百度、阿里、腾讯重逢AI办公,胜负手在数据Baidu, Alibaba, and Tencent Reunite in AI Office, the Deciding Factor is Data


AI free token/AI免费积分

本文来源:https://mp.weixin.qq.com/s/7XpAZ-Fowv0Id70QUKa1Hw
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

以下图文内容源自36氪(ID:wow36kr)。
图片
图片

一股熟悉的紧张气氛迎面而来。

图片

短短一周内,百度、阿里、腾讯三大巨头齐聚AI办公领域。

7月底,腾讯将WorkBuddy(了解更多)与CodeBuddy团队合并并升级,明确WorkBuddy(了解更多)为AI办公旗舰产品;8月3日,阿里将QoderWork、悟空、MuleRun三款产品整合为千问办公,并正式开启公测;8月初,百度完成AIWork产品线新一轮整合,形成以百度搭子、GenFlow、秒哒为代表的AI办公产品矩阵。

图片

熟悉的紧张气氛扑面而来,几年前声势浩大的BAT大战,似乎要以崭新的面貌再次上演。

这不禁让所有人都感到好奇,为什么这些大公司不约而同地在此时整合资源,停止“内部竞赛”?

过去两年的快速扩张表明,多条产品线并行虽然测试出不少单项能力,但也导致了算力分散、定位重复和用户认知混乱。当AI行业从展示技术转变为交付成果,大公司必须将模型、Agent、工具与数据能力整合到同一个旗舰平台上,集中资源进行攻坚战。

更深层次的考量在于,AI办公是当前大模型应用中确定性最高、商业闭环最明确的超级场景。

图片

从“竞争模式”转向“阵地争夺”

2026年AI领域最显著的变化,莫过于对话式交互模式的终结已成为现实。

这一变革的触发点可追溯至年初迅速走红的开源Agent框架OpenClaw(龙虾)。该框架之所以引发广泛关注并催生了全民参与的"养龙虾"现象,源于它开始深度融入用户电脑桌面环境,直接接管浏览器、本地文件及终端应用等核心功能。

人类首次看到了AI从被动应答向主动执行任务的转变潜力。

正是由于"AI能够自主完成任务"这一体验带来的颠覆性冲击,过去半年内市场上涌现出多代Agent创新应用。从自动化报表生成、跨平台数据采集,到全天候待命的虚拟助手,这些热门应用无一例外地瞄准了同一用户群体——以电脑为主要生产工具的职场专业人士。

而这群用户,正是过去十余年互联网行业各巨头不惜重金争夺并培养的核心高价值办公群体。他们掌握着企业最关键的工作文档与业务流程,对效率提升有着最直接的感知需求;同时,他们普遍具备出色的学习能力和持续进步的动力。这些用户的电脑桌面自然成为大模型技术商业化落地的最佳场景、最高粘性应用场域以及最深度数据资产积累的竞争阵地。

对于这些用户群体而言,其电脑桌面数十年来始终是各方势力争夺的战略要地。

从应用场景角度看,办公场景具有最明确的适配性。相较于容错空间小、周期长且流程难以标准化的工业应用场景,办公场景天然具备高频次、标准化程度高和数据沉淀量大的特点。无论是文档处理、数据分析还是代码编写,任务交付的标准极为清晰,Agent技术在此类场景能够率先实现"从对话交互到实际执行"的商业化应用突破。

从商业价值维度看,其市场潜力极其巨大。放眼全球市场,以OpenAI和Anthropic两家公司年化营收(ARR)双双逼近千亿美元规模为代表的国际高付费市场格局,更揭示了万亿人民币级别的长期发展天花板。

在国内现有市场环境中,数以亿计的电脑办公用户的SaaS订阅模式和个人端付费习惯正在迅速形成。

在个人用户端,消费者付费的思维方式正经历深刻变革。随着大型模型逐步接管桌面应用,AI办公工具正迅速从最初的猎奇聊天工具转变为不可或缺的生产力装备。高端白领及专业用户群体开始接受付费订阅模式,一旦电脑办公群体中占比10%-15%的核心高频用户转变为稳定付费群体,仅个人用户端的付费收入就能带来极为可观的稳定现金流。

