Since the start of the new fiscal year on April 1, Alibaba has been mapping out new growth opportunities in e-commerce, with artificial intelligence expected to play a central role.
36Kr reported that Taotian Group’s current AI exploration will focus on B2B use cases. Key targets include the retention rate of AI-driven merchant tools and the gross merchandise value, or GMV, generated through AI. Heading into 2026, OpenClaw has accelerated interest in agentic AI, while group CEO Eddie Wu has led the creation of a new Alibaba Token Hub (ATH) business group. All related businesses are now required to commercialize around it.
Organizational changes have followed. According to 36Kr, Kaifu Zhang, who previously oversaw AI initiatives within Alibaba’s domestic e-commerce business group, no longer holds that role. The smart search and recommendation products division has been split into two units, platform users and products, and algorithms. Meanwhile, the future innovation division, which handled multimodal AI work, has been folded into the ATH business group.
“In the new year, ATH will lead the group’s AI strategy. There’s no need for other businesses to reinvent the wheel,” a source close to Taotian told 36Kr.
At the beginning of last year, the China e-commerce business group established a core search and recommendation products division under Zhang. His primary task was to integrate multiple teams, including those focused on search and recommendation algorithms, user algorithms, merchant algorithms, and content generation algorithms, to improve system precision and efficiency.
The focus was on B2C AI applications, with user-side penetration as the core metric. That required continuous optimization of the user experience, prompting the Taobao app to roll out several features, including universal search, a product selection assistant, a shopping guide, and scan-to-search tools. On the B2B side, the goal was to improve advertising performance and support merchant growth through AI.
Zhang’s most recent public appearance was at the kickoff event for the 2025 Singles’ Day promotion on Tmall. In a speech, he said the platform was using AI to better understand and organize roughly two billion products, generating double-digit gains in traffic-matching efficiency. He cited a 20% improvement in search relevance for complex semantic queries, a 10% increase in clicks in recommendation feeds, and a 12% improvement in merchant advertising return on investment.
In the second quarter of last year, Jiang also said on an earnings call that Taotian was working to improve user experience and rebuild its search advertising system using AI. That effort was a priority and had already produced measurable results. According to 36Kr, the search and recommendation division adjusted its logic using AI to increase the conversion rate from searches to transactions, contributing to higher order volumes.
Still, B2B appears to offer a more direct path to AI monetization than B2C. This helps explain why Anthropic’s annualized revenue has reportedly surged to USD 30 billion, allowing it to overtake OpenAI. With a large merchant base, Alibaba’s e-commerce business is now moving toward a token-based model.
A person close to Taotian told 36Kr that the group is recruiting what it calls “fourth-generation service providers,” referring to agentic AI service providers. Earlier generations included labor-based store operators, custom software providers, and software-as-a-service providers. As part of this shift, the Qianniu platform, originally built for merchants, will be upgraded to “Qianniu Claw.” Taotian plans to promote the product around the 618 shopping festival, with support from these new service providers.
According to an Alibaba source, merchant demand currently centers on three areas: greater traffic support, higher operating efficiency, and lower product selection costs. Qianniu Claw is expected to address these needs by reducing operating costs and improving efficiency.
Under Taotian’s plan, Qianniu Claw will offer customized services for large brands and standardized solutions for small and midsize businesses. It will also deploy merchants’ systems, applications, and data on Alibaba Cloud. In this model, merchants will consume tokens, with costs borne either independently or shared with service providers.
This shift could particularly benefit small and midsize sellers. Merchant feedback indicates that opening a Tmall store remains complex, requiring both sufficient staffing and external operators. With Qianniu Claw, many of these functions could be handled by AI agents. Large brands, whose e-commerce teams can number in the hundreds, may also use the platform to reduce headcount and convert certain roles into automated agents.
In theory, this creates mutual benefits. Merchants could lower operating costs, while Alibaba establishes a practical application for its token economy within e-commerce. Growth metrics may also shift beyond GMV alone, reflecting broader revenue pressures. Alibaba’s fiscal 2025 fourth-quarter earnings show Taotian Group revenue grew 6%, while customer management revenue, or CMR, increased just 1% year-on-year.
According to the same source, Qianniu Claw could drive revenue in two ways. Lower operating costs and improved margins may encourage merchants to increase advertising spending. At the same time, pricing could shift from traditional advertising fees to token-based consumption, supporting overall revenue growth.
RELATED ARTICLE
On a recent earnings call, Eddie Wu provided AI revenue guidance for the first time, stating that AI and cloud revenue could reach USD 100 billion annually within five years. A source close to Alibaba said that under the token model, this figure could also include AI revenue generated by its e-commerce operations.
On the merchant side, Taotian is upgrading its operational index and exploring fully AI-driven governance. AI systems may take on supervisory roles, such as determining penalties or product removals, reducing reliance on manual processes.
A Taotian employee told 36Kr that products and technologies are also being reconfigured around AI agents. Previously, product teams focused on direct business needs. Now, they are building modular “skills” for agents, with tools such as automated refund processing and penalty assessment receiving increased resources.
AI adoption is also expanding internally. Employees across Alibaba’s e-commerce business are now evaluated on AI-related metrics, including tool usage and issue resolution rates. In international operations, staff must complete AI training, track participation, and pass exams to earn points. While these points are not yet tied to performance reviews, they come with incentives and signal the institutional importance of AI.
Management expects that increased use of AI tools will improve efficiency. One employee said their team uses an AI-powered office tool called Qoderwork, similar to Wukong, which functions as a task assistant. “Two years ago, our only AI use case was customer service,” the employee said. “At that time, responses were basic. Now AI can review products, and merchants can apply through AI, which has reduced a lot of repetitive work.”
36Kr also reported that since March 25, employees have been able to queue for tokens to use in their workflows. Business teams have actively applied to use the Wukong system.
In terms of model selection, Taotian’s e-commerce business is now closely aligned with the Qwen large model, partly due to data privacy and security considerations. Last year, employees could still choose between Qwen, DeepSeek, and Kimi models.
After two years of internal evaluation, Alibaba is now advancing its AI strategy with a focus on structural change. Its push into agent-based systems reflects an effort to secure a position in what it sees as the next phase of AI-driven commerce.
KrASIA features translated and adapted content that was originally published by 36Kr. This article was written by Peng Qian for 36Kr.