- Alibaba’s Accio AI sourcing tool surpassed 10 million monthly active users in March 2026, representing roughly one in five users of Alibaba.com.
- The tool is trained on 26 years of Alibaba’s proprietary transaction data and built on multiple frontier models, including the company’s open-source Qwen series.
- One documented case shows Accio’s supplier recommendation reducing a small seller’s per-unit manufacturing cost from $17 to approximately $2.50.
- Stanford HAI researcher Jiaxin Pei cautioned that AI sourcing agents must disclose the data they collect and the incentives built into their outputs.
What Happened
Alibaba’s AI sourcing platform Accio surpassed 10 million monthly active users in March 2026, the company told MIT Technology Review in a report published April 6, 2026. The tool, launched in 2024, lets small e-commerce merchants research products and identify manufacturers via a chat interface. That user figure represents approximately one in five Alibaba.com users consulting AI for sourcing decisions.
Why It Matters
Traditional supplier research for small sellers has required weeks of manual effort: browsing factory listings, comparing manufacturing capacities, negotiating minimum order quantities, and arranging sample shipments. Accio’s scale makes it one of the more concrete mass deployments of AI in business-to-business commerce infrastructure, distinct from consumer-facing retail assistants.
Alibaba Group CEO Eddie Wu told managers in March 2026 that integrating the company’s core services with its Qwen AI model family is a top priority. During a separate Chinese New Year promotion featuring Qwen’s personal shopping agent, customers placed 200 million orders, the company said.
Technical Details
Accio is built on multiple frontier language models, including Alibaba’s own Qwen series—an open-source family of large language models—according to Zhang Kuo, president of Alibaba.com. The system indexes millions of supplier profiles on the platform and is trained on 26 years of Alibaba’s proprietary transaction data.
The interface offers “fast” and “thinking” modes and returns structured outputs—charts, supplier links, and visual comparisons—rather than plain text responses. In one documented case, the tool identified a manufacturer in Ningbo, China, that reduced Illinois-based seller Mike McClary’s per-unit production cost for a redesigned flashlight from $17 to approximately $2.50. McClary, who runs his small outdoor brand from his living room, said he took the process from there by contacting the supplier directly; the redesigned product went live on Amazon within a month.
Who’s Affected
Small and mid-sized US e-commerce merchants sourcing manufactured goods from China and India are the primary user group. Richard Kostick, CEO of beauty brand 100% Pure, told MIT Technology Review that Accio “blows it away” compared to general-purpose tools like ChatGPT for sourcing-specific tasks. Vincenzo Toscano, an e-commerce seller and consultant, used Accio to develop a new sunglasses brand, with the tool suggesting materials and surfacing design references aligned with his brand concept before he reached out to suppliers.
Manufacturers on Alibaba.com are also adjusting their behavior in response. Sally Li, a representative at a makeup packaging company in Wuhan, China, said her firm has begun adding more detailed equipment and manufacturing experience information to its Alibaba listings because it suspects those details improve AI-driven discoverability—though she noted manufacturers cannot tell whether an incoming inquiry was guided by AI.
What’s Next
Alibaba has not yet determined how to monetize Accio at scale. Zhang Kuo confirmed the tool is not currently integrated with Alibaba.com’s paid placement advertising system. “We haven’t had a clear answer in terms of how to monetize this tool,” Zhang told MIT Technology Review. For now, users can pay for additional tokens after their free query allocation runs out.
Jiaxin Pei, a research scientist at the Stanford Institute for Human-Centered AI, said AI agents used in purchasing decisions “need to act transparently, securely, and in the customer’s best interest,” adding that developers should “disclose the data they collect and the incentives built into them to ensure that the marketplace remains fair.”