在企业端,公司的采购方式也在不断升级。企业购买的不只是软件账号,更是能够替代重复工作的数字员工与自动化计算能力。无论是主流协同软件的AI增值服务,还是按Agent执行节点与Token消耗量结算的付费方式,都使企业客户的客单价提升了数倍。

按照这一发展趋势,以亿为单位的核心办公人群叠加企业级服务的算力消耗,国内AI办公市场将稳固形成百亿元人民币起步的基础规模。这不仅是大型科技公司实现商业闭环的现金流来源,更是其将业务拓展至全球万亿市场前最坚实的演练场。

从这一视角来看,互联网巨头布局AI办公,并非源于前几年流行的“FOMO(害怕错过)”心理,而是基于对未来趋势的明确判断。

从大公司竞争的视角分析,此刻集体押注AI办公是顺理成章的。大型模型的能力正在快速趋同。无论是推理能力、长文本处理还是Agent调度,顶级模型之间的差异已经不足以单独决定产品竞争力。因此,对于这些公司而言,持续开发更多Agent不再是重点,如何将模型能力有效整合,完成实际工作任务,已成为新的竞争核心。

多重因素共同作用下,目前百度、阿里、腾讯等公司再次站在同一起跑线上,而他们的目标几乎完全一致:致力于打造下一代生产力入口。

图片

为什么还是大厂?

自2023年ChatGPT掀起全球大型模型浪潮以来,创业公司数量激增。无论是研发基础大模型的独角兽企业、专注于特定场景的AI应用公司,还是主打通用Agent的新锐团队,都曾在聚光灯下备受瞩目。

但时至今日,当AI办公这场硬仗真正展开,人们发现最终占据舞台中心、掌握最高胜算与最大关注度的,依然是那几家互联网巨头。

为什么?

表面上看,AI办公工具似乎只是能够撰写文档、制作PPT、分析数据的Agent;但一旦真正应用于企业场景,Agent要完成一项任务,远不止是调用一次大型模型。它需要理解企业知识库和历史文件,需要连接微信、邮件、文档、日历、数据库等不同系统,在权限范围内调用各种工具,还要满足企业对数据安全、权限管理和私有化部署的要求。最终,它还需要借助成熟的销售和交付体系,真正融入企业的日常工作流程。

这对AI办公服务商提出了极为严苛的要求。再次印证了ToB领域那个经典的铁律:长板决定上限,短板决定生死。不同于容忍度较高的C端娱乐场景,正式的办公场景对准确性、稳定性和数据安全有着近乎零容忍的刚性标准。只有综合能力没有短板的参与者,才能真正获得这张入场券。

这几家大型企业在AI办公领域展现出强大的竞争力。具体分析下来,主要可以归纳为四个方面。

首先,是模型能力。这是所有AI办公产品的核心竞争力要素。文心大模型5.1在LMArena搜索榜中位列国内第一;腾讯混元由姚顺雨带领团队构建了集团级的RL基础设施;阿里通义系列则在Coding/Cowork方面持续加强。BAT对推理、Agent调度及Coding等能力的持续优化,为办公Agent提供了底层技术支撑。

其次,是数据资产。这恰恰是三家公司开始显现差异的关键点。Agent要有效完成工作,必须拥有丰富多元且真实可靠的上下文信息。腾讯最核心的数据来源于通信关系和协作网络;阿里积累的是企业组织、审批和流程数据;百度则拥有百度网盘沉淀的大规模工作文件和知识资产。三家公司掌握的数据类型存在差异,这也决定了它们选择了不同的发展路径。

第三,是工具生态。Agent的价值体现在能够跨工具执行任务。腾讯拥有微信、企业微信、腾讯文档等完备的产品矩阵;阿里则掌握了钉钉和阿里云企业生态;百度正逐步将搜索、网盘、伐谋、秒哒等产品能力整合到统一体系中。生态连接能力越强,Agent可以处理的任务就越复杂。

