"Monkey King" sweeps away "Shrimp Soldiers" Alibaba redefines DingTalk: Wu Yongming kicks off the main battlefield of B2B AI

Every reporter｜Chen Ting Every editor｜Yang Jun

(Provided by the company)

OpenClaw (an open-source AI agent framework, commonly known as “lobster”) has taken the market by storm, making “raising lobsters” a new trend in the industry, but Alibaba has voiced a different opinion.

“I see articles about lobsters running rampant online every day, and while knowledge bloggers are constantly praising lobsters, I feel it’s irresponsible,” said Chen Hang (alias Wu Zhao), CEO of DingTalk, at the AI DingTalk 2.0 annual product launch on March 17. He stated that uncontrolled super-intelligent agents pose a risk of backlash.

Shortly thereafter, Chen Hang unveiled Alibaba’s breakthrough solution by launching the world’s first enterprise-grade AI-native work platform, “Wukong.”

Interestingly, on the big screen at the launch event, a cartoon version of “Wukong” stood with a stick among a group of shrimp soldiers, the image full of metaphor. Alibaba claimed that Wukong is a “lobster army” that works 24 hours a day, which will be directly integrated into DingTalk, used by over 20 million enterprises.

“Today, we have shattered DingTalk and rebuilt it with AI, refining it into ‘Wukong,’” Chen Hang stated. In the past, humans used DingTalk to work; in the future, AI will use DingTalk to work. “Unlike all other lobster agents on the market, Wukong is inherently rooted in enterprise organizations and can be safely used in real business environments.”

On March 16, Alibaba officially announced the establishment of the Alibaba Token Hub (ATH) business group, which is directly overseen by Alibaba CEO Wu Yongming. Among them, the Wukong division has made its first appearance in the public eye, clearly signaling Alibaba’s focus on the B-end AI application market.

This is not only a restructuring of DingTalk but also a crucial pivot in Alibaba’s AI strategy. Behind this well-prepared launch event is a reflection of Alibaba’s latest layout for the commercialization of AI.

(Provided by the company)

Safety and control are the ultimate battleground; Alibaba raises the “Golden Staff”

At the beginning of the launch event, Chen Hang elaborated on the security issues of OpenClaw. He stated that if all the “lobsters” were released, they would definitely harm the entire ecosystem. “In fact, once installed on personal or company computers, they would be full of backdoors, breeding various Trojan viruses.”

Concerns about the security risks of OpenClaw have long existed in the industry.

On March 8, the Ministry of Industry and Information Technology’s cybersecurity threat and vulnerability information sharing platform published an article stating that it had detected high security risks in some instances of OpenClaw’s open-source AI agents under default or improper configurations, which could easily lead to security issues such as cyberattacks and information leaks.

According to official statements, the fundamental difference between Wukong and other AI agents lies in the fact that others focus on making AI capable of working, while Wukong emphasizes making AI work safely, controllably, and transparently in enterprises.

Reporters learned that Wukong’s dual-layer rule system defines the absolute bottom line for AI behavior, with foundational security rules being the highest priority, and no command can breach them; enterprise-customized rules allow administrators to flexibly configure based on industry characteristics.

Based on DingTalk’s 11-year enterprise-level permission system, administrators can precisely control who can use Wukong, what AI skills can be used, and what data can be accessed. All operations are tied to the enterprise’s real identity, and without sufficient permissions, data cannot be accessed.

In group chat scenarios, Wukong’s permissions are derived from the intersection of “user permissions” and “questioner permissions.” Even if Wukong theoretically can access certain data, if the questioner lacks permission, Wukong will not return it.

Furthermore, its full-chain audit logs will record every input, every skill invocation, and every output—who did what, when, with what identity, and on what data, all clear at a glance, ensuring enterprise data security and controllability, and AI behavior can be traced and managed.

Zhang Yi, CEO and chief analyst at iiMedia Research, stated in an interview with the Daily Economic News that OpenClaw indeed has security issues such as unauthorized execution and data leakage, following a development path that prioritizes functionality over security. “Objectively speaking, it lacks the necessary permission system, data isolation, and operational auditing that enterprises require, making it more suitable for individual geek exploration.”

While emphasizing safety and control has its value, can this become the core competitiveness that distinguishes Wukong from its competitors? Zhang Yi believes that achieving safety and control is not easy. “It’s not simply adding safety features; it requires security design from the underlying architecture, involving organizational structure, deeply integrated permission systems, data isolation mechanisms where data does not leave the domain, and controllable operations with traceability, as well as multi-AI collaborative safety governance across multiple dimensions, all of which are crucial,” Zhang Yi said.

(Provided by the company)

DingTalk nurtures “Wukong”; can AI truly take over workflows?

Why did Wukong become the protagonist at a DingTalk launch event?

Reporters learned that as Alibaba Group’s enterprise-grade AI-native work platform launched globally, Wukong was developed by the DingTalk team.

On March 17, after updating the DingTalk App, opening the AI entry in the lower left corner revealed a completely new and greatly changed DingTalk interface to reporters.

Regarding the relationship between DingTalk and Wukong, DingTalk’s new feature “AI Inquiry” stated in response to reporters’ questions that DingTalk and Wukong are deeply integrated and evolve symbiotically; Wukong is not an independent competitor to DingTalk but is Alibaba’s enterprise-grade AI-native work platform, reconstructed, upgraded, and inherently built into DingTalk.

