Investing in stocks means relying on Gold Qilin analyst research reports—authoritative, professional, timely and comprehensive, helping you uncover potential thematic opportunities! This article is sourced from a special report on China's small factories.

Editor's note: With international trade fluctuating, the consumer market cooling and the AI wave arriving, China's small factories are quietly shifting gears. Zhejiang is home to a large concentration of hidden champions, manufacturing firms, supply chain companies, foreign trade enterprises, specialized market operators and consumer brands. Although these small Chinese factories may not appear to be at the very center of the stage, they are the solid foundation of Chinese manufacturing. They are moving from earning processing fees to building brands, from relying on manpower to seeking efficiency from AI, and from obsessing over large orders to embracing small ones, constantly completing self-iteration. This special series focuses on these units at the end of the supply chain, recording the most vivid business stories and changes.

At 6 p.m., Jialin, an operations assistant at a manufacturing company in Zhejiang, closed his laptop and ended his workday. Jialin's work includes store data processing, product maintenance and product listing, which previously required him to stay busy from 8:30 a.m. until the afternoon. After integrating AI, these trivial tasks that once filled most of the day have been compressed to three or four hours, allowing him to devote substantial time to new product development and backend ad placement.

While Jialin has stepped away from repetitive labor, salesman Lin Ke finds it much harder to "clock off." Some of his clients are still students and only have time to discuss products after 10 p.m. "The latest one was a high school student who called after 11 p.m., and we talked about a customized product for more than an hour." According to Lin Ke, clients usually arrive with an AI-generated rendering, but when they reach the workshop entrance, it may not be reproducible due to differences in fabric or craftsmanship. He has to translate the imagination stacked up by AI into production parameters that the workshop can understand, item by item.

The daily lives of Jialin and Lin Ke are the most authentic cross-section of frontline small-factory employees under the AI wave, and such scenes are playing out across industrial belts in Yiwu, Yongkang, Keqiao and other places. Data from the Yiwu Mall Group shows that nearly 30,000 local merchants are now routinely using various AI tools. According to "China's Labor Market in the AI Era" released by the Center for Education, Innovation and Sustainable Development at ShanghaiTech University, the research team built a database from 743 million recruitment records spanning 2022 to 2026, and further calculations based on 40.01 million online recruitment demands from January to April 2026 show that 26.6% of positions fall into the substitution-pressure zone of high exposure and low complementarity, meaning one in every four positions is standing at the threshold of structural adjustment.

China's small factories thus stand at the crossroads of this round of technological replacement: which experiences can be loaded into systems, and which judgments must remain with people; when bosses personally build AI employees, what can workers rely on to stay on the next recruitment sheet?

Becoming a workplace newcomer once again. In the AI era, both tasks and roles have changed. Jialin still sits at the same workstation, and his job title has not changed, but the content of his work has been reorganized. He said that although AI helps, he also needs to reunderstand which things no longer need to be done entirely by himself and which things have become more important. "Some very basic things, including operations, product listing and data processing, I think can absolutely be left to AI." However, Jialin admitted that when buyers are also using AI to generate product ideas, human professionalism becomes more important, because some AI-generated product effects cannot be realized in the factory. "For example, a client wants a pattern design, but in reality the product can only be printed, not embroidered. Yet what AI gives him is an embroidery effect, so we need to give the client more professional advice."

For Jialin, the balance of work is tilting—standardized tasks on one end are being taken over by AI, while the other end consists of responsibilities that are harder to define by rules. He must learn faster and judge more accurately. For Jing Jianghuan, deputy director of the new media home cleaning division at Zhejiang Oukaisi Technology Co., Ltd., this change is most prominently felt as the freshness of becoming a "newcomer" again. Jing Jianghuan, 33, graduated ten years ago, and entering e-commerce was already a career turn for someone who studied tourism management. When AI entered the company, he once again stood before an unfamiliar working language.

