Office SaaS is becoming an adjunct to AI.

This morning, an internal email from ByteDance marked a phased (phased) end to Feishu's identity as an independent product.

The Feishu product team is fully integrated into Doubao, forming a new Doubao product team under the leadership of Doubao's head, Zhao Qi. Feishu's head, Xie Xin, will report to Zhao Qi. Meanwhile, Feishu's entire market, sales, and customer service teams will be incorporated into Volcano Engine, with Tan Dai overseeing the establishment of a unified ToB commercialization front desk.

This marks the largest ToB system restructuring at ByteDance since the implementation of business segmentation in 2021.

However, looking back a month earlier, it becomes evident that Feishu's structural overhaul was not an isolated move within the same circle. Alibaba had just consolidated three AI office agents into Qianwen Office, placing it under DingTalk's leadership. Tencent completed the structural integration of its Hunyuan large model team, with WeCom internally testing an intelligent assistant named 'Dayuan' and simultaneously advancing the commercialization of WorkBuddy Enterprise Edition.

In fact, before Feishu confirmed its structural adjustment, Xinmou had already learned relevant information from internal sources. The organizational adjustments by the three companies occurred almost simultaneously, seemingly as internal optimizations on the surface but actually reflecting a concentrated outbreak of industry consensus: the enterprise AI track has completed the first half of parameter competition and multi-point trial and error, and now it's racing to secure positions for the same endgame, requiring dispersed efforts to be consolidated.

01

AI Leadership:

Who is the Gateway, Who is the Foundation

The strongest signal from ByteDance's adjustment is that AI's leadership within the product system has been elevated to an unprecedented height.

Over the past few years, Feishu has been a star business in ByteDance's ToB portfolio, with an independent team and brand, targeting DingTalk and WeCom, aiming to become the next-generation collaborative office gateway.

However, after the restructuring, Feishu's product sovereignty has been ceded to Doubao, shifting its positioning (positioning) from an independent office SaaS to a carrier for Doubao's enterprise-level AI office scenarios. All of Feishu's capabilities, including documents, meetings, IM, and approvals, will ultimately serve the experience delivery of AI agents, becoming functional modules invoked by Doubao.

This logic has already been validated by the currently beta-testing Doubao Enterprise Edition.

The product natively integrates with all of Feishu's office scenarios, allowing AI to directly read document content, extract meeting minutes, organize group chat information, and complete Q&A and creation based on the enterprise's proprietary knowledge base, eliminating the need for users to switch between multiple systems.

ByteDance likely judges that in the future, the first work software employees open will not be a collaboration tool but an AI assistant, with office functions merely extending AI capabilities.

This mirrors Alibaba's integration path.

In July, Alibaba consolidated its three parallel AI agent products—QoderWork, Wukong, and MuleRun—into 'Qianwen Office,' placing it under the overall leadership of DingTalk CEO Chen Yusen.

Within Alibaba's ecosystem, DingTalk remains the core product, with large model capabilities serving as value-added items injected into office scenarios. DingTalk's accumulated organizational relationships, permission systems, and enterprise data are indispensable for AI implementation. The core task of Qianwen Office is to add AI paid options for existing customers, enhancing single-customer value.

The difference in leadership between the two companies essentially stems from their differing foundations.

DingTalk has Cumulative (accumulated) over 800 million users, serving more than 28 million organizations, and has established a deep customer base in government, manufacturing, education, and other industries.

By 2025, DingTalk's revenue exceeded 4 billion yuan, with a sufficiently solid foundation that it doesn't need to acquire customers from scratch with AI. Instead, it can convert massive free and low-paying customers into high-paying AI users, driving significant revenue growth.

However, ByteDance lacks such an existing customer base.

QuestMobile data shows that by 2026, Feishu's monthly active users will reach approximately 30 million, less than one-sixth of DingTalk's, ranking third in overall market share. Relying on Feishu's gradual customer expansion cannot keep pace with the competitive rhythm of the AI industry.

Conversely, leveraging Doubao's product reputation and AI technical capabilities accumulated in the C-end market to bring Feishu into the enterprise market may offer a differentiated breakthrough—after all, for many young users, Doubao's AI experience advantages are more strongly perceived than differences in traditional office functions.

