On March 17, 2026, a light haze lent a somewhat languid air to Yunqi Town in Hangzhou. However, what appeared to be a routine technology launch day for DingTalk could potentially become a thunderclap in the process of enterprise digitalization. At 9:30 AM, as CEO Wu Zhao (Chen Hang) took the stage, DingTalk officially unveiled its enterprise-grade, AI-native work platform, "Wukong," to the world. This was not a standard product iteration but a quiet revolution aimed at fundamentally restructuring the underpinnings of enterprise productivity. While the industry was still debating the parameter scale of large language models, DingTalk had already shifted its focus to a more essential dimension: a paradigm shift from "humans operating software" to "intent directly leading to results," successfully making the daring leap from an "enterprise collaboration tool" to an "AI-native work system."
The core of this launch hinged on a decision that seemed radical yet targeted first principles: a comprehensive move towards CLI-ization (Command-Line Interface). In an era where Graphical User Interfaces (GUIs) have dominated computing for three decades, DingTalk chose to "regress," encapsulating thousands of features into a standardized set of machine instructions. The logic behind this move is stark and profound: GUIs are designed for human vision and muscle memory, filled with redundant clicks and uncertain interactions; the essence of AI, however, is code and logic. Only through deterministic command interfaces can intelligent agents take over enterprise processes with zero loss and high security. The birth of "Wukong" signifies that DingTalk is no longer merely an APP for people to use; it is evolving into an operating system for AI.
In this new operating system, the "Lobster" serves as the most vivid metaphor. The so-called "Lobster" entity is, in fact, an army of AI Agents with powerful execution capabilities. Like lobsters in the deep sea, they possess precise "claws" capable of delving into the fabric of enterprise operations to complete complex tasks ranging from approval workflows to supply chain procurement. This creates a stark contrast with general-purpose Agent frameworks like OpenClaw available on the market. General frameworks provide a universal robotic arm but lack understanding of specific contexts; DingTalk's "Wukong," however, is born from and grows within an organizational ecosystem. It inherently understands a company's structure, permission boundaries, and business context. It embeds a vast library of commercial skills from within the Alibaba ecosystem, such as those from Taobao, 1688, and Alipay, and has launched an enterprise-grade AI capability market that is compatible with open-source Skills, aiming to become the world's largest toB Skill market.
From an analytical perspective, "Wukong" is not intended to "defeat" the currently popular OpenClaw. Instead, on the enterprise track, it accomplishes what OpenClaw cannot—providing a secure, controllable, governable, and deeply integrated AI-native platform. If OpenClaw represents a "personal lobster," then "Wukong" represents an "enterprise-grade lobster army." This strategic move essentially constitutes a significant expansion of the enterprise boundary. In the past, DingTalk's scope was limited to collaborative office work, with its commercial value anchored in "user seats" and "storage space." Now, with the entry of "Wukong," DingTalk's boundary extends to the operation of a "digital workforce." Enterprises are no longer purchasing software licenses but a tireless, 24/7 standby intelligent army. The payment logic is set to evolve from "per-user fees" to "task execution volume" and "depth of skill invocation."
A more profound impact is that DingTalk is attempting to become the "infrastructure" for the enterprise AI era. By abstracting complex business logic into standardized APIs, DingTalk is defining the communication protocol for enterprise-grade Agents. In the future, any AI application seeking to enter a company's core business processes may need to run on this native protocol built by DingTalk. Viewed from a broader historical perspective, the launch of DingTalk's "Wukong" represents the first complete implementation of the Silicon Valley "Agent Native" concept within the soil of China's enterprise services sector. This also aligns with Wu Zhao's vision of "AI using DingTalk to work": a shift from "humans operating software" to "software collaborating autonomously," and from "connecting people" to "connecting AI with organizations." DingTalk's boundaries are being redefined.
This is just one facet of Alibaba's comprehensive commitment to the AI field. The day before, on March 16, Alibaba announced the official establishment of the Alibaba Token Hub (ATH) business group, incorporating the Qianwen and Wukong business units. The Wukong business unit is positioned to "create a B-side AI application entry point, deeply integrating model capabilities into enterprise workflows." The goal of "Wukong" from its inception has been global. During the launch event, DingTalk also released a global version of "Wukong," which will subsequently support integration with major global IM platforms like Slack, WhatsApp, and WeChat, breaking through DingTalk's traditional boundaries. Users will be able to remotely summon Wukong to complete tasks from either a computer or a mobile phone.
The name "Wukong" carries profound meaning. It symbolizes liberation from constraints and the limitless possibilities akin to its namesake's seventy-two transformations. DingTalk's "Wukong" aims to transform the "Lobster" from a personal toy into an enterprise-grade AI productivity army, thereby redefining DingTalk's boundaries and future.
Comments