I have sat across the table from brand managers who told me AI was “on the roadmap for Q3.” That was eighteen months ago. Their Chinese competitors spent those eighteen months running 1,000 SKU tests per year with AI, cutting design team headcount by two-thirds, and building an intelligence layer that compounds every quarter. Western brands are not losing in China because of pricing or distribution. They are losing because they are scheduling training workshops while local players are already at L4 AI adoption. The gap is not technical. It is a willingness to start before everything is perfect.
What This Article Is About
This article is based on a Baidu whitepaper analyzed and published on Woshipm, covering AI adoption across Chinese e-commerce merchants in 2026. The research segments merchants by revenue and maps their AI usage across five levels, from no adoption to full AI-native operations.
The numbers that matter:
- 39% of merchants are stuck at L2: they use AI regularly but only for surface tasks. Copy generation, image resizing, basic Q&A. They feel like they are using AI. They are not getting the compound returns.
- Merchants above 1 billion yuan in revenue show 73% AI adoption at meaningful levels. Merchants below 1 million yuan: 54%. The gap is 19 percentage points. The bigger players are pulling further ahead, not because they have more money, but because they started earlier and built institutional knowledge.
- The optimal AI investment range for measurable returns: 500 to 2,000 yuan per month. Brands in this range see 10 to 30% performance improvement. Below that, the investment is too thin. Above it, you are paying for tools you have not learned to use yet.
Three real case studies from the whitepaper: Old Boss Appliances used AI to identify keyword opportunities competitors missed. Duoji Jewelry went from 100 SKUs tested per year to 1,000. Naison Furniture cut its design team from three people to one without reducing output. These are not pilot programs. They are operational realities for brands that started in 2024.
What the Brands That Are Actually Winning Do
The L4 and L5 merchants treat AI like infrastructure, not a feature. They did not wait for the perfect tool. They picked what was available, built workflows around it, and let the intelligence accumulate over time.
Old Boss Appliances did not ask their agency to run AI experiments. They built internal capability. Their team learned to use AI for keyword gap analysis, found opportunities their competitors were ignoring, and moved budget there. The tool was not the advantage. The speed of execution was.
Duoji Jewelry understood that AI’s real value in product testing is volume. You cannot test 1,000 SKU variations manually. With AI-assisted image generation and copy testing, you can. They identified winning products faster, reduced waste on slow movers, and reinvested in what worked. The cycle time for product decisions went from months to weeks.
Naison Furniture made the hardest call: they restructured the team. One designer with AI tools replaced three designers without them. That is a cultural and operational decision, not just a technology decision. Most Western brands will not make that call until they are forced to.
The real competitive advantage from AI is not access to the tools. Everyone has access. The advantage is the accumulated intelligence: the prompt libraries, the workflow automations, the institutional knowledge of what works for your specific products, categories, and customers. That only comes from starting early and building consistently.
Where Most Brands Go Wrong
L2 is the comfort zone with diminishing returns. Most brands that “use AI” are at L2. They asked ChatGPT to write product descriptions once. They used an image tool for a campaign. They feel modern. They are not building anything durable.
L2 usage does not compound. You get a slight efficiency gain on the tasks you already do. You do not fundamentally change what is possible. The brands at L4 are not doing the same things faster. They are doing things that were simply impossible before at their budget level.
The other failure mode: buying expensive tools without changing workflows. I have seen brands spend 50,000 yuan on an AI subscription they use for email subject lines. That is not an AI strategy. That is a procurement decision pretending to be transformation.
The third mistake: treating AI as a cost-cutting exercise only. Yes, you can cut headcount. But the brands that win ase AI to increase output volume and speed, not just reduce spend. Duoji Jewelry did not fire their product team. They multiplied what their product team could test. The frame matters.
What to Do This Week
1. Diagnose your AI level honestly. Are you at L1 (no use), L2 (surface use), L3 (workflow integration), or L4 (compounding intelligence)? Most teams that think they are at L3 are at L2. Be honest.
2. Pick one workflow to fully rebuild with AI. Not “use AI more.” Rebuild a specific workflow end to end. Product copy: from brief to final text, AI-assisted. Keyword research: from category to long-tail gaps, AI-driven. One workflow, fully rebuilt, in the next 30 days.
3. Set a monthly AI budget between 500 and 2,000 yuan. The data says this range produces the best ROI. If you are spending less, you are not testing seriously. If you are spending more without workflow clarity, you are wasting it.
4. Document what you learn. Every prompt that works, every workflow that saves time: write it down. This is how you build the intelligence layer that compounds. It is not in the tool. It is in your team’s accumulated knowledge.
5. Stop waiting for the industry-standard AI playbook. It does not exist yet. The brands building competitive advantage are writing it right now, through testing. Start testing.
GMA helps foreign brands build AI-ready e-commerce operations in China. See what that looks like on our China e-commerce agency page.
Sources
- Woshipm: Baidu AI Adoption Whitepaper for Chinese E-Commerce Merchants (2026) – Analyse de l’adoption IA par niveau de revenus, cas concrets Old Boss Appliances, Duoji Jewelry, Naison Furniture, et ROI par tranche d’investissement.
- Alibaba Group Investor Relations – Donnees officielles sur l’adoption IA dans l’ecosysteme Alibaba, Taobao et Tmall, publiees dans les rapports trimestriels 2025-2026.
- 36Kr – Media tech chinois couvrant l’adoption IA dans le e-commerce, les startups et les grandes marques chinoises avec analyses regulieres du marche.
Further Reading
- Xiaohongshu Organic vs Paid Traffic: What Works for Foreign Brands in 2026
- Tmall 618 2026: What the Rankings Tell You About China’s Market Right Now
- China’s Emotional Economy: What the 2.31 Trillion Yuan Market Means for Western Brands
Your competitors are not smarter. They just started testing 6 months before you.
Marcus Zhan
About Marcus Zhan
Marcus Zhan is Partner at GMA, a China digital marketing agency founded in 2012 in Shanghai. GMA has worked with over 600 foreign brands on Xiaohongshu, WeChat, Douyin, Tmall, and Baidu. He has helped foreign e-commerce brands build AI-ready operations in China at a time when most were still debating whether to start. He has no patience for brands that plan for months before running a single test. Follow his work on LinkedIn: “15 years in China marketing. No patience for excuses.”
