The first time I heard Huang Xufeng say “170 to 50,” I thought I’d misheard. “You went from 170 employees to 50,” I repeated. “And revenue went from 100 million to 300 million.” He nodded. “That’s right.”
I checked my notes. Then I checked my notes again.
Six months later, I heard Zhang Feng say something similar. From 5 square meters to 40 million RMB in annual revenue. From his wife spending two hours per day on data entry to twenty minutes. From a team that couldn’t scale to a business that could.
Both men run traditional industries—wholesale clothing, cross-border e-commerce. Both achieved what should be impossible: more output, fewer people, more money. And both attribute the same thing as the catalyst: AI.
Not in the way you’d expect. Not “we bought AI and got results.” More like “we found the parts of our business that felt stupid, and AI made them disappear.”
The Pattern Nobody Talks About
There’s a meme in Chinese business circles right now. It’s about the “30% Club”—founders who’ve achieved roughly 30% annual growth while cutting headcount significantly. It’s considered impossible by old-guard business advisors. “You have to choose,” the conventional wisdom goes. “Either you grow, or you automate. You can’t do both.”
Except these founders are doing both. Consistently.
Zhang Feng’s warehouse in Guangzhou still looks like a traditional wholesale operation from the outside. But inside, his wife no longer spends two hours entering SKU data. She spends twenty minutes. The other hundred minutes? She focuses on what only humans can do—curation, relationship management, strategic decisions about what to stock.
Huang’s cross-border operation processes data from 187 countries automatically. When a listing’s conversion rate drops, the system pushes an alert to the relevant person before they’ve even opened their laptop. Decision cycle: from 4-6 hours to real-time. “Before, you had to actively look for problems,” he explained. “Now the problems find you.”
The pattern is consistent across both stories: identify the work that is fundamentally automatable—repetitive, rule-based, high-volume—and build systems that handle it. Then redirect human energy toward judgment, creativity, and relationship-building.
Why It’s Terrifying
Here’s why most business owners won’t do this, even when they see it working.
Automation requires admitting that some of your best employees are doing work that machines can do. And if those employees have been with you for years—if they’ve earned your trust and loyalty—the admission feels like betrayal.
Zhang talked about this candidly. His wife was doing “machine work” for years before they found AI.录入 data, managing inventory, repetitive tasks that required no judgment. She was good at it. She was dedicated. But she was also working late every night while the business struggled to scale.
“Once we automated that part,” Zhang said, “she finally had time to do the work that actually mattered—her judgment, her relationships, her understanding of what customers wanted.”
The terror comes from a different direction too. When you automate a job, you eliminate a job. And in China especially, there’s an emotional weight to that. “These people trusted us,” a business owner might think. “We owe them.”
But here’s the reframe that neither founder mentioned explicitly, but both demonstrated: keeping someone in a job that machines should do isn’t loyalty. It’s slow-motion betrayal. You’re keeping them in a role that has no future, while the market moves past the skills they’re developing.
The founders who succeed at this aren’t the ones who automate and lay off. They’re the ones who automate and redirect. Huang’s team went from 170 to 50—but those 50 people are doing work that requires actual human judgment. The company grew, but it also became a place where humans do human work, not machine work pretending to be human work.
The Real Starting Point
Both founders mentioned the same first step, independently: SOP.
Before Zhang built his AI inventory system, he documented every step of the入库 process. Before Huang automated his listing pipeline, he had a 500-page SOP for how to upload a single product.
This feels tedious. It feels like “pre-work” that slows you down. But it’s the opposite—it’s the foundation that makes everything else possible.
“Most companies fail at AI adoption because they think you can skip this step,” Huang told me. “They want the magic tool. But you can’t automate what you haven’t documented.”
The logic is simple: if you don’t know exactly how a process works—step by step—you can’t write the rules that let a machine do it. And you can’t improve the process by automating it if you don’t understand it in the first place.
Zhang’s 500-page SOP wasn’t just documentation. It was discovery. When he mapped out every step of listing a product, he found that 60% of the steps were either redundant, automatable, or indicative of a deeper organizational problem. “The document was supposed to help employees,” he said. “But it ended up helping me understand that the process itself was broken.”
The Learning Tax
Both founders paid what I call a “learning tax”—time and money invested upfront that won’t show returns for months or years.
Huang spent 9 hours per employee getting 160 people UiPath certified. That’s roughly 1,440 hours of learning time before a single automation was deployed. Zhang spent months building his inventory system before it connected to anything else. Neither approach produces immediate ROI.
But the learning tax compounds. Once Huang’s team understood RPA, they started identifying automation opportunities everywhere. Once Zhang’s systems were connected, new possibilities opened up that he hadn’t imagined.
The learning tax is also a filtering mechanism. If you’re not willing to pay it, you’re not serious about transformation. You’re just looking for someone to tell you which magic tool to buy.
“Most small business owners want the big lobster,” Huang said, using a Chinese idiom for expecting a windfall. “They want the solution to fall from the sky. But transformation doesn’t work that way.”
What Stays Human
Neither founder treats AI as a replacement for human judgment. They treat it as a replacement for human drudgery.
