AI Rewrites Recruitment Forms for Small Factories: An 8-Person Team Manages 30 Stores, Produces 800 Images Per Day, Yiwu Business Owners "Manually Craft" Digital Employees
Editor's Note: Against the backdrop of volatile international trade, a sluggish consumer market and the sweeping AI wave, small Chinese manufacturers are quietly shifting gears. Zhejiang province is home to a large number of hidden champions, manufacturing enterprises, supply chain enterprises, foreign trade enterprises, specialized market operators and consumer brands. Though seemingly not at the "center of the stage", these small Chinese manufacturers form the solid foundation of China's manufacturing sector. They keep iterating themselves, moving from earning OEM fees to building their own brands, from relying on manpower to tapping AI for higher efficiency, and from chasing large orders to embracing small-batch customization.
The Time Weekly has launched the special feature "Small Chinese Manufacturers" to focus on these units at the end of the supply chain, and record the most vivid business stories and changes taking place.
At 6 p.m., Jialin, an operation assistant at a manufacturing company in Zhejiang, closed his laptop and wrapped up his work for the day.
Jialin's job includes store data processing, product information maintenance, product listing and other related tasks, which used to keep him busy from 8:30 a.m. all the way to the afternoon.
After the company adopted AI tools, these trivial tasks that once took up most of the day are now compressed into only 3 to 4 hours, leaving him plenty of time to devote to new product development and backend traffic investment operations.
While Jialin has been freed from repetitive labor, Lin Ke, a sales representative, cannot easily "clock off work". Some of his clients are still students, who only have time to discuss products after 10 p.m. "The latest time I took a call was after 11 p.m. from a high school student, and we communicated about a customized product for more than an hour." Lin Ke said.
According to Lin Ke, clients usually come with an AI-generated render, but when they arrive at the factory workshop, the design may not be replicated due to differences in fabric or craft. He has to translate the AI-generated imaginative design item by item into production parameters that the workshop workers can understand.
The daily routines of Jialin and Lin Ke are the most authentic portrayal of frontline employees at small manufacturers amid the AI wave, and similar scenarios are playing out in industrial belts across Yiwu, Yongkang, Keqiao and other regions. Data from Yiwu China Commodity City Group shows that nearly 30,000 local merchants have been using various AI tools on a regular basis.
According to *China's Labor Market in the AI Era* released by the Center for Education, Innovation and Sustainability Development of ShanghaiTech University, the research team built a database with 743 million recruitment records from 2022 to 2026, and further calculation based on 40.01 million online recruitment demands from January to April 2026 shows that 26.6% of jobs fall into the high-exposure, low-complementarity zone under substitution pressure, meaning one in every four jobs is at the crossroads of structural adjustment.
Small Chinese manufacturers are now standing at the crossroads of this round of technological renewal: which experiences can be programmed into systems, and which judgments must be reserved for humans? When business owners build AI employees by themselves, what can ordinary workers do to secure a place in the next recruitment list?
Become a Workplace Newcomer All Over Again
In the AI era, the tasks and roles of jobs have all changed.
Jialin still sits at his original workstation, and his job title remains unchanged, but his work content has been completely rearranged.
He said that even with the help of AI, he needs to re-evaluate which tasks do not require full manual involvement, and which tasks have become more important. "Some very basic work, including operation, product listing, and data processing, I think can be fully handled with the help of AI."
However, Jialin admitted that when buyers also use AI to generate product ideas, human professionalism becomes even more critical, because some AI-generated product effects cannot be realized in actual factory production.
"For example, a client wants a pattern design, but the product can only support printing instead of embroidery in practice, while AI gives them an embroidery effect. In this case, we need to provide more professional suggestions to the client."
For Jialin, the balance of his work is shifting: standardized tasks on one end are taken over by AI, while the other end is filled with responsibilities that are harder to be defined by fixed rules. He has to learn faster and make more accurate judgments.
△ Image source: Tuchong
For Jing Jianghuan, acting director of the New Media Home Care Division at Zhejiang Oukesi Technology Co., Ltd., this change is most prominently reflected in the sense of freshness of being a "newcomer" all over again.
33-year-old Jing Jianghuan graduated ten years ago. Majoring in tourism management, entering the e-commerce industry was already a career shift for him. When AI was introduced to the company, he once again faced a set of unfamiliar work terms and logics.