第四,则是企业交付能力。AI办公本质上属于To B业务。无论Agent能力有多强,都需要借助成熟的云服务体系、行业解决方案和销售渠道,才能将Agent成功部署到企业内部。这也是腾讯云、阿里云和百度智能云共同具备,而多数初创公司短期内难以构建的能力。

尽管如此,虽然BAT都将AI办公提升到集团级战略高度,但它们的起步点并不相同,这与各公司的企业特质密切相关。

腾讯选择了连接“人”的路径。

从微信、企业微信到腾讯文档,腾讯最显著的专长始终是建立人与人之间的连接。

WorkBuddy(了解更多)延续了这一理念,它更多扮演协作者角色:理解沟通内容、整合聊天记录、调用文档和会议工具,在人与人之间创造新的协作模式。关系链是腾讯最宝贵的资源,也是WorkBuddy(了解更多)最自然的切入点。

阿里则选择了连接“组织”的路径。

钉钉积累了丰富的企业内部组织架构、审批机制、项目协同及管理模式。千问办公并非白手起家,而是依托现有企业运营体系构建,使Agent能深度融入审批、项目管理、知识管理等核心业务流程。相较于个人效率应用,阿里巴巴更倾向于将Agent打造成企业数字化转型的有机组成部分。

百度则选择聚焦于“数据”的连接。更准确地说,是连接工作场景下持续产生的数据资源。

无论是文档、表格、图像、视频,还是会议纪要、项目构思,这些最终都会转化为工作文件。百度网盘长期积累的海量文件数据,构成了其进军AI办公领域的核心优势。

这一点决定了百度搭子(DuMate(了解更多))从设计之初就不是以聊天机器人的身份出现,而是被定位为能够理解文件、调用文件、处理文件,并能进一步执行任务的桌面级全能Agent。

图片

这三种路径并无绝对优劣之分。

人际通信关系,让Agent更懂团队协作模式;组织架构关系,让Agent更理解企业运作逻辑;文件数据关系,则让Agent更清楚具体工作的完成方式。

不过随着Agent执行任务能力的持续提升,数据的重要性正迅速凸显。因为判断一个Agent是否真正“高效”,往往看它能否获得足够丰富的上下文信息,而这些信息主要来源于实际工作中积累的数据,如文档、表格、代码、设计稿和项目文档等。

图片

百度的“数据反攻”

连接人员、连接组织、连接数据,代表三种不同的竞争策略。

百度接下来的规划是围绕“数据+Agent”重新打造其AI Work产品矩阵。今年8月完成的新一轮产品线整合,正是该战略落地的关键一步——3月份全面上线的桌面级通用Agent百度搭子(DuMate(了解更多))被正式推向核心位置,成为整个AI Work产品线的统一标杆与主要入口。

这确立了百度独一无二的竞争定位:百度的“百度搭子”直接建立在数据与执行力之上。它不依赖任何现有软件,而是一个自带工具、能直接掌控桌面的全局智能中心。

图片

百度搭子之所以能抢占先机,关键在于其背后那套高度整合的集团军协同策略:

其基础拥有海量数据资源。百度网盘积累的10亿用户与超1000亿GB工作文件,为Agent履行职责提供了最关键的“原材料”。在制作PPT或分析表格时直接提取并解析网盘内容的能力,形成了竞争对手短时间难以逾越的数据护城河。

左侧有GenFlow强化深度场景应用。基于1亿月活用户验证的场景能力,GenFlow擅长利用网盘资源进行长文档精炼、报告创作与视频剪辑,这些场景化功能为百度在AI办公领域的竞争提供了独特的技术支撑。

右侧有秒哒拓展功能边界。拥有3500万+用户与350万+应用创建的秒哒平台,提供了无门槛的零代码开发方式,使Agent在应对个性化、跨平台的长尾业务需求时,能够即时生成微应用与轻量化工具来完成执行任务。