Jing Jianghuan admitted that at first, when younger colleagues born after 2000 began trying new tools, he felt somewhat resistant. "It's not about whether the tool is difficult, but about not wanting to learn something new from scratch." But after he really started learning, he found that AI could indeed "lift people up a notch"—information that originally needed repeated organizing could be quickly gathered together. In the end, he settled down, started from the basic concepts and operating models of AI, tried letting AI teach him AI, and then gradually gained hands-on experience through trial and error.

Xu Linfu, head of the Pinduoduo project at Oukaisi, felt the efficiency improvement from AI tools even more deeply. According to Xu Linfu, although the team has only 8 operations staff, it can maintain about 20 to 30 stores. He revealed that a competitor's link may accumulate 50,000 to 60,000 comments, and merely breaking down dozens of links used to take 10 days or even half a month. After integrating AI, it can be completed in a morning or an afternoon. Xu Linfu estimated that the knowledge base and analysis tools save the team about half to 60% of their time. At the same time, he believes this efficiency has already entered the company's staffing calculations. "As business continues to grow, the team no longer needs to expand at the original ratio."

Listing products, organizing spreadsheets and handling replies were once the entry point for young people to become familiar with the e-commerce business, but this kind of repetitive work is also the first to be hit by AI. Research by ShanghaiTech CEISD shows that among newly added online recruitment demand, positions with monthly salaries of 5,000 to 8,000 yuan and requiring one to three years of experience are under the deepest pressure. However, while AI brings changes to small factories, it also brings job anxiety. In response, Jing Jianghuan said frankly: "When the automobile era arrived, no one would ask the coachman for his opinion. Technology will not wait for everyone to be ready; those who drive carriages can only relearn how to drive cars. Only this time, the new driver's cab is on a computer screen."

Faced with the worry that teaching AI could lead to their own replacement, Xu Linfu brought the question back to whether personal capabilities can continuously iterate. In his view, no one can control being replaced; he can only make himself the kind of person who is always learning something new. "If the emergence of every new thing is enough to constitute a threat, then what eliminates you may be more than just AI."

A thousand drawings reaching the workshop still require human experience. AI tools have liberated the "hands" at the front end, allowing job changes to transmit all the way backward along orders, but for a rendering to land in the workshop, there is still a stretch of road that people must "push along." Lin Ke told reporters that when looking for clients on Xiaohongshu, he often sees people first posting an AI-generated rendering, then testing whether it can become an intended order based on views and intent deposits, and only after initial demand appears do they look for a factory. A shaped hair-drying cap, a throw pillow that does not yet exist—these are first tested by the market on screen, and only then sent to salespeople.

After taking an order, the next step is to verify the drawing. AI may draw embroidery effects on materials suitable only for printing; the image may look complete, but the workshop may not necessarily be able to produce it. In such cases, Lin Ke will let AI assist in estimating quotations for simple logos, but when it comes to custom sizes, special weights and embroidery stitch counts, he still has to find a master familiar with costs. He thus discovered that fast image generation and fast quotation can only win the first round of communication; to truly retain an order, what matters is the professional judgment accumulated in the workshop.

An image can be generated quickly on screen, but after entering the factory it must go through inquiry, price verification, sampling and sample revision, and then circulate among printing, embroidery and packaging. Especially under the "small orders, quick response" model, orders start at dozens or hundreds of pieces, styles are more scattered, revisions are more frequent, and the scheduling of outsourced factories cannot speed up along with AI. This mismatch in rhythm between online generation and offline implementation is ultimately reflected in the staffing structure.