Tencent, meanwhile, has taken a third route. After consolidating its technical team for the Hunyuan large model, its capabilities are simultaneously output to two product lines:

One is the WorkBuddy Enterprise Edition, focusing on pure AI collaboration, and the other is the intelligent assistant 'Dayuan' embedded in WeCom. Tang Daosheng has publicly explained the positioning differences between the two: WeCom emphasizes human-to-human communication, business processes, and customer connections, while WorkBuddy focuses on human-AI collaboration, with the two running in parallel and complementing each other.

This strategy aligns highly with Tencent's DNA. WeCom's ecological connectivity with WeChat is a barrier that the other two companies lack, allowing AI to extend from internal office operations to external customer management, helping enterprises drive growth rather than merely reducing costs—a unique differentiation space.

Microsoft's commercialization data may point the way for all players. As of the third quarter of fiscal year 2026, Microsoft 365 Copilot's paid enterprise seats surpassed 20 million, with 5 million added in a single quarter, and user weekly activity levels now on par with Outlook.

This earnings report has completely dispelled industry doubts about enterprise AI's willingness to pay, confirming the endgame: AI agents embedded in complete office workflows will become the core gateway for next-generation enterprise productivity.

All three domestic companies' organizational adjustments are essentially vying for this gateway in the Chinese market, albeit starting from different points and choosing vastly different paths.

02

The End of the Race: An Industry Shift from Trial and Error to Critical Breakthroughs

The near-simultaneous conclusion of internal competitions among the three companies essentially reflects that the industry has reached a stage where consolidation is necessary.

In the early days of generative AI's outbreak (breakout), no one knew the final product form or the optimal implementation direction. Major companies widely adopted a parallel team competition mechanism, encouraging internal trial and error to avoid missing any potential routes.

Alibaba simultaneously ran three highly overlapping AI office agents, each targeting desktop, enterprise embedding, and overseas markets, with separate R&D and operations teams;

Tencent's AI products were scattered across different business groups, with Yuanbao, WorkBuddy, CodeBuddy, and WeChat Xiaowei each forming independent systems;

ByteDance's Feishu, Doubao Enterprise Edition, and Volcano Engine each had independent product and sales teams, requiring enterprise customers to coordinate with three separate teams for full-link service procurement.

During the exploration phase, this model offered more advantages than disadvantages, enabling rapid iteration through internal competition and clarifying genuine market demands. However, by 2026, the industry's direction had become increasingly clear:

Office operations represent the most mature paid scenario for enterprise AI, with desktop intelligent agents being the highest-certainty product form. The focus now shifts to scalable delivery, commercialization efficiency, and customer retention. Continuing to maintain multiple competing teams would only cause internal resource drain, raise customer cognitive costs, and even lead to internal order snatching.

Alibaba felt this most directly. The three agent products had varying promotional narratives, confusing customers, while internal R&D resources were duplicated.

Shortly after Chen Yusen took over DingTalk, he initiated integration within his first week, completing the consolidation of the three product lines in 46 days and launching the unified Qianwen Office. The rapid progress underscores the group-level consensus that the trial-and-error phase has ended, necessitating concentrated resources to build a flagship product without further dispersion.

ByteDance's integration focuses more on commercialization efficiency. Previously, Feishu had its own sales and customer success systems, Volcano Engine focused on cloud services and compute power sales, and Doubao Enterprise Edition had separate business dock (connections).

Medium-to-large enterprise customers wanting to procure ByteDance's compute power, large models, and office software had to coordinate with three teams, leading to high communication costs and issues like customer overlap and inconsistent pricing internally.

Merging Feishu's GTM team entirely into Volcano Engine to establish a unified 'Creativity Service Platform' aims to provide customers with a single entry point, enabling a single sales team to deliver full-link solutions from compute infrastructure to large models and office applications.

Unifying the sales system is a prerequisite for scalable commercialization.

Alibaba Cloud possesses a mature sales network for government and large enterprises, allowing Qianwen Office to be directly bundled into existing cloud orders with extremely low marginal sales costs.

ByteDance lacks such a ready-made foundation. Volcano Engine's past customers were predominantly in the internet sector, with sales capabilities concentrated on compute power and technical services, lacking delivery and customer success experience for office SaaS. Incorporating Feishu's GTM team comprehensively supplements SaaS sales, delivery, and service capabilities, paving the way for large-scale promotion of Doubao Enterprise Edition.

Industry data also supports this shift. IDC predicts that by 2026, China's digital office platform software market will reach 57.69 billion yuan, growing 18.3% year-on-year, roughly in line with 2025's growth rate, indicating that the pure collaborative office market has transitioned from a Outbreak period (boom phase) to steady growth.