凭借产品版本的快速迭代(实现“每日更新”并迅速整合微信、飞书、如流等主流应用)以及“安全沙箱+文件夹级别权限管控”的保护机制,百度搭子不仅在WAIC 2026上获得了通用Agent领域唯一的“镇馆之宝”认证,还成功实现了从单一产品到平台化生态的转变。

这种产品矩阵的整合对整个行业的发展具有启发作用。

在此之前,互联网巨头的AI办公产品普遍存在“创新多但专精不足、用户体验割裂感强”的问题,用户需要在线下盘查找文档,上文库撰写大纲,再切换到第三方应用制作PPT,数据在不同应用间反复流转。百度将这类深度场景能力整合到百度搭子中,本质上是在大厂内部停止了重复开发与入口内耗。通过构建统一的桌面级Agent,“百度搭子”成为全局调度中枢,所有复杂计算、数据调用与工具生成都在后台自动闭环,实现了真正的一站式交付。

在商业角度,这一重构方案能够从根本上解决SaaS时代遗留下来的付费困境与交付难题。

传统的B端SaaS模式在中国长期遭遇付费意愿薄弱、ARPU(每用户平均收入)难以提升的挑战。百度搭子不再向用户售卖零散的单点插件或“聊天机器人”账号,而是依托“桌面接管+实时工具构建+私有资产调阅”的功能,直接向用户交付具备端到端执行能力的“数字员工”。对于高净值白领及企业客户而言,为直接节省的工作时间与确定性成果交付付费,其商业价值远超单纯的软件订阅。

这种产品架构的整合,与李彦宏在Create 2026大会上提出的“AI竞争核心已从模型能力转向执行能力”的观点高度一致。

百度搭子作为统一的接入点,使得每一次对长文档的精炼、复杂表格的分析以及微应用的开发,都转变为高价值的智能体任务。这种高频率的桌面级交付,不仅能够构建“任务越复杂交付价值越高使用者粘性越强”的商业闭环,还能在高频率的互动中积累高质量数据,反馈到底层大模型的推理与规划能力,形成持续的正向循环。

图片

终局仍有悬念

AI的加速发展态势持续加强,站在当前时刻回望,BAT三大公司大模型的参数量差异与排行榜位次的变动,早已成为上一时代的过时议题。

竞争焦点悄然转移至AI办公领域。

从三大公司的战略举措来看,整合已成为必然选择:腾讯已完成组织架构升级,将WorkBuddy(了解更多)提升为云机型六部;百度则在8月加速完成新一轮整合,将百度搭子打造为统一接入点;而阿里的千问办公则于8月3日全面启动公测,整合工作预计于年底完成。

然而截至目前,三大公司仍需继续解答一系列难题。

例如,百度的智能云的企业级销售渠道,能否匹配百度搭子每日更新的机型进展?百度搭子又能否真正与GenFlow和秒哒实现深度协同,在AI办公领域创造1+1+1>3的协同效应?

在PC互联网时期,BAT依靠搜索、电商与社交三大领域划分市场,持续竞争十年未决出最终胜负。但在AI办公这一代,由于Agent机型的迭代速度与数据飞轮效应是上一代的十倍,这场竞争或许在两年内就能分出胜负。

而百度搭子能否将机型优势快速转化为“企业客户最多”的商业优势,将直接决定百度在此次竞争中,是成为领跑者,还是跟随者。

图片
以上图文转载自36氪(ID:wow36kr)。

【图文来源】微信公众号:36氪


本文图文来源于网络,版权属于原作者或网站,内容为作者观点,并不代表本公众号认同其观点及对其真实性负责。如有版权等问题,请与小编邮箱联系,将即刻启动相应处理。

【远方的家】精选中国科学院汪寿阳研究员及其学生的日常随笔,分享各类热点资讯、趣闻轶事、最新消息。
微信号:homeofcas
投稿或提出建议,请直接回复或发送邮件至:amssmadis@163.com

Source: https://mp.weixin.qq.com/s/7XpAZ-Fowv0Id70QUKa1Hw. If you have copyright questions, please contact 13951431913#139.com or submit a comment in the bottom right.

