Picture this: an accountant’s office floor completely covered in paper forms—hundreds of them, from nearly 800 workers. Every piece of data about production, materials, equipment—all handwritten, then manually entered into a system, then double-checked. The deputy plant manager stood at the door and asked himself one question: How many human hours are we burning on something a machine should have done years ago?
That’s the scene that kicked off a quiet revolution at Jinhong Crankshaft, a factory in Sichuan that started as a gear factory in 1957. Clients like Chery, BYD, and Geely. 25 production lines. 800 workers. And an IT system that was essentially paper and hope.
But here’s the thing: the revolution didn’t start with robots, or a big-budget digital transformation project. It started with one person and a single spreadsheet.
Meet Lü Wang (coworkers call him "Wang-zai"), a 90s-born accountant. His job: collect daily production data from 170 workers in Workshop 6, type it all into the system. Sounded simple—except every worker wrote their numbers on paper. Every piece of paper had to be manually read, interpreted, and entered. That took him two to three hours every single day. And the errors? Endless. A worker writes “yuántǒngtǒng” (Sichuan dialect for “filter cylinder”), but the system uses a code nobody remembers. Or a machine leak report gets lost in a stack and surfaces weeks later. One cost engineer recalled manually checking 10,000 records for a single audit—his hand cramped for days.
In early 2024, the company introduced Feishu (Lark), a collaboration tool. Most people saw it as just another chat app. But Lü Wang discovered the multi-dimensional table feature. A lightbulb went off: What if workers filled in their own data, so I don’t have to type it?
He tried it. And it worked. Instantly. Those two to three hours of manual entry disappeared. He felt that “why didn’t I find this earlier” shock.
He didn’t stop. He built a performance review table, a material requisition table, a labor competition tracker. Each table saved somebody hours. Soon he had 30 automated workflows covering cost, quality, wages, equipment, materials—pretty much the entire core operation of Workshop 6.
But then a new problem popped up: too many tables. Workers couldn’t remember where to go. So Lü Wang built an AI bot—nicknamed "Xiao Liu" (Little Six)—using Feishu’s agent feature. Workers could just speak to their phone in Sichuan dialect, say “yuántǒngtǒng,” and the system knew what they needed. No Mandarin. No codes. No running to the accountant.
Now, you might think: “So a capable guy automated his own job. Big deal.”
But the real challenge came next: getting everyone else to use it.
Jinhong has about 800 workers; many are over 40, set in their ways. Resistance came from three directions:
- Habit resistance. “I’ve done it this way for 10 years. Why change?”
- Cognitive resistance. Middle managers thought AI was for tech companies, not old factories.
- Interest resistance. Data transparency made it harder to slack off. Some even sent emails requesting exemption from the new system.
Lü Wang’s strategy? A mix of soft and hard. Soft: he recorded tutorial videos, taught people one-on-one, posted them in group chats. Hard: he tied data entry to salary calculation. “If you don’t fill in your output in the system, I can’t calculate your pay.” Reasonable, isn’t it? You want your wage based on data? The data has to come from somewhere—and it’s not going to come from a paper that might get lost.
The story teaches something practical: the best digital transformation starts from the ground up. Not from a CEO’s vision document, but from one person’s daily pain point. Wang didn’t need an MBA in digital strategy. He just needed to notice a tool that was already there, and have the courage to try it.
Here’s the actionable takeaway: if you have a repetitive, data-heavy task that eats your time, don’t accept it as inevitable. Look for a tool you already have—a spreadsheet, a bot builder, a simple form—and ask: “Can a machine do this instead?” That’s the first step of 知行合一, knowing and doing as one.
The 67-year-old factory didn’t need a miracle. It needed a 90s accountant who refused to waste his hours on what a machine could do. And maybe that’s the closest thing to a practical guide for AI in the real world.