As the decisive factor in the artificial intelligence (AI) race shifts from “model capability” to “engineering application capability,” how can enterprises find the value in building proprietary Agents at a time when personal AI assistants are already widely adopted? Tencent Cloud Vice President Wu Yunsheng provided a key answer at the 2026 World Artificial Intelligence Conference (WAIC): governance.
During WAIC, Tencent showcased its full Agent matrix to the public and officially launched the overseas version of its Tencent Cloud enterprise-grade Agent development platform, ADP 4.0. Facing the reality that personal desktop AI assistants like WorkBuddy are already widely used by employees, Wu explained the fundamental reason for enterprise-specific Agents’ existence during a media interview, while also highlighting several hurdles enterprises still need to overcome when building AI agents.
Wu pointed out that when employees use WorkBuddy locally to design an effective sales workflow or call specific systems, enterprise managers often cannot see it, nor can they easily enable other colleagues to reuse it. However, if an employee builds a “sales assistant” on ADP, it can be published with one click, allowing 30 employees to directly invoke it. Senior management can also clearly grasp usage and permissions, and even embed the Agent into the enterprise’s existing Customer Relationship Management (CRM) or Office Automation (OA) systems. He stressed that while using edge-side desktop AI assistants for individual and team efficiency gains is convenient, deploying them in enterprise business and operational scenarios demands higher standards, with greater emphasis on integration and governance. Deploying through ADP’s “cloud-edge integrated” architecture is a more suitable choice.
Legacy System Integration is the Biggest Challenge; Agents Will Not Erode SaaS
The challenges enterprises face in building proprietary Agents cannot be overlooked. Wu stated bluntly that the biggest difficulty in deploying AI agents currently lies in integrating with existing legacy systems. Much data and many processes are buried within various old systems, and some enterprises with weaker prior IT infrastructure have legacy systems lacking mature Application Programming Interfaces (APIs) or Model Context Protocol (MCP) interfaces for invocation.
However, he believes the trend is driving change. Wu said: “Legacy systems were previously designed for humans; interfaces, interactions, and outputs were all meant for people to see. In the future, the design philosophy for many systems must consider designing for Agents.” He noted that the design philosophy for Agent-oriented systems is fundamentally different from that for human-oriented systems. As Agent Loop capabilities improve, some tasks previously constrained by retrieval and workflow capabilities now have the conditions for practical deployment.
Regarding the market’s concern about whether “Agents will erode existing SaaS,” Wu gave a clear answer: “No.” In his view, Agents are more like a layer of capability superimposed on existing systems, responsible for orchestration and scheduling, not a tear-down and rebuild. The real incremental value lies in enabling those legacy systems, which have been accumulating for decades, to develop interfaces that can be called by Agents.
ADP Demand Doubles; Version 4.0 Positioned as an AgentOps Platform
Targeting the strong demand for enterprise Agent building, Tencent comprehensively upgraded its large model knowledge engine to the Tencent Cloud Agent Development Platform (ADP) in May 2025. Wu 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, with Tencent Cloud ADP customer demand at least doubling year-over-year. It is reported that Tencent Cloud ADP has now been deployed across more than 30 industries, covering business scenarios such as intelligent customer service, knowledge management, and media production.
The newly upgraded ADP 4.0 is positioned as an AgentOps platform, providing enterprises with a one-stop service for building, distributing, and governing cloud-based AI agents. This version has achieved bidirectional interoperability between Agents and Workflows, balancing open intent with rigorous processes. It ensures process rigor while reducing token consumption, and the entire Workflow chain is traceable, meeting compliance and audit requirements. Additionally, ADP opens up module capabilities such as spaces, applications, Skills, plugins, knowledge bases, documents, and conversations via OpenAPI, making it convenient for enterprises to embed Agent capabilities into their original business processes.
Wu emphasized that enterprise-grade AI agents are not about who can build faster, but about who can run more stably and securely in the business field, forming an operational system that continuously creates value, where both capability and knowledge are indispensable. To this end, Tencent Cloud has built two foundational pillars: the capability foundation, “Tencent Cloud Agent Development Platform ADP,” and the knowledge foundation, “Tencent Lexiang Agentic Knowledge Base.” Wu explained: “ADP solves the problems of how to build, run, and manage AI agents, while the knowledge base solves the problems of what the agent understands, trusts, and uses.”
The conference also saw the simultaneous launch of ADP 4.0’s overseas version, along with upgrades to three core modules: the Intelligent Workbench, Claw Mode, and Skill Square, achieving comprehensive upgrades across four major capabilities: reach, interaction, ecosystem, and connectivity. In terms of omnichannel reach, the Claw application supports mainstream overseas channels such as LINE and Telegram. For interactive experience, the Intelligent Workbench supports custom time zones and multi-region scheduling, with model outputs automatically responding in the user’s input language. Regarding open ecosystems, the model marketplace has integrated multiple international mainstream large language models, and platform Skills have undergone international adaptation. For deep connectivity, the platform deeply integrates with mainstream SaaS applications such as Google Workspace, Confluence, and Jira, embedding deeply into overseas enterprise office and collaboration scenarios. Tencent Cloud also invited ecosystem partners at the conference to jointly build the “Ten Industries, One Hundred Scenarios Ecosystem Plan.”