The following images and text are from 36Kr (ID: wow36kr).

图片

图片

A familiar sense of tension is filling the air.

图片

Within just a week, the three giants of Baidu, Alibaba, and Tencent have all made significant moves in the AI office field.

At the end of July, Tencent merged the WorkBuddy and CodeBuddy teams and upgraded them, clearly positioning WorkBuddy as the flagship AI office product. On August 3, Alibaba integrated QoderWork, Wukong, and MuleRun into Thousand Ask Work (Qianwen Office) and officially launched a public beta test. In early August, Baidu completed a new round of integration for the AIWork product line, forming an AI office product matrix represented by Baidu Buddy (DuMate), GenFlow, and MioDa.

图片

The familiar tension is back, reminding us of the high-profile BAT wars from a few years ago, which now appear to be resuming with a fresh look.

This has naturally piqued the curiosity of everyone. Why are these major companies simultaneously integrating resources and ending their "internal competitions" at this specific moment?

The rapid expansion over the past two years has shown that while running multiple product lines in parallel tests out many individual capabilities, it also leads to scattered computing resources, overlapping positioning, and confusing user cognition. As the AI industry shifts from demonstrating technology to delivering results, big companies must integrate modeling, agent, tool, and data capabilities into a single flagship platform to concentrate resources for a breakthrough battle.

A deeper consideration is that AI office is currently the super-scenario with the highest certainty and the most clearly defined commercial closed loop in large-model applications.

图片

Shifting from "Competition Mode" to "Territory Seizure"

The most noticeable change in the 2026 AI landscape is that the era of dialogue-based interaction modes is coming to an end.

This transformation was triggered by the rapid rise of the open-source Agent framework OpenClaw earlier this year. The framework gained widespread attention and sparked a nationwide "raising lobster" phenomenon because it began deeply integrating with users' desktop environments, directly taking control of core functions like browsers, local files, and terminal applications.

Humanity saw, for the first time, the potential for AI to shift from passive responses to active task execution.

It is precisely this disruptive experience — the realization that "AI can autonomously complete tasks" — that has led to the proliferation of innovative Agent applications over the past six months. From automated report generation and cross-platform data collection to full-time virtual assistants, these hot new applications are all targeting the same user group: white-collar professionals who primarily use computers for their work.

These users are the high-value office workers that internet giants have been spending significant resources to attract and cultivate over the past decade. They hold the key to critical business documents and workflows, have a direct need for efficiency improvements, and generally possess strong learning abilities and a drive to progress. Their computer desktops have naturally become the optimal battleground for the commercialization of large-model technology, serving as the best scene for application, the highest-engagement platform, and the most profound source of data asset accumulation.

For this user group, the computer desktop has historically been a strategic point of contention for various forces.

From the application-scenario perspective, office environments are the most obviously suitable candidates. Unlike the industrial sector, which has low error margins, long cycles, and difficult-to-standardize processes, office scenarios are characterized by high frequency, high standardization, and massive data deposits. Whether it is document processing, data analysis, or code writing, the standards for task completion are extremely clear, making office spaces the ideal place for Agent technology to first achieve a breakthrough in commercialized "talk-to-action" applications.

Looking at the commercial value dimension, the market potential is enormous. The global market, exemplified by the $100 billion annual recurring revenue (ARR) figures reported by OpenAI and Anthropic, demonstrates an ultimate ceiling in the tens of trillions of yuan.

In the domestic market, the SaaS subscription model and individual user payment habits for hundreds of millions of computer office users are rapidly forming.

On the individual user side, consumer payment mindsets are undergoing a significant shift. As large models increasingly take over desktop applications, AI office tools are quickly evolving from novelty chat tools into indispensable productivity equipment. High-end white-collar workers and professional users are beginning to accept subscription models, and once the core 10% to 15% of high-frequency users within the office community become stable payers, the personal user sector alone will generate a substantial, stable cash flow.

