When Wang Ming unlocked the door to his new office in mid-January, it marked a sharp break from the career path he had followed for more than a decade.

Just two months earlier, Wang had been the youngest vice president at Alibaba-backed DingTalk, one of China’s largest enterprise software platforms. Now he is the founder of K2 Lab, an artificial intelligence startup betting that the next phase of the AI boom will be driven less by tools.

“In this era, many of the real opportunities belong to startups,” Wang said in an interview. “Large companies are good at building platforms, but innovation around business models and new scenarios usually happens at the edge.”

Wang formally launched his company, Panfeng Intelligence, known in English as K2 Lab, in October 2025. The name, he said, reflects both “picking low-hanging fruit” and “climbing to the peak of intelligence.”

The company’s first product, Moras, is scheduled to launch in the coming months. It is an AI agent designed for overseas TikTok creators that automates the full process of content-driven e-commerce — from selecting products and generating scripts to producing short videos, publishing them, and tracking sales. Instead of charging a fixed subscription, K2 Lab plans to take a share of creators’ revenue.

The approach reflects Wang’s view that the AI industry is entering a turning point.

“Tools can only charge subscriptions,” he said. “Results allow you to share in the profits. Only scenarios that are close to revenue and form a closed loop can support long-term value.”

During five years at DingTalk, Wang oversaw the platform’s expansion into China’s largest enterprise service ecosystem and later led its transition toward AI-powered workflows. He helped open the platform to large language models and build an ecosystem of AI applications.

In early 2025, he publicly described DingTalk’s ambition to become “China’s largest AI startup incubator.” But after speaking with hundreds of founders, Wang decided he would rather build than mentor.

“The hardest part was deciding to leave,” he said. “The economic environment isn’t easy, and large tech companies still offer security that startups can’t match.”

That calculation — leaving stability for uncertainty — is increasingly common among senior executives in China’s technology sector, where slowing growth and intense competition are pushing experienced operators to look for new frontiers.

Rather than targeting China’s domestic market, K2 Lab is focused on creators in Europe and the United States.

Wang said overseas markets offered stronger willingness to pay and clearer monetisation paths. He also argued that content-driven e-commerce outside China remains underdeveloped, leaving room for automation.

“Many overseas creators don’t know what products to sell, how to create commercial content, or how to operate at scale,” he said. “These are exactly the gaps AI can fill.”

Moras uses a system of multiple AI agents working in sequence. One analyses the creator’s account and market data to recommend products. Another generates scripts and videos, selecting the appropriate AI tools for editing or avatars. A third handles publishing, tagging and timing. A fourth analyses performance and feeds results back into the system.

Humans remain in the loop to review outputs that AI handles poorly, such as identifying physical inconsistencies or potential violations of platform rules.

The goal, Wang said, is to reduce the creator’s involvement to a minimum while producing predictable commercial returns.

K2 Lab’s business model is based on revenue sharing rather than fixed fees.

Creators pay a small base fee and then share a portion of sales generated by the AI. In another version of the model, the AI assigns simple tasks to users — such as moderation or labelling — and pays them.

“We don’t sell software licences,” Wang said. “We act more like a partner.”

The model also helps control computing costs, he said. Videos are kept under 30 seconds, where production costs are lower and conversion rates are similar. High-cost tasks are handled by humans where possible. Advanced models are used selectively rather than universally.

The company says it has validated that commissions from product sales can, on average, cover computing expenses, making the system economically viable.

Wang predicts 2026 will mark a shift away from pure AI tools toward applications directly tied to revenue.

“Most of the tools that need to exist already exist,” he said. “Now the question is how to turn capability into value.”

That view echoes a broader debate in the global tech industry over whether generative AI can justify its heavy infrastructure costs through enterprise subscriptions alone, or whether it must become embedded in commercial workflows.

K2 Lab’s long-term ambition extends beyond e-commerce.

Wang envisions what he calls “the Shopify of the AI era” — a platform that provides individuals with agent infrastructure covering traffic, supply chain, payments and operations, allowing them to run digital businesses supported by AI.

“In the future, everyone could have a digital avatar and a business system,” he said. “We want to be the infrastructure behind that.”

Despite the vision, Wang is acutely aware of the risks.

The AI sector is crowded, capital-intensive and moving quickly. Large technology firms continue to pour resources into foundation models and integrated ecosystems.

“The window is short,” he said. “Startups have to grow fast and stay focused to survive between the redwoods.”

For now, K2 Lab remains in product validation, refining Moras for different markets and regulatory environments. It has not disclosed funding details.

The company’s team includes former DingTalk executives and engineers with experience in AI infrastructure, content creation and e-commerce, as well as staff with overseas backgrounds.

Wang believes that combination — technical depth plus commercial pragmatism — is essential.

“In the end, AI is not about replacing people,” he said. “It’s about amplifying what people can do.”

As the AI industry moves into its next phase, Wang is betting that the future belongs not to those who build the smartest tools, but to those who turn intelligence into income.