Agent Lightweight Entrepreneurship Experiment: A "Team of 10" Prying into a Trillion-dollar Market

The "entrepreneurial battle" in which small teams collaborate with AI technology to tap into the multi - billion - dollar market is underway.

On June 25th, Li Zhifei, the founder of Mobvoi, shared a mind - blowing "one - person company" experiment: Using AI as his sole assistant, he independently developed a prototype of a collaboration platform designed specifically for AI organizations - the "AI version of Feishu" in just two days.

This is an office software with core functions such as login, private chat, group chat, file upload, message forwarding, and reply. "To be honest, I almost gave up after working for half a day at the beginning because I couldn't solve the database problem and there were always various key errors. But I finally managed to finish it within two days," Li Zhifei said.

To promote this Agent product, Li Zhifei also had the AI build a website highlighting the product's features and unique functions in just five minutes. In the past, such work usually took a team of five or six people a week to complete.

In Li Zhifei's view, this experiment validates his prediction: "The era of AGI has arrived, but it is not evenly distributed yet." As early as the 2024 performance press conference in March, he predicted that a company that originally required 200 people may only need 20 to 30 core employees in the future, and the other positions will be filled by AI Agents.

Now, this "wave of micro - teams" is sweeping through the startup ecosystem. More and more new companies with a scale of 1 - 2 people can achieve the efficiency of a team of dozens in the past. Recently, there have been more and more AI startups composed of one or two people.

Tan Dai, the president of Volcengine, also noticed that startups born in the AI era are characterized by a small number of people. He said in an interview with Phoenix Tech in June: "With AI, one person can do the work of many. A company may have less than 20 people, but it can still develop at a high - speed continuously."

The startup in the AI for IoT (Internet of Things) field, Mighty AI, is a typical example. Its founder, Sun Zhuojian, revealed that the core team of the company only has 2 people (including himself and CTO Zhang Qi), and only 2 temporary operation employees were added this year due to the surge in the number of customers. With in - depth AI collaboration, tasks that originally took a week to complete can now be compressed to one day.

Dario Amodei, the CEO of Anthropic, predicted at the first developer conference held this year: "The first company valued at one billion dollars, operated by a human employee in cooperation with AI, will be born in 2026."

The "entrepreneurial battle" in which small teams collaborate with AI technology to tap into the multi - billion - dollar market is underway.

The New Logic of Agent Entrepreneurship

Currently, AI technology is reshaping the work paradigm of startups with exponential efficiency. The most core aspect is the disruptive transformation of programming efficiency by AI.

Li Zhifei cited his own experience as an example: "During my time at Google, writing 300 lines of complex algorithm code in a day was considered highly productive for me. Now, a general Agent can produce 3000 lines of industrial - level Python (a computer programming language) code in just 3 hours - pure backend logic without a UI (user interface), and the quality far exceeds my level." This means that the coding ability of an AI Agent in 3 hours is equivalent to his workload in 10 working days in the past.

Moreover, the team scale will also be disrupted by AI. "Back then, Google Translate required 20 top - notch PhDs to work on it for a long time. Now, I can complete a system of the same complexity on my own," Li Zhifei said.

Li Zhifei Source: The interviewee

Li Zhifei claims to be a "wild spokesperson" for Cursor (an AI programming tool). He currently spends up to $50 per day using Cursor, far exceeding the monthly membership fee of $25. "If there were no Cursor, I wouldn't dare to say that I can do the work of 100 people on my own," Li Zhifei said.

For AI - native companies in the AI era, the efficiency leap brought by AI goes far beyond code generation. According to Sun Zhuojian, AI has been deeply integrated into their entire chain from customer engagement to R & D implementation. Whether it's writing proposals, implementing code, or daily email communication, it can all be handled by AI employees. After simple in - house development, each person is equipped with three computers running around the clock, and humans do the feedback and supervision work, which has also brought more than 300% improvement in labor efficiency.

Therefore, in the more than two years of entrepreneurship, instead of expanding, the size of the startup team led by Sun Zhuojian has been streamlined from the initial 10 people to a core of 2 people, which has overturned his core perception: "We once accepted the view of investors that a startup team needs to have a complete set of roles. But in our company's practice of focusing on R & D, we found that with the help of AI, two people can also create a leading product in the industry."