In simple terms, DingTalk serves as the soil and skeleton, while Wukong is the intelligent nerve and execution engine that grows within it.

As Chen Hang stated, for Wukong, DingTalk rewrote its underlying code, fully CLI-enabled (Command-Line Interface), ensuring that all capabilities can be called and operated by AI.

“AI Inquiry” noted that this transformation has achieved true “communication equals execution.” For example, in a DingTalk group, a command such as “generate last week’s sales report and synchronize it with management” allows Wukong to automatically pull approval flows, attendance, and CRM data to generate reports and push them, all without manual operation of the interface.

Additionally, Wukong itself is an independent application, and it will be directly embedded into DingTalk’s latest AI 2.0 version, allowing all DingTalk users (covering over 20 million enterprises and 800 million users) to use it out of the box without needing additional installation.

Moreover, reporters learned that Alibaba’s B-end (merchant) capabilities such as Taobao, Tmall, Alipay, and Alibaba Cloud will gradually connect to Wukong, making Wukong the unified outlet for Alibaba’s AI capabilities in enterprise work scenarios.

It can be seen that by embedding Wukong into the user base of DingTalk, Alibaba aims to rapidly reach a vast number of enterprise users with its enterprise-grade AI capabilities.

Furthermore, on the day of the launch event, Wukong simultaneously released ten industry solutions for OPT (One Person Team), which is the world’s first solution to turn AI Skills from a technical concept into industry-level, out-of-the-box products. The first batch covers ten scenarios, including e-commerce, cross-border e-commerce, knowledge bloggers, development, and stores, allowing users to enable them with a single click.

According to reports, by the end of 2025 and the beginning of 2026, the global AI industry is shifting from the “Agent arms race” to “Skills ecosystem building,” with industry consensus becoming increasingly clear. The core competitiveness of AI no longer solely depends on the scale of model parameters but on Skills, which are standardized, reusable, and evolvable capability modules that can truly be embedded in business processes and realized on a large scale.

Cui Lili, deputy director and professor at the Digital Economy Research Institute of Shanghai University of Finance and Economics, told the Daily Economic News that Wukong is more like a platform that integrates B-end Skills, still exploring more micro team work scenarios and is currently in the exploration stage.

It is worth mentioning that in the updated DingTalk App, many requests made to Wukong after tagging it were refused on grounds of permission isolation and security sandbox limitations. For instance, based on feedback, Wukong will not automatically scan or archive users’ historical information, which also means that users will need time to learn how to use this application.

(Provided by the company)

After the “lobster” frenzy, B-end AI will determine the outcome

Various signs also indicate that at this stage, Alibaba’s AI strategy may have anchored the direction for the next stage of development.

On March 16, Alibaba officially established the Alibaba Token Hub (ATH) business group, creating a new organization focused on “creating Tokens, delivering Tokens, and applying Tokens,” directly overseen by Alibaba CEO Wu Yongming.

Alibaba Token Hub includes Tongyi Lab, MaaS (Model as a Service), Qianwen Division, Wukong Division, and AI Innovation Division, covering a complete layout from foundational model research and development, model service platforms, to AI applications for individuals and enterprises. Among them, the Wukong division has made its first appearance in the public eye. From the action of establishing an independent business division, it is evident that the B-end AI application market has become one of the core elements of Alibaba’s AI strategy.

In the long term, where will this emerging market head?

Zhang Yi believes that in the future, enterprise-level B-end AI applications will inevitably become the focus of core layouts for major technology companies.

On one hand, the commercial closed loop of the B-end market will become clearer, as enterprises are willing to pay for cost reduction, efficiency enhancement, and security compliance, with a more stable monetization path compared to the C-end; on the other hand, the B-end scenarios have clear boundaries, and the business processes are more standardized, which is conducive to controllable implementation of AI. Additionally, enterprises have mandatory requirements for data sovereignty, permission control, and compliance auditing, which can create higher industry barriers. Therefore, B-end AI will undoubtedly become the main battleground for giants.

Zhang Yi also stated that the recent national discussions and experiences triggered by OpenClaw have made the market clearly aware of the real demands of enterprises and users for AI.

He believes that the current core competitive points in the AI industry mainly focus on the following aspects: first, the degree of integration between the model base and the MaaS platform; second, the service capabilities of C-end AI entry and personal assistants; third, the smoothness of the integration between AI agents and workflows in B-end scenarios.

He emphasized that the key to winning or losing in future industry competition does not solely depend on the strength of the models themselves, but more importantly on user acceptance and usage rates, which cannot be separated from the guarantees of safety and compliance, as well as the ability of AI to solve practical problems in specific scenarios. Furthermore, the ability to monetize commercially and integrate ecosystems is also an important factor determining a company’s competitiveness.

At the current stage, Alibaba’s approach in the AI battlefield has become increasingly clear, penetrating users and iterating models through the C-end “Qianwen” App, exploring commercialization and building competitive barriers through the B-end “Wukong,” while relying on the ATH business group to connect the AI capabilities of both C-end and B-end.

From “lobsters everywhere” to “Wukong’s emergence,” the battle for B-end AI applications is ushering in a critical turning point. In the consumer-grade market, AI can be an all-powerful “magician”; but in enterprise-level scenarios, it must be a safe, reliable, and clearly defined “professional manager.”

Perhaps only when AI truly evolves from being “able to work” to “able to do well within enterprises” can the commercialization loop be genuinely opened, and the industry competition will have a concrete starting point.