On the enterprise side, procurement methods are also upgrading. Companies are not just buying software accounts; they are purchasing digital workers and automation computing power that can replace repetitive labor. Whether it is premium services for mainstream collaborative software or payments based on Agent execution nodes and Token consumption, the average order value for enterprise clients has increased significantly.

Following this trend, the combination of a 100+ million core office population and enterprise-level service computing power will solidify a foundational scale of at least tens of billions of yuan for the domestic AI office market. This is not only the cash flow source for big tech companies to achieve commercial closure, but also the most solid training ground for extending their reach to the trillion-dollar global market.

From this perspective, the major internet companies' investment in AI offices is not driven by the "FOMO (fear of missing out)" psychology that was prevalent a few years ago, but rather by a clear judgment of future trends.

Analyzing the competition between large companies, the collective bet on AI offices at this moment is a logical step. The capabilities of large models are rapidly converging. The differences between top models in terms of reasoning, long-text processing, or Agent dispatching are no longer sufficient to independently determine product competitiveness. Therefore, for these companies, the focus has shifted from developing more Agents to effectively integrating model capabilities to complete actual work tasks.

Under the influence of these multiple factors, Baidu, Alibaba, and Tencent now find themselves on the starting line once again. Their goals are almost identical: to build the next generation of productivity gateways.

图片

Why are big companies still leading the charge?

Since the global large-model wave triggered by ChatGPT in 2023, the number of startup companies has surged. Whether they are unicorns developing foundational models, companies specializing in specific vertical-scene AI applications, or new teams focusing on general-purpose Agents, all have been under the spotlight.

However, as the hard battle for the AI office space begins, it is clear that the final contenders occupying the center stage and holding the highest stakes and attention are still the established internet giants.

Why is this the case?

On the surface, AI office tools appear to be simply Agents capable of writing documents, making PowerPoint presentations, and analyzing data. But once these tools are applied to enterprise scenarios, the reality is much more complex. For an Agent to complete a task, it is not just a matter of making a single call to a large model. The Agent must understand the corporate knowledge base and historical files, connect with various systems like WeChat, email, documents, calendars, and databases, invoke tools within the appropriate permissions, and satisfy the enterprise's requirements for data security, access control, and private deployment. Furthermore, the product must be delivered through mature sales and implementation channels to truly integrate into the company's daily workflow.

These requirements place extremely high demands on AI office service providers. The classic rule of ToB (business-to-business) markets is proving true once again: strengths determine the upper limit, while weaknesses determine survival. Unlike the consumer entertainment sector, which is highly tolerant, formal office scenarios have near-zero tolerance for errors, instability, or data insecurity. Only participants with a full suite of capabilities and no obvious shortcomings can secure a valid ticket to the game.

These large enterprises demonstrate significant competitive advantages in the AI office area. Specifically, this boils down to four key factors.

First is model capability. This is the core element of competitiveness for any AI office product. Baidu's Model 5.1 ranks first in China on the LMArena search ranking; Tencent's Hunyuan was built by Yao Shunyu's team, leveraging the group's RL infrastructure; Alibaba's Tongyi series continues to strengthen in the Coding/Cowork domain. The continuous optimization of reasoning, Agent dispatching, and coding capabilities by BAT provides the fundamental technological support for office Agents.

Second is data assets. This is where the three companies begin to show unique divergences. For an Agent to effectively complete work, it must have access to rich, diverse, and reliable contextual information. Tencent's most critical data comes from communication relationships and collaboration networks; Alibaba's data is centered around enterprise organizations, approvals, and process information; Baidu's strength lies in the massive accumulation of work files and knowledge assets on Baidu Netdisk. The differences in data types dictate that the companies will follow distinct development paths.

Third is the tool ecosystem. The true value of an Agent lies in its ability to execute tasks across multiple tools. Tencent has a complete product matrix including WeChat, Enterprise WeChat, and Tencent Docs; Alibaba controls the DingTalk ecosystem and Aliyun enterprise services; Baidu is gradually integrating search, netdisk, Vamoo, and MioDa into a unified system. The stronger the ecological connectivity, the more complex the tasks an Agent can handle.