Many small factories' recruitment sheets therefore show both subtraction and addition at the same time. Shu Kai, general manager of Zhejiang Yifan Daily Necessities Co., Ltd., remembers that previously a designer could complete at most two or three images a day. Now, AI can produce 800 to 1,000 candidate images a day, and a set of product images can be completed in about one hour at the fastest. Based on the original business volume, work that once required two or three designers to complete can now be handled by one person with the help of AI. At present, Yifan Daily Necessities has reduced recruitment for designer positions, and there are fewer and fewer operations assistants doing only basic tasks; as small-order customization increases, dedicated sample makers have been added, and Shu Kai also plans to recruit more capable supply chain managers to monitor production progress segment by segment.

Bosses hand-building AI employees. The driving force behind changes in positions and staffing structure is the owners of China's small factories. When Wu Xianmin, founder and CEO of Zhejiang Duopin Daily Necessities Co., Ltd., first used AI software, he first had it read through the materials on his computer, so company documents and the business judgments he had left entered the knowledge base, and the system also produced an analysis of him and the company. What he values is whether AI can accumulate memory. He treats AI as an employee whose experience can be continuously fed; the longer they spend together, the better it should understand the boss and the business.

This idea, however, did not receive a response from all employees. The company equipped salespeople with AI phones to record and summarize customer chats. But Wu Xianmin found that "the usage rate of the phones is not high; employees do not know how to use them, and they do not want to use them." To lower the threshold for use, Wu Xianmin began building a simpler control panel, breaking product selection, market research, product image generation, store operations and customer replies into skills that can be directly called. Employees upload images and click the corresponding function, and the system executes according to preset processes. Wu Xianmin used a set of proportions to describe the ideal division of labor—the first 10% is for people to think about what to do and whom to sell to, about 85% of standard execution in the middle is handed to AI, and the final 5% is checked by people.

Along these proportions, he envisions that one set of software plus one person can support a 1688 store. The system is still being built and tested, and there are still many problems to solve before it can stably cover the entire company, but Wu Xianmin has already pushed the goal from getting employees to use tools to getting AI to take on a set of job responsibilities. He also wants to make himself an AI boss. According to Wu Xianmin's plan, daily conversations, meetings and decisions will be recorded, and the system will then analyze which tasks were not executed and which arrangements were not logical, generating a review report. This AI cannot sign for him, nor does it bear business consequences; it is more like a questioner who never loses focus, bringing the boss's own judgments into the check as well.

Unlike Wu Xianmin, who writes experience into the system, Oukaisi general manager Wu Xiangju advances the transformation from within positions. Since the beginning of this year, Wu Xiangju has established an AI department, brought in AI engineers and digital talent, and required every employee to complete one AI application related to their position each month: HR developing interviews and performance systems, operations reforming store inspections, review replies and ad placement processes. According to Wu Xiangju's estimate, even if this year's performance doubles, Oukaisi's employee scale will remain at more than 200 people, while requirements for employee capabilities and educational backgrounds are rising.

Whether it is Wu Xianmin hand-building "AI employees" or Wu Xiangju requiring every position to re-deconstruct its own work, both point to the same organizational transformation: processes formerly completed through individual proficiency and departmental collaboration are beginning to be written into tasks that can be read, called and checked. Standard actions gather toward the system, while people retreat to more essential positions such as problem definition, exception handling and responsibility for results. Under the impact of AI, the owners of China's small factories have realized that for enterprises to achieve their next growth, what needs to improve is system capability, not simply addition or subtraction of manpower.

At 6 p.m., Jialin closes his laptop as usual; on some nights, Lin Ke still has to answer customer calls after 11 p.m.; in another office, Wu Xianmin continues adding materials and skills to his AI employee. Technological iteration will not stop, and AI's capabilities will continue to grow stronger. But AI cannot perceive the real feel of fabric, the subtext of customers, or the experience in the workshop that cannot be replaced by code. The relationship between people and tools still must slowly find its answer through daily磨合 in every office and every factory. When the next round of orders arrives, whether the factory will add a sample maker, an AI engineer, or simply stop hiring, no one can predict in advance. On the iterated recruitment sheet, every requirement corresponds to a small Chinese factory's renewed weighing of efficiency and human nature.