China's Factories Skip the Software Era and Sell AI Agents to the World's Supply Chains

China's factory floors never finished the software revolution. They are skipping it entirely — and selling the shortcut abroad.

That is the argument Zhou Yuxiang, chief executive of Black Lake, makes about the country's manufacturing base. Black Lake builds cloud-based artificial intelligence tools for factories, and Zhou told the South China Morning Post that Chinese plants are jumping straight into the era of agentic AI — software that does not just report on production but acts on it — rather than grinding through the multi-year digitalisation projects that Western manufacturers typically treat as a prerequisite. His analogy is deliberately local: China never built a credit-card culture, it went from cash to QR codes almost overnight. Factory software, he suggests, is following the same curve.

The claim matters beyond one company's sales pitch. China runs the densest manufacturing ecosystem on earth, and its small and mid-sized plants have long been criticised for running on spreadsheets, paper and the institutional memory of a veteran foreman. The conventional path out of that — installing ERP systems, connecting machines, hiring integrators — is slow, expensive and frequently abandoned. Zhou's contention is that AI agents collapse several of those steps. Instead of a plant first learning to digitise and then learning to automate, the agent layer arrives first and works around the mess. That is a genuinely different adoption sequence from the one Western vendors have been selling for two decades, and it is why the phrase 'leapfrog' keeps surfacing in Chinese industrial-tech circles.

Black Lake sits in a crowded but fast-maturing corner of China's tech economy: industrial AI aimed at the country's vast base of factories rather than at its internet consumers. The sector has attracted both established cloud players and a wave of startups promising to make shop-floor data useful without a full systems overhaul. These firms are riding two forces at once. Beijing's industrial policy has pushed smart manufacturing up the agenda, while the domestic market's sheer scale — thousands of plants making everything from auto parts to appliances — gives AI vendors a testing ground that no other country can match. Success at home is the credential; exports are the ambition.

What Zhou is describing is really a capital story. Chinese industrial-tech companies have spent years proving their tools inside domestic supply chains, and they now want to sell those tools to factories in Southeast Asia, South Asia, Latin America and beyond — the same manufacturing belt that global brands have been quietly diversifying into. If an AI agent can be dropped into a Vietnamese or Indonesian plant without a lengthy digitalisation programme, the addressable market is enormous. It also hands Chinese vendors a narrative advantage: they are not selling a mature Western product at a premium, they are selling a workaround built for factories that never had the luxury of a clean technology stack.

Sceptics will note that 'agentic AI' is a phrase doing heavy lifting right now, and that factories are unforgiving places for software that hallucinates. A chatbot that errs is embarrassing; an agent that mis-schedules a production line is expensive. Zhou's answer, in effect, is that the alternative — waiting for full digitalisation — is not happening at most plants. The leapfrog is less a strategic choice than a pragmatic one.

For anyone tracking Asian wealth and capital, the signal is worth watching. China's next export may not be a physical good but a management layer for the world's factories, sold by companies most outsiders have never heard of. The plants that adopt it first will not be the showpieces of German or Japanese industry. They will be the ones that never had a legacy system to protect — which, increasingly, describes the manufacturing map of the developing world.