Zhang still curates products with his wife. The decision about what to stock, what to push, what to drop—that’s still human. Huang still has a team doing selection—figuring out which products to launch on which platforms. The data informs the decision, but the decision is human.
The automation handles the work that would otherwise limit scale. With Huang, that’s listings, monitoring, data aggregation. With Zhang, that’s入库, customer follow-ups, inventory management.
The key insight: these founders didn’t try to replace humans with AI. They tried to remove the ceiling that prevented humans from doing meaningful work.
“I don’t want my wife working until midnight entering data,” Zhang said. “I want her making decisions that require her judgment and experience. The data entry was just taking up space in her head.”
The 2026 Goal
Huang has a specific target for 2026: 95% automation of routine operations. His vision is an “operating system” where he tells the AI “I want to launch this product on 5 platforms, with these margin targets, expect a report in 3 months”—and the system handles everything in between.
It’s ambitious. But the foundation he’s building—data infrastructure, SOPs, internal tools, trained team—makes it achievable.
Zhang is focused on productizing what he’s built. “My tools solve my problems,” he said. “Now I’m finding they solve problems for my customers too.”
Neither approach is fast. Neither is glamorous. Both require the kind of patient, systematic work that most business advisors won’t talk about because it doesn’t fit in a keynote slide.
The Question Nobody Asks
When I listen to these founders, I keep coming back to the same question: why aren’t more people doing this?
The answer I’ve settled on: it’s not about intelligence. It’s about tolerance for uncertainty.
Building an automation system means admitting you don’t know exactly what you’re building. It means experimenting, failing, iterating. It means your team asking “why are we spending time on this instead of selling?” and not having a satisfying answer for months.
Most business owners have survived by being certain. Certain about their product, their market, their customers. That certainty is a liability when the world changes.
The founders in the 30% Club? They’re comfortable not knowing. They’re comfortable saying “let’s try this and see.” They’re comfortable with the ambiguity of building something that doesn’t exist yet.
“I was terrified at first,” Zhang admitted. “I didn’t know if any of this would work. But I knew the old way wasn’t sustainable. So I just kept going.”
What I Keep Noticing
In both conversations, the same phrase kept coming up: “找到爽感”—find the feeling of breakthrough.
Huang describes the moment he first saw his computer running automation scripts at 2am: “You watch the screen and the machine is working for you. The feeling is incomparable.”
Zhang describes the moment his wife stopped working at midnight: “She could finally sleep. And the business kept running.”
These aren’t just efficiency gains. They’re liberation. From work that machines should do, back to work that humans should do.
The 30% Club isn’t about automation. It’s about getting humans back to human work.
FAQ
Q: How can traditional businesses achieve 30% annual growth while cutting staff?
A: The key is identifying which work is fundamentally automatable—repetitive, rule-based, high-volume—and building systems that handle it. This frees human capital for judgment, creativity, and relationship-building. Both Zhang and Huang achieved this by: 1) documenting all processes as SOPs, 2) automating routine operations via RPA/AI, 3) redirecting human energy to higher-value work, and 4) building data infrastructure that enables continuous optimization. The results: 70% headcount reduction with 2-3x revenue growth.
Q: Why do most AI transformations fail in small businesses?
A: Most small business owners want a “magic tool” solution and skip the foundational work: documenting SOPs, cleaning data, understanding process boundaries. Huang Xufeng puts it bluntly: “They want the big lobster—the solution to fall from the sky and solve everything at once.” Without documented processes, you can’t automate them. Without data infrastructure, AI has nothing to analyze. Without team-wide understanding of capabilities, automation requests don’t get communicated effectively. The learning tax (9+ hours per employee for RPA certification in Huang’s case) is seen as optional rather than foundational.
Q: What’s the actual first step for traditional businesses starting AI adoption?
A: Don’t start with AI. Start with SOP documentation. Map out every step of your most painful, repetitive process. Zhang Feng built a 500-page SOP for listing products; Huang documented 60+ steps for product uploads. This discovery process reveals what’s broken, what’s redundant, and what can actually be automated. Only after documenting the process can you identify which parts should be automated—and automation requires clear rules, not creative freedom. The SOP is the foundation everything else is built on.
Q: How does RPA differ from AI in business transformation?
A: RPA (Robotic Process Automation) handles rule-based, repetitive tasks—clicking, typing, following structured workflows. It’s deterministic: same input produces same output. AI handles pattern recognition, prediction, and judgment-based decisions. In practice, the most successful transformations use both: RPA for execution (uploading listings, moving data between systems), AI for optimization (identifying which listings underperform, predicting inventory needs, pushing real-time alerts). Huang’s company uses RPA to pull platform data into dashboards, then AI to analyze patterns and trigger actions—combining deterministic automation with intelligent decision support.
Q: What makes these founders different from others trying AI transformation?
A: Both Zhang Feng and Huang Xufeng share three traits: 1) They started by solving their own problems before selling solutions—internal validation before external productization; 2) They embrace the “learning tax”—investing months of upfront time before seeing ROI (Huang’s 1,440 hours of team training, Zhang’s months of system integration); 3) They treat AI as liberation for humans, not replacement of humans—automating machine work so humans can do human work. Most failed transformations try to skip steps, demand immediate ROI, and view automation as cost-cutting rather than capability-building.