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 businesses, 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 demanding efficiency from AI, and from obsessing over big orders to embracing small ones, constantly completing their own self-iteration.

At 6 p.m., Jialin, an operations assistant at a manufacturing company in Zhejiang, closes his laptop and ends his workday. Jialin's work includes store data processing, product maintenance, and product listing, which used to keep him busy from 8:30 in the morning until the afternoon. After integrating AI, these trivial tasks that once filled most of the day have been compressed to three or four hours, freeing him to devote a large amount of time to new product development and backend ad投放.

Jialin has stepped away from repetitive labor, but salesperson Lin Ke does not find it so easy to "clock off." Some of his clients are still in school 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 custom product for over an hour." According to Lin Ke, clients usually arrive with an AI-rendered image, but when they reach the workshop entrance, it may not be reproducible because the fabric and craftsmanship are different. He has to translate the imagination piled up by AI, item by item, into production parameters that the workshop can understand.

The daily lives of Jialin and Lin Ke are exactly the most authentic cross-section of frontline small-factory employees under the AI wave, and such scenes are playing out across industrial belts in places like Yiwu, Yongkang, and Keqiao. Data from the Yiwu China Commodities City 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 from 2022 to 2026, and further estimates based on 40.01 million online recruitment demands from January to April 2026 show that 26.6% of positions fall into the high-exposure, low-complementarity substitution pressure zone, meaning one in every four positions is standing at the threshold of structural adjustment.

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

Becoming a workplace newcomer once again. In the AI era, the tasks and roles of positions have both changed. Jialin still sits at his original desk, 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 do not 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 admits that when buyers are also using AI to generate product ideas, human professionalism becomes even 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 this 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 business unit 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 having studied tourism management, entering e-commerce was already a career turn. When AI entered the company, he once again stood before an unfamiliar working language. Jing Jianghuan admits that at first, when the post-2000s colleagues around him began trying new tools, he was somewhat resistant. "It has nothing to do with whether the tool is difficult or not, but rather not wanting to learn something new from scratch." But after he truly studied it, he found that AI really "can give a person a lift"—information that originally needed repeated sorting can be quickly gathered together. He eventually settled down, starting from the basic concepts and operating models of AI, and also tried letting AI teach him about AI, then gradually experimenting through trial and error.

Xu Linfu, head of the Pinduoduo project at Oukaisi, has an even deeper appreciation of how AI tools improve efficiency. 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 listing may accumulate fifty or sixty thousand reviews, and just breaking down dozens of links could take ten days or even half a month. After integrating AI, it can be completed in a morning or an afternoon. Xu Linfu estimates that the knowledge base and analysis tools have saved 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. "If 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 through which young people became familiar with the e-commerce business, but this kind of repetitive work is also the first to be hit by AI. Research by ShanghaiTech University's CEISD shows that among newly added online recruitment demand, positions with monthly salaries of 5,000 to 8,000 yuan and requiring 1 to 3 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's opinion. Technology will not wait for everyone to be ready; those who drove horse carriages can only relearn how to drive cars. It is just that this time, the new driver's cab falls on a computer screen."

Facing the worry of teaching AI and then being eliminated themselves, Xu Linfu brings the question back to whether personal ability can continue to iterate. In his view, no one can control being eliminated; he can only make himself the person who is always learning new things. "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 blueprints reach the workshop, but human experience is still needed. AI tools have liberated the "hands" at the front end, allowing changes in positions to continue traveling backward along the order chain, but there is still a stretch of road that people must "push along" before a rendering can land in the workshop. Lin Ke told reporters that when looking for clients on Xiaohongshu, he often sees people first posting an AI rendering, then testing whether it can become an intended order based on views and intent deposits, and only after preliminary demand exists do they look for a factory. A uniquely shaped hair-drying cap, a pillow that does not yet exist—first they undergo market testing on the screen, and only then are they brought to the salesperson.

After taking the order, the next step is to verify the drawing. AI will 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 have AI assist in estimating quotes for simple logos, but when it comes to custom sizes, special fabric weights, and embroidery stitch counts, he still has to find a master familiar with costs. From this he discovered that fast image generation and fast quoting can only secure the first round of communication; truly retaining an order still depends on 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, sample making, 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 schedules of outsourcing factories cannot speed up along with AI.

This rhythm mismatch between online generation and offline implementation is ultimately reflected in the staffing structure. Many small factories' hiring 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 run out 800 to 1,000 candidate images a day, and a set of product images can be completed in as little as about an hour. Based on the original business volume, work that previously required two or three designers to complete together 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-batch customization increases, dedicated sample makers have been added, and Shu Kai also plans to recruit stronger 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 bosses 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, and the company's documents and the business judgments he had left entered the knowledge base, with the system also providing 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, and 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 client chats. But Wu Xianmin found that "the usage rate of the phones is not high; employees do not know how to use them, and 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 invoked. Employees upload images and click the corresponding function, and the system executes according to the preset process. Wu Xianmin described the ideal division of labor with a set of proportions—the first 10% is for people to think about what to do and to whom to sell it, about 85% of the standard execution in the middle is handed to AI, and the final 5% is checked by people. Following this set of proportions, he envisions that one set of software paired with 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 build himself an AI boss. According to Wu Xianmin's vision, daily conversations, meetings, and decisions will be recorded, and the system will then analyze which tasks were not executed and which arrangements were illogical, 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 all day, bringing the boss's own judgments into the inspection as well.

Unlike Wu Xianmin, who writes experience into the system, Oukaisi general manager Wu Xiangju advances reform from within the positions themselves. 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 own position each month: HR developing interviews and performance systems, operations transforming store inspections, review replies, and ad投放 processes. According to Wu Xiangju's estimate, even if this year's performance doubles, Oukaisi's workforce will remain at just over 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 that previously relied on individual proficiency and departmental collaboration are beginning to be written as tasks that can be read, invoked, and checked. Standard actions gather toward the system, while people retreat to more fundamental positions such as problem definition, exception handling, and responsibility for results. Under the impact of AI, the bosses of China's small factories have realized that for a company to achieve its next growth, what needs to improve is system capability, not simply adding or subtracting manpower.

At 6 p.m., Jialin closes his laptop as usual; on some nights, Lin Ke still has to pick up client calls after 11 p.m.; in another office, Wu Xianmin continues adding materials and skills to his AI employees. 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 a client's words, or the experience in the workshop that cannot be replaced by code. The relationship between people and tools still has to be slowly figured out 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 hiring sheet, every requirement corresponds to a small Chinese factory's renewed weighing of efficiency and humanity.