Alibaba is sharpening the commercial focus of its artificial intelligence work by centring development on agents - AI-driven digital assistants that connect services across the company's extensive ecosystem. In recent months the group has introduced multiple agent integrations and announced an organisational change that separates its AI operations from its cloud business.
The reorganisation created the Alibaba Token Hub business group, which will be overseen by Chief Executive Eddie Wu. Company executives and outside analysts view the move as a signal that Alibaba is prioritising AI agents that rely on much higher token consumption - the units of data processed by language models to generate outputs - compared with conventional question-and-answer chatbots.
Alibaba did not provide comment on the restructuring request. The $325 billion e-commerce giant is due to report quarterly results on Thursday, and observers are watching AI monetisation closely as technology firms in China and abroad attempt to convert the new generative AI wave into consistent revenue. For the quarter, analysts forecast third-quarter revenue growth of 3.8% and a 42.5% decline in net income. The reporting period included Singles' Day, China’s largest shopping festival.
Alibaba’s push toward agent-driven commerce comes as broader consumer demand remains subdued. Faced with weak household spending, a challenging macroeconomic outlook and a property market slump that has reduced household wealth, the company has explored new models to stimulate consumption. One example is the heavy investment made last year to acquire users for its instant retail platform, a service that competes in the one-hour delivery market with Meituan.
Beyond logistics and delivery, Alibaba has started adapting its Qwen chatbot from a primarily informational tool into an interface that can directly help users complete purchases. In February, the company rolled out the first phase of a 3 billion yuan coupon programme designed to encourage users to make in-app purchases across Alibaba-owned retail properties using chatbot prompts alone. The initiative was so widely used that the app supporting it was temporarily shut down to manage the demand.
Former Alibaba employee and author Brian Wong noted the potential behavioural change if consumers can accomplish multiple daily tasks through a single conversational interface. Wong described the company's ecosystem as encompassing e-commerce, food delivery, travel, movie ticketing and more, and suggested that integrating those functions into a unified text-based assistant could fundamentally alter how people interact with digital services. He compared the scope of integration enabled by Alibaba's restructuring to combining a range of major digital services into a single natural-language interface.
Other Chinese technology firms are also experimenting with agent-based approaches, but analysts say Alibaba’s in-house ecosystem gives it unique advantages. Ed Sander, an analyst at China Digital Retail Report, highlighted Alibaba’s integrated fulfilment and logistics capabilities and its cloud infrastructure as differentiators. According to Sander, no other company in China is positioned to orchestrate a transaction from chatbot interaction through to final delivery in the same end-to-end manner that Alibaba can.
On the enterprise side, Alibaba introduced an automation-focused AI platform named Wukong. The platform is designed to coordinate multiple AI agents within a single environment in order to execute complex business processes such as document editing, spreadsheet updates, meeting transcription and research. The launch is positioned as part of Alibaba’s broader move to embed agent workflows across both consumer and enterprise use cases.
A central commercial rationale for prioritising agents is their substantially higher token usage. Poe Zhao, a China technology analyst and founder of Hello China Tech, estimated that agents performing continuous tasks could consume tens to hundreds more tokens per day than a typical chat session. That higher token throughput can create more monetisation opportunities, a key concern as many Chinese companies provide open-source AI models that are free to download and have faced steep declines in token pricing amid intense domestic competition.
Despite the strategic push, Alibaba is managing internal turbulence in its AI ranks. Lin Junyang, who headed the Qwen model division, departed in early March, marking the third senior executive from the Qwen team to leave so far this year. The exits have raised questions about morale and Alibaba’s capacity to retain top AI talent and maintain leadership in the AI model development race. Morningstar analyst Chelsey Tam warned that losing Lin and other core Qwen members to rivals would be a setback.
At the same time, some observers pointed to Alibaba’s broader engineering bench as a mitigating factor. Brian Wong said that AliCloud’s talent pool is deep and broad enough to cover gaps created by departures, particularly in light of the recent restructuring that consolidated AI functions.
The company’s strategic emphasis on agents, combined with new organisational structures and product launches like Wukong, represents a concerted effort to turn heavy token usage into a clearer revenue path. But Alibaba’s success in doing so will be judged against several headwinds: a weak consumer environment, the competitive dynamics of open-source model pricing and internal talent turnover.
($1 = 6.8851 Chinese yuan renminbi)
Key points
- Alibaba has separated its AI operations from its cloud unit and created the Alibaba Token Hub, led by Eddie Wu, to prioritise agent-driven AI that consumes higher volumes of tokens.
- New products and experiments include Qwen’s expanded shopping capabilities and Wukong, an enterprise automation platform that coordinates multiple AI agents for business tasks.
- The strategy aims to monetise higher token consumption amid falling token prices for open-source models and challenging consumer spending conditions, impacting e-commerce, cloud services and logistics.
Risks and uncertainties
- Leadership turnover in the Qwen division - including the departure of Lin Junyang and other senior executives - raises concerns about talent retention and could affect AI model development.
- Open-source model availability and plunging token prices reduce pricing power, making it more difficult to monetise AI unless higher token consumption by agents offsets lower unit prices.
- Weak consumer confidence and a prolonged property sector crisis that has eroded household wealth could constrain demand for Alibaba’s commerce and instant retail services, limiting revenue upside from new AI-driven features.