China’s AI Gap Is Organisational

The striking thing about China’s AI race is not that Chinese people are slow to adopt new tools. It is the opposite. Consumer uptake can look almost manic. Zack Kass, who led go-to-market work at OpenAI before starting his own consultancy, told the *South China Morning Post* that China has a “techno-centric consumer” culture even as its companies remain slower than American firms to absorb AI into everyday work. He pointed to a recent frenzy around an AI agent product called OpenClaw, which drew crowds to Tencent’s Shenzhen headquarters and even helped push up Mac Mini prices in China. (scmp.com) That split matters because it changes the usual story. The bottleneck is not simply chips, models, or talent. China has all three in serious quantity. The gap shows up when a company has to let many teams test a tool, change a workflow, and make a decision before the quarter is over. Kass’s argument is that Chinese firms often struggle there. Hierarchies are steeper. Approval chains are longer. Managers are less willing to let small teams run loose with unfinished systems. (scmp.com) A new survey from Deloitte China and the University of Hong Kong lands in almost the same place. It asked more than 100 C-suite leaders across mainland China and Hong Kong about enterprise AI. More than one-third of organizations said they were still only exploring AI. Another 56 percent said they were in limited implementation. Just 23 percent reported measurable financial impact, and only 4 percent described their use as transformational. The report is blunt about why. Most initiatives stall because of silos, cultural resistance, weak governance, poor data, and vague business cases rather than missing technology. (deloitte.com) That diagnosis also helps explain why the United States still looks stronger in enterprise AI even when consumer excitement is weaker. OpenAI’s own enterprise report says weekly messages in ChatGPT Enterprise increased roughly eightfold over the prior year, while use of structured workflows such as Projects and Custom GPTs rose nineteenfold. Workers in surveyed companies said AI improved the speed or quality of their output in three out of four cases. Deloitte’s U.S. enterprise survey finds the same general pattern at a broader level: worker access to AI rose by 50 percent in 2025, and companies are moving from pilots toward production, even if many still feel operationally unprepared. (openai.com) China’s enterprise lag is especially visible in AI agents, the software category meant to do multistep work rather than just answer prompts. A July 2025 report cited by the *Post* said overseas AI agent makers were ahead of Chinese firms by an order of magnitude in orders and revenue. The reason was not superior imagination. It was higher corporate IT budgets and more mature digital infrastructure among their customers. Chinese firms, the report said, were still dealing with weaker digitization and tighter spending. (scmp.com) This is why the most useful way to think about the U.S.-China AI contest is not as one race on one track. Some analysts now describe two different races. The U.S. still has much deeper private capital and a more mature enterprise software market. China, facing tighter constraints, has pushed harder on efficiency, open models, and targeted deployment. That can produce fast industrial wins in some sectors, but it does not automatically create the internal habits needed for company-wide adoption. (ai-frontiers.org) The result is a peculiar picture. In China, ordinary users can stampede toward a new AI app before most large employers know what to do with it. In the U.S., consumers may grumble about Big Tech while their employers quietly wire AI into sales, support, coding, and internal workflows. The gap is not between a country that understands AI and one that does not. It is between a market that can fall in love with a tool overnight and a corporation that still needs permission slips. (scmp.com)