Dialogue with Wu Yunsheng, Vice President of Tencent Cloud: Why Enterprise-Exclusive Agents Remain Crucial

Author | Huang Yu

As the decisive factor in AI competition shifts from "model capabilities" to "engineering application capabilities," Agents have replaced large AI models as the protagonist of this World Artificial Intelligence Conference (WAIC).

At WAIC, Tencent comprehensively showcased its Agent matrix to the outside world and newly launched the overseas version of Tencent Cloud's Enterprise-grade Intelligent Agent Development Platform (ADP) 4.0.

An unavoidable question arises: Against the backdrop where personal agents like WorkBuddy are already widely adopted, why should enterprises still struggle to build their own exclusive Agents?

In a dialogue with media outlets including Wall Street News, Wu Yunsheng, Vice President of Tencent Cloud, not only explained the fundamental rationale for the existence of exclusive Agents but also highlighted several hurdles that enterprises still need to overcome when building intelligent agents.

Wu Yunsheng's answer points to one word: Control.

He compared two usage scenarios. When employees use WorkBuddy locally to create efficient sales processes or access various systems, enterprise managers may remain unaware, making it impossible for colleagues to reuse these workflows. However, if an employee creates a "Sales Assistant" on ADP, it can be published with a single click, allowing 30 employees to invoke it directly. Senior management can clearly monitor usage and permissions, and it can even be embedded into the enterprise's existing CRM or OA systems.

In other words, while using desktop-side intelligent agents is convenient for improving individual and team efficiency, applying them within enterprise business and operational scenarios imposes high requirements on agent construction, emphasizing integrability and controllability. In such cases, building a "cloud-integrated" solution based on ADP is more suitable.

Of course, enterprises also face challenges when building their own Agents.

Wu Yunsheng mentioned that a major difficulty currently faced by enterprises in deploying intelligent agents is how to interface with existing legacy systems. Much data and many processes are entrenched in these old systems, and some enterprises with less mature prior informatization efforts lack mature APIs or MCP interfaces for invocation.

"Legacy systems will also undergo changes driven by current trends, because they were previously designed for humans—their interfaces, interactions, and outputs were all intended for human consumption. In the future, the design philosophy of many systems must consider designing for Agents." Wu Yunsheng stated that the system design philosophy for Agents differs from that for humans.

As Agent Loop capabilities improve, tasks previously limited by retrieval and workflow capabilities are now ready for implementation.

A highly concerned question is: Will Agents erode existing SaaS?

Wu Yunsheng's answer is "No." In his view, Agents are more of a capability layer stacked on top of existing systems to connect and orchestrate, rather than tearing everything down to start over. The true incremental value lies in enabling those legacy systems, which have accumulated over decades, to develop interfaces that can be invoked by Agents.

Targeting the demand for enterprises to build Agents, Tencent fully upgraded its large model knowledge engine to the Tencent Cloud Intelligent Agent Development Platform (ADP) in May 2025.

Wu Yunsheng revealed that after more than a year of development, coupled with the emergence of new technologies this year, customer acceptance of Agents has significantly increased. The demand for Tencent Cloud ADP has also grown markedly, at least doubling year-over-year compared to last year.

It is reported that Tencent Cloud ADP has currently been implemented in over 30 industries, covering business scenarios such as intelligent customer service, knowledge management, and media production.

Focusing on "implementability, scalability, and operability," Tencent Cloud ADP continues to strengthen its capability foundation. According to introductions, the latest upgraded ADP 4.0 is positioned as an AgentOps platform, providing enterprises with a one-stop service for building, distributing, and governing cloud-based intelligent agents.

ADP 4.0 has achieved bidirectional invocation between Agents and Workflows, balancing open intent with rigorous processes. Furthermore, ADP exposes module capabilities such as spaces, applications, Skills, plugins, knowledge bases, documents, and sessions via OpenAPI, facilitating enterprises to embed intelligent agent capabilities into their original business processes.

Wu Yunsheng also stated that enterprise-grade intelligent agents are not about who can build them faster, but rather about who can run them more stably and securely in actual business operations, forming an operational system that continuously creates value. In this context, both capabilities and knowledge are indispensable.

Wu Yunsheng believes that the application of Tencent Lexiang's Agentic Knowledge Base is also crucial. "ADP solves the problems of how to build, run, and manage intelligent agents, while the knowledge base addresses what the intelligent agents understand, trust, and utilize."