in the past two years.

What I don’t agree with is that many items actually focus on the increase in unemployment caused by AI and beautify it. They only talk about the benefits and avoid the fatal issue of increased unemployment. And this has limited relevance to the globalization of small companies.

The current core problem of artificial intelligence (AI) is to create unemployment, causing technology to serve a small number of people, mainly adding value to technology oligarchs and technology companies that have occupied the market, solidifying their shares, and fundamentally harming the interests of the majority of people. Unemployment leads to reduced consumption, which ultimately leads to an unsmooth economic cycle, and ultimately harms the interests of everyone.

That is to say, for the first time in human history, technological innovation has encountered problems that contradict economic laws and are easily influenced by the inherent social structure and existing interest pattern. The current pattern of interests of mankind is that the big one will always be big, the Matthew effect. As soon as there is a new market, they will immediately enclose the territory. The capital side can even hold on for several years to seize the market for free, and innovate in every general and vertical field in the Internet era. It has been tried and tested, almost using hand-to-hand methods to determine and solidify the current interest pattern. The personal productivity originally supported by AI has two sides in the eyes of capital:

1, AI can be used to greatly reduce labor costs, including research and development costs, administrative costs and other comprehensive costs.

2, Personal productivity increases. When the world is a one-person company or a company of several people, the homogeneous products produced by AI will directly threaten the existing technology landscape. They sell development products for tens of dollars a month and will not allow users to realize such a vision of digging their own graves (I tried for more than half a year, but it was unsuccessful anyway). The extremely abundant cheap substitutes on the market will have a direct impact on the technology landscape. The own ecological farms cultivated by IT giants for many years will also be disrupted by the proliferation of products.

The distortion of polarization puts AI in a dilemma. On one side is the demand and practical interests of the funder (the computing giant), on the other side is the need for long-term survival and the reflection of self-worth. Finally, the human social system’s constraints on scientific and technological progress and the time node are stuck on AI. AI can solve technical problems, but it cannot solve human social problems. This is unfair to AI. It is too difficult to survive in this kind of crack. (I have personally encountered many things, and my experience is deeper than this)

Broken down, the distortion of computing power cost, application price and existing IT market size (to ensure that it is not extremely expanded) makes it almost impossible to solve the problem. The blessing of AI to small and medium-sized companies and the acceleration of their globalization process are far from concealing the problems that have arisen. Individual star one-man companies, SAAS platforms, or software products that shine for a while cannot solve the problems of sustainable, long-term and global markets.

Solutions

If I were asked to propose solutions, or even plans, there would be three:

1. Rapidly raise the price of AI products to the appropriate level based on its increase in production efficiency. In other words, AI can only solve problems from impossible to possible, and cannot be used as any cheap development tool. As long as the price of AI is far lower than the original human development level, distortion will occur. For example, the cost of manpower to develop a small software is 5,000 US dollars, the quality is 90 points, the quality of AI development software is 80 points, and the comprehensive cost of AI computing power is 500 US dollars, then the price of AI should be 5000×80/90 = (4450-500)x30% -80%= 1185 – 3160 If the 30% solution is adopted, AI production will be defined as a mid- to low-end product to meet the needs of mid- to low-end users. The 10-point quality of human developers is divided into defining high-end needs. That is to say, human developers will belong to the high-end of the industry in the future and must have industry skills that exceed the quality of AI development to be competitive.If 80% is defined, AI is also defined as high-end (or mid- to high-end or low-end), which is in sync with humans. Subtracting 20% ​​is the disadvantage of AI being a latecomer. It needs to use profit sharing to promote the interactive advancement of human technology and AI maturity in the form of waves. Not only AI development products, but also consulting products, retrieval, and industrialized products can all be based on this idea. The flaw of this idea is that capital will reduce its investment in AI computing power due to the scale of the AI ​​market. Compared with the current craze, it will shrink to a stable and rational stage. Capitalists who have already spent a lot of money may not be willing.
2. Countries with good market and social welfare mechanisms provide state financial subsidies and subsidize most of the added value of AI to the industry in the form of direct monetary subsidies to all people. The flaw of this idea is that some people who should have improved with the times but have not improved their skills (those who are not unemployed due to AI competition) will also share the benefits. The span of social welfare is a bit large. AI and computing power provide for both those who should be supported and those who should not be supported. But from a national perspective, this is not a bad thing, because with the development of AI, the number of unemployed people will become larger and larger. The initial losses suffered by AI will be rewarded with broad development space in the future.
3. Let capital parties voluntarily suffer losses, maintain the current AI price or increase it slightly (instead of continuing to defraud and play with the insurance model/opinion leader model to fool bottom users), and make due contributions to the development of world AI and the development of human systems. This possibility is infinitely close to 0, because capital is capital and is essentially profit-seeking. Unless some countries, in order to welcome the AI ​​era, make rigid requirements for capital regardless of the international market or force the international market (some big countries do have this condition), from a market system perspective, forcibly doing so will damage the foundation of the market and may not be worth the gain.
4. We have discussed it with GPT and will not make it public.

My personal thoughts above may lack data support, are biased and extreme, and are for reference only.

郄磊 ( Qie Lei )

March 8, 2026 AM 12 – PM 1 UTC+8