China’s manufacturing sector is leapfrogging straight into the era of agentic artificial intelligence and is now exporting that playbook to global supply chains, according to the CEO of Black Lake, a Chinese provider of cloud-based AI tools for factories.

“Just as China skipped credit cards and jumped directly to mobile payments, many factories are skipping full traditional digitalisation and jumping directly to AI agents,” Zhou Yuxiang said in a recent interview with the South China Morning Post.

The shift marks a pivotal stage for China’s manufacturing sector. It was moving beyond its traditional advantages of low costs and reliable quality to compete on “agility”, deploying AI tools and agents from design to production so factories could meet small-batch and customised demand in time, Zhou said.

The trend aligns with Beijing’s national “AI plus” initiative, which includes a target of integrating AI into 90 per cent of China’s economy by 2030.

Shanghai-based Black Lake has launched a series of role-based agents that handle tasks ranging from design files and workflows to procurement, quality assurance, supply chain coordination, planning and scheduling, and equipment maintenance.

Up to 40 per cent of those agents had reached full autonomy, making decisions under human supervision, according to Zhou.

The remaining 60 per cent remained in a co-pilot stage where they helped employees make decisions faster, but Zhou said he expected them to evolve to full autonomy within 12 to 18 months through reinforcement learning based on engineering feedback.

“Factories buy the solution for speed and, once embedded in the factory, the agent learns from human feedback,” he said. “Over one to two months, the agent catches up, eradicates errors and reaches parity with human capability.”For factory owners, AI agents were increasingly valued as manufacturers grappled with an ageing workforce and a shrinking pool of young workers willing to learn from veteran technicians, according to Zhou.

Once software reaches full autonomy and robots operate without human supervision, that will represent the ultimate picture of future manufacturingZhou Yuxiang, Black Lake As China’s manufacturing sector embraces agentic AI, Black Lake is exporting the playbook overseas, where demand is growing rapidly in Southeast Asia and Latin America.

The firm’s overseas footprint initially stemmed from Chinese factories expanding operations in markets like Vietnam, Indonesia, Malaysia and Mexico.

According to Zhou, this expansion later broadened as multinational companies using Black Lake’s technology in China replicated the system across their international operations, followed eventually by local, native factories overseas seeking to integrate Black Lake’s AI agents into their own production.

“Agents are the result of industrial knowledge, skills and expertise [accumulated in] China’s industrial sector, which has been leading global manufacturing,” Zhou said.

Demand for greater manufacturing agility was being fuelled by a new wave of globally successful “Made in China” brands such as Pop Mart’s Labubu, Zhou said.

Unlike traditional mass-produced toys, Labubu products came in hundreds of distinct designs, requiring manufacturers to make rapid adjustments across their supply chains, he said.

To tap into the trend, Black Lake launched Popzao in April, a platform aimed at riding on the explosion of demand from celebrities, museums and tourist attractions seeking to create merchandise even in small batches despite having little or no manufacturing experience.

Popzao used AI agents to tokenise and automate processes ranging from design and 3D simulation to mould-making for clients, slashing fixed engineering costs from about 300,000 yuan (US$44,800) to 2,000 yuan in tokens, Zhou said. The model also lowered minimum order quantities from 20,000 units to just 100 units.

The growing robotic deployment in manufacturing has also prompted Black Lake to deepen integration between AI agents and hardware systems. “We act as the central dispatcher and brain,” Zhou said.

Robotics companies primarily targeted 70 per cent of factory work involving manual and low-skilled tasks, while AI agents addressed the remaining 30 per cent performed by high-skilled experts and engineers, Zhou said.