Finally, there is the enterprise delivery capability. At its core, AI office is a ToB business. No matter how powerful an Agent is, it needs to be deployed within enterprises via mature cloud service systems, industry solutions, and sales channels. This is a capacity that Tencent Cloud, Alibaba Cloud, and Baidu Smart Cloud possess, but which most startups will struggle to build in the short term.

Despite these advantages, the starting points for BAT are not identical, as they reflect the distinct characteristics of each company.

Tencent chose to connect the "people."

From WeChat to Enterprise WeChat and Tencent Docs, Tencent has always excelled at establishing connections between individuals.

WorkBuddy inherits this philosophy. It acts more as a collaborator, understanding conversation content, integrating chat histories, calling document and meeting tools, and creating new collaboration patterns between people. The relationship chain is Tencent's most valuable resource and the most natural entry point for WorkBuddy.

Alibaba chose to connect the "organization."

DingTalk has accumulated rich data on internal corporate structures, approval mechanisms, project collaboration, and management modes. Qianwen Office is not starting from scratch; it is built on top of existing enterprise operation systems, allowing the Agent to deeply integrate into core business flows such as approval, project management, and knowledge management. Compared to individual efficiency applications, Alibaba is aiming to make the Agent an integral part of corporate digital transformation.

Baidu chose to focus on connecting the "data." More precisely, it connects the continuous flow of data generated in work scenarios.

Whether it is documents, spreadsheets, images, videos, meeting minutes, or project drafts, all these eventually become work files. The massive file data accumulated on Baidu Netdisk forms the core advantage for Baidu's entry into the AI office market.

This decision means that Baidu Buddy is not designed as a simple chatbot; it is a desktop-level omnipotent Agent that understands, calls, processes, and executes tasks based on files.

图片

There is no absolute superiority among these three paths.

Interpersonal relationship data helps the Agent understand team collaboration patterns; organizational structure data helps it comprehend corporate logic; file data makes it aware of how specific work is actually done.

However, as the capability of Agent execution continues to improve, the importance of data stands out. The efficiency of an Agent is often measured by whether it can access sufficient contextual information, which is primarily derived from the accumulated data of actual work, such as documents, spreadsheets, code, design drafts, and project papers.

图片

Baidu's "Data Counterattack"

Connecting people, connecting organizations, and connecting data represent three distinct competitive strategies.

Baidu's upcoming plan is to rebuild its AI Work product matrix around the axis of "Data + Agent." The recent integration of product lines in August 2026 is the key step in implementing this strategy. The desktop-level general-purpose Agent Baidu Buddy, which was fully launched in March, has been elevated to the core position, becoming the uniform standard and main portal for the entire AI Work product line.

This establishes a unique competitive position for Baidu: Baidu Buddy is built directly on data and execution power. It does not depend on any existing software; it is an independent intelligence center that manages its own toolbox and directly controls the desktop.

图片

Baidu Buddy's ability to seize the initiative is thanks to a highly integrated group collaboration strategy:

First is the foundation of massive data resources. The accumulation of 1 billion users and over 100 billion GB of work files on Baidu Netdisk provides the crucial "raw material" for the Agent to perform its duties. The ability to directly extract and parse Netdisk content when making PPTs or analyzing spreadsheets creates a data moat that competitors cannot easily cross in the short term.

Second is the left-flank reinforcement by GenFlow. Based on the verified capabilities of 100 million monthly active users, GenFlow specializes in leveraging Netdisk resources for long-document refining, report writing, and video editing, offering unique technical support for Baidu's competition in the AI office space.

Third is the extension of boundaries by MioDa on the right flank. With over 35 million users and 3.5 million applications created, the MioDa platform provides a no-code development method that allows the Agent to generate micro-apps and lightweight tools on demand to meet the needs of specific, cross-platform long-tail business scenarios.

