腾讯汤道生 WorkBuddy将开放架构 支持企业自建Agent网络Tencent's Tang Daosheng: WorkBuddy will open architecture to support enterprises in building their own Agent networks


AI free token/AI免费积分

中证报中证网讯(记者 张兴旺)10月29日,在2026腾讯全球数字生态大会上,腾讯集团高级执行副总裁、云与智慧产业事业群CEO汤道生表示,腾讯将围绕办公协同、数据管理与业务执行三大情形,持续开放WorkBuddy(腾讯出品的AI办公工作台)Agent(智能体)平台能力,推动企业构建可运营、可进化的智能生产力平台。

会上,汤道生宣布,以WorkBuddy为工作台的一体化套件Buddy One正式发布;混元全新模型Hy4以及Hy4 flash即将推出。同时,他也透露,自今年3月上线以来,WorkBuddy企业版已落地50多个行业,覆盖中国企业500强中的50%、企业100强中的60%,开放平台伙伴和开发者超过5万家,上架Skills(技能包)超20万。

汤道生认为,当前企业落地Agent最大的难题,不是模型不够强,而是断点太多。Agent落地的关键是连接——把人、工具、平台和数据连接起来,让大模型理解业务上下文,真正进入工作流程。

要接入企业上下文,开放是关键。汤道生表示,WorkBuddy将全面开放机型架构,同时兼容企业自研Agent与生态中第三方Agent,涵盖第三方IM(即时通信)平台的接入。“选什么模型、接什么Agent、用什么IM,挑选权都在企业手里。”汤道生表示,已经有越来越多企业基于WorkBuddy组成自己的企业Agent网络,搭建新一代的企业智能中枢。

在办公情形,汤道生提出,多人协作将进入“人机共创”的新范式:团队讨论完把需要提交给AI,内容由大模型生成,由Agent来整合,团队来验收,然后沉淀到知识库。将来的协作不再围绕文件反复流转,而是围绕任务持续推进。会上发布的Buddy One,正是面向这一趋势推出的一体化套件。它以WorkBuddy为工作台,整合腾讯文档、腾讯ima(AI智能工作台)、腾讯外观等工具,借助统一账号、统一积分池、统一数据池,让大模型依据任务需要调用不同工具,生成文档、外观、图片等多模态内容,再由团队共同修改、验收和沉淀。

在数据管理情形,腾讯介绍称,WorkBuddy与企业微信深度融合,使用者在企业微信里的沟通、文档和会议记录,都能在授权范围内,成为Agent理解任务、开展工作的依据。企业不用换平台、搬数据、重做流程,直接可以把沉淀多年的数字化资产变成AI的生产资料。员工也可以像和同事对话一样交办任务,Agent全天候执行,并把结果发回企业微信。“谁能更好使用AI作出更快决策、更早行动,就可能在市面更早发现商机。”汤道生说,“在AI时代,速度本身就是竞争力。”

在业务执行情形,腾讯企点营销云MAGIC智能体已上线WorkBuddy开放平台,一个入口即可串起营销的准备、执行、成交和复盘。企业也可以借助腾讯云ADP(智能体开发平台)定制业务Agent,并开展统一纳管和调度,让智能体真正进入客服、营销、运营等核心业务流程。

Agent落地全情形,也离不开基础设施与安全能力支撑。汤道生表示,腾讯云具备高表现与极高性价比的Agent基础设施,稳定的异构算力集群,与多家芯片厂商深度合作,构建国产算力的新生态。模型维度,腾讯混元Hy大模型持续在真实业务情形中进化,同时兼容客户基于Tokenhub(腾讯云大模型服务平台),借助API(应用程序接口)调用混元、DeepSeek、GLM、MiniMax、Kimi等主流开源大模型。

腾讯汤道生 WorkBuddy将开放架构 支持企业自建Agent网络

本文来源:https://www.cs.com.cn/ssgs/01/2026/10/29/detail_2026102910048290.html
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

China Securities Journal / CSN (Reporter Zhang Xingwang) On October 29, at the 2026 Tencent Global Digital Ecosystem Conference, Tang Daosheng, Senior Executive Vice President of Tencent and CEO of the Cloud and Smart Industries Group, said Tencent will continue to open up capabilities of the WorkBuddy (an AI office workbench produced by Tencent) Agent (intelligent agent) platform around three scenarios—office collaboration, data management, and business execution—to help enterprises build operable, evolving intelligent productivity platforms.

