Alibaba unveils its newest agentic AI platform in collaboration with the international unit, Accio Work
Alibaba has advanced in the global competition for agentic artificial intelligence, with its international commerce division introducing Accio Work, a plug-and-play “AI taskforce” that claims to autonomously manage complex business operations for small and medium-sized enterprises.
The launch occurs during a surge in China surrounding agentic AI, sparked by OpenClaw. This has led to a diverse group of consumers, from students to retirees, eager to participate in the “lobster raising” trend, resulting in companies hastily developing OpenClaw-based tools and escalating security concerns.
Accio Work stands in stark contrast to the consumer-driven frenzy, as the company states that it utilizes cross-functional AI teams that require no coding or setup.
“We set ourselves apart by being a specialized B2B tool instead of a generalist platform,” stated Alibaba International Vice President Kuo Zhang. “We establish a distinct boundary regarding high-stakes operations … any action that involves financial transactions, payment execution, or access to private files necessitates clear and detailed permission from the user.”
The launch occurs just under a week after another division of Alibaba unveiled Wukong, an AI platform designed for enterprises that can manage multiple AI agents to execute intricate business tasks, such as document editing, spreadsheet updates, meeting transcription, and research, all within one interface.
Last week, Alibaba announced its decision to separate its AI businesses from its cloud computing division. The recently established Alibaba Token Hub business group, under the leadership of Chief Executive Eddie Wu, represents a significant shift in the company’s direction towards digital assistants driven by AI models that utilize a greater number of tokens — units of data employed to generate language — compared to conventional Q&A chatbots.
Zhang emphasized that the critical global effort to define agentic AI comes with significant risks that can only be addressed through the use of controlled, specialized models that strike a balance between automation and security.
“We think the most significant risk comes from employing horizontal, generalist models for vertical business tasks. By concentrating on specialized B2B agents and integrating AI with human approval layers, we can provide the advantages of an autonomous workforce while mitigating the usual risks linked to unrestricted AI,” he stated.