Jing Jianghuan admitted that he was a little resistant at first when his post-2000s colleagues around him started to try new AI tools. "It's not about whether the tool is difficult to use, but I just didn't want to learn a completely new thing from scratch."
But after he really started learning, he found that AI can indeed "lift people to a higher level": information that used to require repeated sorting can be gathered quickly. He finally settled down to learn the basic concepts and operation modes of AI, tried to let AI teach him how to use AI, and gradually practiced and made mistakes on his own.
Xu Linfu, head of the Pinduoduo project at Oukesi, has a deeper experience of the efficiency improvement brought by AI tools. According to Xu Linfu, although his team only has 8 operation staff, they can maintain around 20 to 30 stores.
He revealed that a single competitor product link may accumulate 50,000 to 60,000 reviews, and it used to take 10 days or even half a month to disassemble dozens of such links. After adopting AI, the whole process can be finished within one morning or afternoon.
Xu Linfu estimated that the knowledge base and analysis tools saved the team about 50% to 60% of the working time. Meanwhile, he said this efficiency gain has been taken into the company's manpower planning. "As business continues to grow, the team does not need to expand in the original proportion."
Listing products, sorting out forms, and processing replies used to be the entry point for young people to get familiar with e-commerce business, but these repetitive tasks are the first to be impacted by AI. The research by CEISD, ShanghaiTech University shows that among the new online recruitment demands, positions with a monthly salary of 5,000 to 8,000 RMB and requiring 1 to 3 years of work experience are under the greatest pressure.
However, while AI brings changes to small manufacturers, it also brings job anxiety.
In response to this, Jing Jianghuan said frankly: "When the automobile era arrives, no one will ask for the opinions of carriage drivers. Technology will not wait for everyone to be fully prepared. People who used to drive carriages can only learn how to drive cars all over again. Only this time, the new 'driver's cab' is on the computer screen."
Facing the concern that people may be eliminated after training AI themselves, Xu Linfu attributed the problem to whether individual abilities can keep iterating. In his view, no one can avoid being eliminated by the times, and the only thing he can do is to keep learning new things. "If every new thing that emerges is enough to become a threat, then AI will not be the only thing that eliminates you."
Human Experience Is Indispensable for Turning Thousands of Drawings into Actual Workshop Outputs
AI tools have freed the frontline workforce from repetitive manual work, and the changes of jobs are transmitted downstream along the order chain. But there is still a long way to go for a design render to be turned into actual production in the workshop, which requires human efforts to push the process forward.
Lin Ke told the reporter from *The Time Weekly* that when he looks for clients on Xiaohongshu, he often sees people releasing an AI design render first, testing whether it can be turned into a potential order according to page views and earnest money from potential buyers, and then looking for factories after preliminary demands are confirmed. A special-shaped hair drying cap, a pillow that does not even exist in the market, will first be tested by the market on screen, and then sent to sales representatives.
After receiving the order, the next step is to verify the design drawing. AI may generate embroidery effects on materials that are only suitable for printing. The picture looks complete, but the workshop may not be able to produce it in reality.
In this case, Lin Ke will let AI assist in estimating the quotation for simple logos, but for customized sizes, special fabric weights and embroidery stitch counts, he still needs to consult skilled workers who are familiar with cost calculation. He thus found that fast drawing and fast quotation can only help win the first round of communication. To actually secure the order, it still relies on the professional judgment accumulated in the workshop.
A design can be generated quickly on the screen, but after entering the factory, it has to go through processes including inquiry, price verification, proofing, and sample modification, and circulate between printing, embroidery and packaging links. Especially under the "small-batch quick response" mode, orders usually start from dozens or hundreds of pieces, with more scattered styles and more frequent modifications, and the production scheduling of cooperative factories cannot be accelerated along with AI speed.
This rhythm mismatch between online generation and offline implementation is ultimately reflected in the employment structure. Many small manufacturers have made both cuts and additions to their recruitment lists.
△ Image source: Tuchong
Shu Kai, General Manager of Zhejiang Yifan Daily Necessities Co., Ltd., recalled that in the past, a designer could complete at most 2 to 3 drawings a day. Now AI can generate 800 to 1000 candidate drawings a day, and a full set of product images can be completed in as fast as about an hour. For the original business volume, the work that used to require 2 to 3 designers to cooperate can now be handled by one person with the help of AI.