Fourth is rapid iteration. By achieving daily updates in the product version and quickly integrating mainstream applications like WeChat, Feishu, and Ruxin, Baidu Buddy has transformed from a single product into a platform-based ecosystem. Coupled with a "safe sandbox + folder-level permission control" protection mechanism, Baidu Buddy not only won the sole "Town House Treasure" certification for the general-purpose Agent field at WAIC 2026 but also successfully transitioned to a platform-level service.

The integration of this product matrix is highly instructive for the industry.

Before this, the AI office products from tech giants often suffered from being "innovative but lacking focus, with disjointed user experiences." Users would search for documents on one platform, write outlines in a knowledge base, and then switch to a third-party application to make a PPT, causing data to flow back and forth between different applications. Baidu addressed this by integrating deep scene capabilities into Baidu Buddy. Essentially, it is putting an end to repeated development and entry-point fragmentation within the company. By building a unified desktop-level Agent, Baidu Buddy becomes the central dispatcher where all complex calculations, data calls, and tool generations are automatically closed-loop, realizing true one-stop delivery.

From a commercial perspective, this restructuring fundamentally solves the payment difficulties and delivery challenges of the SaaS Era.

The traditional B-end SaaS model in China has long suffered from poor payment willingness and low ARPU. Baidu Buddy is no longer selling scattered plug-ins or "chatbot" accounts; it is delivering "digital workers" capable of end-to-end execution based on desktop control, real-time tool building, and private asset review. For high-net-worth white-collar workers and enterprise clients, paying for the direct saving of work time and the certainty of results is far more valuable than a simple software subscription.

This integration of the product architecture aligns perfectly with Robin Li's statement at the Create 2026 conference that "the core of AI competition has transitioned from model capability to execution capability."

As a unified entry point, Baidu Buddy turns every task of refining long documents, analyzing complex tables, or developing micro-apps into a high-value AI task. This high-frequency delivery not only builds a business cycle where "the more complex the task, the higher the delivery value, and the stronger the user stickiness" but also accumulates high-quality data from interactions that feed back into the bottom-layer model's reasoning and planning abilities, creating a continuous positive cycle.

图片

The Endgame is Still Unsettled

The acceleration of AI development continues to intensify. Looking back from this current moment, the differences in parameter counts between the BAT large models and their positions on the leaderboards are already outdated topics from a previous era.

The focus of the competition has quietly shifted to the AI office domain.

Based on the strategic moves of the three companies, integration has become inevitable: Tencent has upgraded its organizational structure to promote WorkBuddy to the sixth major division of cloud computing; Baidu accelerated its integration in August, positioning Baidu Buddy as the unified entry point; while Alibaba's Qianwen Office entered a full open beta on August 3, with integration work scheduled to complete by the end of the year.

However, as of now, there are still a series of questions that the three companies need to answer.

For instance, can Baidu's enterprise-level sales channels in Baidu Smart Cloud keep up with the rapid, daily updates of the Baidu Buddy model? Can Baidu Buddy truly achieve deep synergy with GenFlow and MioDa to create a 1+1+1>3 effect in the AI office space?

During the era of PC internet, BAT dominated separate markets (search, e-commerce, and social networking) and fought for over a decade without a clear winner. In this AI office era, however, the iteration speed of Agent models and the data flywheel effect are orders of magnitude faster. It is highly likely that the outcome of this battle will be determined within the next two years.

Whether Baidu can transform the advantages of its "Buddy" model into the commercial advantage of having the "most enterprise customers" will directly determine whether Baidu emerges as the leader or the follower in this competition.

图片

The above text and images are reprinted from 36Kr (ID: wow36kr).

[Article Source] WeChat Public Account: 36Kr

The images and text in this article are sourced from the internet. Copyright belongs to the original authors or websites. The content represents the views of the authors, and this public account does not endorse the opinions or guarantee the authenticity of the information. If there are any copyright issues, please contact the editor's email, and appropriate action will be taken immediately.

[Distant Home] Curated by Chinese Academy of Sciences researcher Wang Shouyang and his students, this account shares daily essays, trending news, anecdotes, and the latest information.

WeChat ID: homeofcas

For submissions or suggestions, please reply directly or email: amssmadis@163.com

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