The next wave of growth will inevitably come from AI value-added services. Whoever completes internal integration and streamlines product and sales systems first will be the first to reap this growth dividend.

For startups, this is likely not a positive signal. When major companies were internally disorganized, startups had room for niche breakthroughs, surviving by differentiating in specific scenarios.

Now, with major companies consolidating resources and marketing full-suite solutions of 'compute power + models + office scenarios,' enterprise customers increasingly prefer one-stop procurement to reduce coordination costs, squeezing the survival space for pure tool-based startups. The track is rapidly concentrating toward top players, with the Matthew effect becoming more pronounced.

03

The Battle for Desktop Entry: A Life-and-Death Race in a Three-Year Window

All organizational adjustments ultimately point to the same battlefield: corporate employees' desktops.

The core gateway for next-generation productivity tools will not be a standalone document, IM, or approval software but a permanently resident desktop AI agent.

It can understand natural language instructions, automatically invoke all internal enterprise system data, and complete tasks like information organization, report writing, process approvals, and schedule coordination. Employees won't need to switch between a dozen software applications; most work can be accomplished through AI dialogue.

The window to secure this gateway is extremely short, with the industry generally estimating it at two to three years. The switching costs for enterprise-grade products are prohibitively high—once employees develop usage habits and enterprises accumulate corresponding workflows and data assets, replacement costs rise exponentially.

Microsoft has already gained a global first-mover advantage through Office's monopoly. The three domestic companies are now vying for China's first-mover position, knowing that falling behind even slightly may make catching up impossible.

Each company has its strengths and challenges.

Alibaba's advantage lies in its solid foundation. DingTalk boasts the broadest customer coverage and deepest industry penetration, especially in sectors like government, manufacturing, and education, which have stable payment capabilities, along with a mature service system. As long as Qianwen Office delivers a satisfactory product experience, it can leverage existing channels for push conversions with guaranteed efficiency.

However, challenges remain: SMEs dominate DingTalk's customer base, with varying payment willingness and capabilities. The penetration rate of AI value-added services remains uncertain. More critically, DingTalk has long been positioned as a management tool—whether it can create a native AI experience that sheds its cumbersome tool-like feel remains questionable.

ByteDance's strength is its outstanding AI product capabilities.

According to internal sources, Feishu had already internally embedded Doubao Enterprise Edition.

Doubao has accumulated excellent C-end user reputation, with interaction design, understanding capabilities, and response speeds ranking among China's best, highly accepted by young users. Leveraging its AI advantages to enter the enterprise market can easily create differentiated product perceptions, attracting tech, internet, and creative clients with high AI experience demands.

However, within the industry, Feishu's B-end foundation is relatively weak, with insufficient sales and delivery experience in government, traditional manufacturing, and building customer trust takes time. While Feishu has cultivated high-quality flagship customers among medium-to-large enterprises, its overall customer base remains small. Supporting Doubao Enterprise Edition's scalable growth requires a long accumulation period.

Tencent's ecosystem is irreplaceable. WeCom's connectivity with WeChat represents a barrier the other two cannot replicate. AI capabilities can not only enhance internal office efficiency but also extend to front-end customer service, sales conversion, and other operational scenarios, directly generating revenue for enterprises rather than merely cutting costs.

This value proposition is highly attractive to retail, catering, and service industry clients. Relatively, Tencent's challenge lies in its relatively dispersed product lines, with WorkBuddy and WeCom running in parallel. Whether internal collaboration efficiency can be maximized without internal competition remains to be seen over time.

All players face the same hurdle: activation rates.

Microsoft Copilot's paid seats are growing rapidly, but third-party statistics show that actual active usage rates for paid licenses hover around 35%, with many enterprises purchasing in bulk but employees failing to form usage habits, rendering them idle procurement items.

The domestic market will face the same issue. The concept of AI office is appealing, but truly integrating it into daily workflows and getting employees to actively and frequently use it is far more challenging than stacking parameters or giving demos. This requires deep industry scenario understanding and extensive, granular adaptation work—no shortcuts exist.

This is the deeper reason behind the three companies' simultaneous organizational integrations. Only with unified products, sales, and services can they focus on deep scenario cultivation, refine details, and improve activation and retention rates. The first half was about concepts and parameters, where even a single company's product launch could become a focal point. However, the second half will depend on whose solutions genuinely solve problems and make AI truly usable within enterprises.

From this perspective, this round of dense (intensive) organizational adjustments merely signals the halftime whistle.

— END — For more, click below to follow