Shang Yanyi, the founder of Cross Starry Sky, also experienced the efficiency revolution brought by AI: Her team transformed to develop the travel Agent - iMean AI at the beginning of 2025. In less than half a year, they completed the R & D and launch of the first - generation product and the version update on June 16th. "Now, everyone in our company uses a large number of AI tools. For example, the development team only needs to send the task target to the AI programming assistant, and a very high completion rate can be achieved," Shang Yanyi said.

In Shang Yanyi's view, there may no longer be traditional development positions in the future. The functions of positions will move forward comprehensively, that is, there will be no development, only products. More job requirements will fall on high - level "design", that is, doing a good job in "managing" AI.

The education Agent company, Ke You Ji, has also achieved significant efficiency improvements in two aspects with the help of AI. Li Chun, the founder, said that at the beginning of the company's establishment, he deployed AI programming tools for the engineering team to improve programming efficiency; civilian positions can also use AI to accelerate content creation and break through creative bottlenecks.

On the other hand, the leap in large - model technology has brought a massive increase in efficiency. Shang Yanyi said that two or three years ago, when they were doing private deployment for bank customers, the whole team could only deliver one model (based on the early LLaMA) in more than half a year. Now, based on many models such as DeepSeek, more than 100 vertical models (such as in the travel scenario) can be trained in the same amount of time.

While AI technology improves operational efficiency, it also enhances revenue efficiency. With the support of Shenzhen's policies, Sun Zhuojian's team operates at a very low cost. He revealed that the goal of the enterprise in 2025 is to serve only one customer well, which can not only polish the product but also cover the costs. Shang Yanyi's team's product was launched just over a month ago, and the US travel search engine service provider KAYAK has already come to discuss a revenue - sharing cooperation. The next - stage goal of the company is to reach a monthly active user base of tens of millions. Li Chun has set the annual revenue target of Ke You Ji at 15 million yuan this year, which has changed the profit model of "small team + AI leverage" in the Agent era.

Zhang Lu, the founder of Fusion Fund, believes that the opportunity brought by the era of artificial intelligence is that in the future, many companies can increase their revenue without building a very large team. 70% of the invested enterprises of its third - phase fund increased their revenue by 20 times last year, and most of them are small teams of less than 10 people. Two companies have achieved a leap in revenue from zero to tens of millions of dollars with less than 10 people, and a team of 30 people has increased its annual revenue from 500,000 to 150 million dollars.

This "exponential revenue curve" is reshaping the capital relationship - startups' dependence on financing has dropped sharply. They can selectively introduce high - quality capital based on rapid profitability, give up less equity, and strengthen their control over the company.

However, open - source models represented by DeepSeek have had a disruptive impact on AI entrepreneurship and also brought competitive pressure that needs attention. Sun Zhuojian believes that currently, the training and inference costs of many products have dropped sharply, but the core capabilities are brought about by model updates rather than the key know - how of entrepreneurs. Therefore, to face the increasingly intense competition, his advice to other entrepreneurs is to focus on vertical scenarios and build data barriers; construct a hybrid architecture that is decoupled from the basic model and can be compatible with multiple models; pay attention to the latest trends in Agents and at the same time improve the enterprise's engineering and commercialization speed. Only in this way can they stay sober and maintain the vitality of the enterprise in the upsurge.

From Technical Feasibility to Business Sustainability

"If the definition of AGI is to liberate 80% of human work, then at least in the software industry, AGI has been achieved," Li Zhifei said.

The popularization of the underlying technology of large models is inspiring more people to join the upsurge of AI application entrepreneurship. The AI application field in China is experiencing explosive growth, and the consensus of the "Year of the Agent" is accelerating. In the spring roadshow of Miracle Plus in 2025, among the 57 short - listed enterprises, nearly half (28) focused on the Agent track.

Sun Zhuojian also believes that the opening of DeepSeek has triggered the domestic AI application market for the second time, bringing benefits to both the market side and the technology side.

Sun Zhuojian Source: The interviewee

Capital and entrepreneurs are also moving towards each other. Zheng Xuanle, the founder and CEO of Light Source Capital, has been in more frequent contact with entrepreneurs since this year. He also said in an interview with "China Entrepreneur" at the end of April: "In the past two months, the investment actions of institutions have accelerated significantly, and the market sentiment is significantly recovering this year."