At the conference, Tang Daosheng announced that the integrated suite Buddy One, with WorkBuddy as the workbench, has been officially released; the new Hunyuan model Hy4 and Hy4 flash will be launched soon. He also disclosed that since its launch in March this year, WorkBuddy Enterprise Edition has been deployed across more than 50 industries, covering 50% of China’s Top 500 enterprises and 60% of the Top 100 enterprises, with over 50,000 platform partners and developers and more than 200,000 Skills (skill packs) listed.

Tang Daosheng believes the biggest challenge enterprises face in deploying agents today is not that models are not strong enough, but that there are too many breakpoints. The key to agent deployment is connectivity—connecting people, tools, platforms, and data so large models can understand business context and truly enter workflows.

To connect to enterprise context, openness is key. Tang Daosheng said WorkBuddy will fully open its model architecture while also being compatible with enterprises’ self-developed agents and third-party agents in the ecosystem, including access to third-party IM (instant messaging) platforms. “Which model to choose, which agent to connect to, and which IM to use—the choice is in enterprises’ hands.” Tang noted that more and more enterprises are already building their own enterprise agent networks based on WorkBuddy, forming a new-generation enterprise intelligent hub.

In the office scenario, Tang proposed that multi-person collaboration will enter a new paradigm of “human-machine co-creation”: after team discussion, what needs to be submitted to AI is determined; content is generated by large models, integrated by agents, accepted by the team, and then accumulated into a knowledge base. Future collaboration will no longer revolve around files being repeatedly circulated, but around tasks being continuously advanced. Buddy One released at the conference is an integrated suite launched for this trend. With WorkBuddy as the workbench, it integrates tools such as Tencent Docs, Tencent ima (an AI intelligent workbench), and Tencent Wenchuang, and via unified accounts, a unified points pool, and a unified data pool, it enables large models to call different tools as needed to generate multimodal content such as documents, Wenchuang files, and images, which the team then collaboratively revises, accepts, and accumulates.

In the data management scenario, Tencent said WorkBuddy is deeply integrated with WeCom (Enterprise WeChat), and users’ communications, documents, and meeting notes in WeCom can, within authorized scope, become the basis for agents to understand tasks and carry out work. Enterprises do not need to switch platforms, move data, or rebuild processes; they can directly turn digital assets accumulated over many years into AI production resources. Employees can also assign tasks as if talking to colleagues; agents execute around the clock and send results back to WeCom. “Whoever can use AI better to make faster decisions and act earlier may discover business opportunities in the market sooner,” Tang said. “In the AI era, speed itself is competitiveness.”

In the business execution scenario, Tencent Qidian Marketing Cloud’s MAGIC intelligent agent has been launched on the WorkBuddy open platform, enabling one entry point to connect marketing preparation, execution, deal closing, and review. Enterprises can also use Tencent Cloud ADP (Agent Development Platform) to customize business agents and carry out unified management and scheduling, so agents truly enter core business processes such as customer service, marketing, and operations.

Deploying agents across all scenarios also requires infrastructure and security capability support. Tang Daosheng said Tencent Cloud has high-performance and highly cost-effective agent infrastructure, stable heterogeneous computing clusters, and deep cooperation with multiple chip vendors to build a new domestic computing ecosystem. On the model side, Tencent’s Hunyuan Hy large model continues to evolve in real business scenarios, while also supporting customers via Tokenhub (Tencent Cloud large-model service platform) to call mainstream open-source large models—including Hunyuan, DeepSeek, GLM, MiniMax, and Kimi—through APIs (application programming interfaces).

Tencent's Tang Daosheng: WorkBuddy will open architecture to support enterprises in building their own Agent networks

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry