AI agents move into the workplace with Alibaba’s Wukong

Alibaba's Wukong points to a wider shift toward AI agents in the workplace, where systems can plan and carry out tasks with less human input.

A new push toward AI agents in the workplace is taking shape as companies look for ways to automate routine tasks and manage more complex workflows. Alibaba's latest platform, called Wukong, reflects that shift by focusing on how multiple AI agents can work together inside everyday business tools.

Wukong is designed as an enterprise system in which different agents handle tasks in a shared environment. Instead of relying on a single assistant, the platform can assign work across several agents, each handling a part of the process. These tasks include document editing and spreadsheet updating, as well as approval processing and meeting transcription.

The system also supports more involved work, such as research tasks that require pulling and organizing information from different sources. It can run across local machines, online browsers, and cloud systems, allowing it to interact with tools that employees are already using.

This release comes after an internal restructuring at Alibaba, where the company formed a new unit called the Alibaba Token Hub. Wukong sits within this group and acts as its main enterprise AI product. The move suggests a broader focus on building tools that center on AI agents rather than standalone applications.

Expanding access through workplace tools

Access to Wukong is limited for now. It is available through an invitation-only beta and can run as a desktop app or within DingTalk, Alibaba's workplace messaging platform. DingTalk already supports tens of millions of users, which gives Wukong a built-in user base. Over time, the company plans to connect the platform with other tools such as Slack, Microsoft Teams, and WeChat, which would extend its reach across different work environments.

Security is a core part of the system. Wukong includes identity checks, access controls, and isolated environments for running tasks. These features aim to reduce risks when agents handle company data or perform actions across systems.

Alibaba also plans to link Wukong with its wider ecosystem. Services such as Taobao, Tmall, 1688, Alipay, and Alibaba Cloud are expected to appear as modular "skills" that agents can call on when needed. Third-party integrations are also part of the plan, which could allow companies to tailor the system to their own tools and workflows.

Industry-focused tools and system design

To make the platform easier to use in specific industries, Alibaba has introduced a set of packaged solutions called One-Person Team, or OPT. These are designed for sectors such as retail, manufacturing, finance, legal services, and software development. Each package groups together tools and agent functions suited to that field, which may help small teams or startups handle work that would normally require more staff.

Under the hood, Wukong relies on a redesigned version of DingTalk's interface. The system now includes a command-line style setup and an open API layer, giving the platform deeper control over tasks. This structure allows agents to plan steps, generate instructions, and carry out actions with less direct input from users.

Along with its enterprise efforts, Alibaba has redesigned its consumer-facing AI software, Qwen. While Wukong focuses on workplace use, the Qwen update shows that the company is also exploring how similar AI technologies may be integrated into everyday tasks outside of the office.

Taken together, these moves point to a wider effort to place AI agents at the center of both business and consumer tools, with an emphasis on handling real tasks rather than just providing responses.