On this basis, Zheng Xuanle pointed out that China will welcome four major dividends in the AI era. First, the entrepreneur dividend: The improvement of model capabilities, the spread of engineering capabilities, and the decline in computing power costs have lowered the threshold for entrepreneurship - from the "theory of only technical experts" to the entry of product experts, commercialization talents, and industry integrators; Second, the industry dividend: DeepSeek has promoted the awakening of the public's awareness of AI. Entrepreneurs in traditional industries are actively embracing technological changes, and a large number of industrial scenarios are accelerating their opening to AI, giving rise to a large number of innovation opportunities in "AI + industry"; Third, the capital dividend: The value of Chinese technology assets is being re - evaluated, driving global funds to flow back to China. At the same time, the smooth and optimized exit channels also bring capital dividends, and the technology investment ecosystem continues to improve; Fourth, the globalization dividend: As an advanced productive force, AI naturally has the potential for globalization, providing new impetus for Chinese enterprises to expand into the international market.

Moreover, according to people close to Qi Lu, when Qi Lu went to the United States to raise funds this year, he found that the perception of Chinese AI in the US investment community had undergone a fundamental reversal - from the inherent prejudice that "China will never catch up with OpenAI" to the view that "China's AI development momentum has gradually surpassed that of the United States."

Sequoia Capital proposed at the beginning of the year that "the next step of AI is no longer to sell tools, but to directly sell results," which further encourages entrepreneurs to focus on their core capabilities - locking in scalable business scenarios in their areas of expertise or planning a clearer product direction.

In this context, the implementation of AI applications will unleash great market potential. According to data from Guanyan Report Network, the market size of the Chinese AI Agent in 2023 was approximately 55.4 billion yuan, and it is expected to reach 852.035 billion yuan by 2028, with a compound annual growth rate of up to 72.7%.

Although the market potential is huge, AI application entrepreneurship currently faces great challenges. There is an inevitable exploration period between technical feasibility and business sustainability, and entrepreneurs often need to go through multiple rounds of product - market fit trials and errors.

In response, some investors said that AI entrepreneurship has distinct "generational evolution" characteristics: The first - generation products are often for technical verification, and the real market value anchor points are often established in the second, third, and fourth - generation iterations.

In 2021, after observing the opportunities in large - scale AI models, Shang Yanyi's team initially targeted the AI track for operating computers (software). After succeeding with some bank and financial customers in China, they targeted software giants in the United States. However, due to insufficient localization and time - zone limitations, the communication efficiency was low, and the team gradually realized their weak competitiveness in this field.

At the beginning of 2025, Shang Yanyi's team found that the penetration rate of AI applications in professional fields (such as design and programming) was high, while there was a significant gap in the mass - life scenarios. They decided to shift to the C - end market. After research and discussion, the team regarded the travel track with a high commercial ceiling, strong user pain points, and low AI penetration rate as the breakthrough point. Less than half a year after the launch of their travel Agent product, the number of overseas organic growth users has exceeded 110,000.

Shang Yanyi Source: The interviewee

Li Chun also admitted that at the beginning of entrepreneurship in October 2023, the team had decided to enter the AI education track, but the specific entry point still went through multiple rounds of exploration. In 2024, they also tried to launch products such as AI college entrance examination volunteer filling. Li Chun believes that "extensive trial - and - error in the early stage is not a bad thing - it can quickly verify the direction and deeply understand the industry, laying the foundation for subsequent subtraction in business and strategic focus."

One - Person Unicorn: A Gimmick or the Future?

Data from QuestMobile shows that in March 2025, the monthly active users of AI - native applications reached 270 million. Compared with the total monthly active users of 1.259 billion in China's mobile Internet, the penetration rate is only 21.4%; the average daily usage time per user is only 123 minutes. This means that the public has not yet formed a rigid demand for AI products, and more AI products are needed to fill the "blank market".

Sun Zhuojian is optimistic and takes a long - term view of the future development of AI. He believes that the basic model technology will continue to iterate for at least the next 3 - 5 years, indirectly promoting the continuous release of the potential of Agents; at the same time, it will take at least ten years for Agents to complete the technological upgrade, product refinement, and social perception reconstruction of an "all - around employee".

Peng Zhiqiang, the chairman of Shengjing Network and the founding partner of Shengjing Jiacheng Venture Capital, believes that with the upgrade of AI technology, when the cost of developing a software and delivering an Agent service approaches zero, a large number of small scenarios that were previously not covered by software will emerge, and there will be a huge number of...