At present, Yifan Daily Necessities has reduced the recruitment of designer positions, and operation assistants who only do basic tasks are getting fewer and fewer. After small-batch customization business increased, the company added full-time proofing staff, and Shu Kai also plans to recruit more capable supply chain management personnel to track the production progress at each stage.
Business Owners Build AI Employees on Their Own
The driving force behind the changes in jobs and employment structure are the owners of small Chinese manufacturers.
When Wu Xianmin, Founder and CEO of Zhejiang Duopin Daily Necessities Co., Ltd., used AI software for the first time, he asked it to read all the documents on his computer. The company's files and his accumulated business judgments were imported into the knowledge base, and the system also generated analysis of him and the company.
He values whether AI can accumulate memory. He regards AI as an employee that can be continuously fed with experience, and the longer they work together, the better the AI should understand the business owner and the operation of the company.
However, not all employees responded positively to this idea. The company distributed AI mobile phones to sales representatives to record and summarize customer chat records. But Wu Xianmin found that "the mobile phones are not widely used, employees can't operate them, and they don't want to use them."
To lower the usage threshold, Wu Xianmin started to build a simpler control panel, splitting product selection, market research, product image generation, store operation and customer reply into directly callable functions. Employees only need to upload images and click the corresponding function, and the system will execute according to the preset process.
Wu Xianmin described the ideal work allocation with a set of proportions: the first 10% of the work, that is, deciding what to produce and who to sell to, is completed by humans, the middle 85% of standard execution work is handed over to AI, and the last 5% of the work is checked by humans. Following this allocation, he envisions that one set of software paired with one employee can support a whole 1688 store.
The system is still under construction and testing, and there are many problems to be solved before it can stably cover the whole company, but Wu Xianmin has already shifted his goal from making employees able to use tools to letting AI take over a series of job responsibilities.
△ Image source: Tuchong
He also wants to make an AI version of himself as the business owner. According to Wu Xianmin's vision, daily conversations, meetings and decisions will be recorded, and the system will analyze which tasks are not executed and which arrangements are illogical, to generate a review report. This AI cannot sign documents for him, nor does it bear business risks. It is more like a non-distracted supervisor all day long, who also includes the business owner's own judgments into the check scope.
Different from Wu Xianmin who writes his experience into the system, Wu Xiangju, General Manager of Oukesi, promotes the transformation from inside each position.
Since the beginning of this year, Wu Xiangju has established an AI department, introduced AI engineers and digital talents, and required every employee to complete one AI application related to their own position every month: the HR team developed AI interview and performance management systems, and the operation team transformed the store inspection, comment reply and traffic investment processes with AI.
According to Wu Xiangju's forecast, even if the company's performance doubles this year, the number of Oukesi's employees will still remain at more than 200, while the requirements for employees' capabilities and educational background are on the rise.
No matter Wu Xianmin builds "AI employees" by himself, or Wu Xiangju requires every position to re-deconstruct their own work, they both point to the same organizational transformation: processes that used to rely on personal proficiency and inter-departmental collaboration are being programmed into readable, callable and verifiable tasks. Standard operations are gathered into the system, while humans are pushed back to the more essential responsibilities including problem definition, exception handling and result accountability.
Under the impact of AI, the owners of small Chinese manufacturers realize that to achieve the next round of growth, enterprises need to improve their system capabilities, rather than simply adjust the number 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 keeps adding materials and functions to his AI employees.
Technological iteration will never stop, and the capabilities of AI will continue to improve. But AI cannot perceive the real hand feel of fabrics, the implied meaning of customers' words, and the set of experience in the workshop that cannot be replaced by codes. The relationship between humans and tools still needs to be explored slowly in the daily running-in of every office and every factory.
When the next batch of orders arrives, no one can predict in advance whether the factory will recruit a proofing specialist, an AI engineer, or simply not hire new staff at all. Every demand on the iterated recruitment list represents a re-evaluation of efficiency and humanity by a small Chinese manufacturer.
(Jialin and Lin Ke are pseudonyms)
Reporter | Liu Ting
Intern | Zhang Xiaorui
Editor | Yang Chunxia
Operation | Song