On the morning of July 30, 2026, an internal email from ByteDance effectively rewrote a decade of Feishu's history. The message was straightforward: the Feishu product team would merge with the Doubao product team to form a new Doubao product team, led by Doubao's head Zhao Qi, with Feishu CEO Xie Xin reporting to Zhao Qi. Meanwhile, Feishu's GTM (go-to-market, sales, and customer service) teams would integrate with the Volcano Engine team to create a new ToB organization called the "Creativity Service Platform." Feishu, once a standalone business unit alongside Douyin and Volcano Engine, thus bid farewell to its independent era, becoming a single piece in ByteDance's AI puzzle.

This restructuring came at a time when Feishu appeared to be at its strongest, with reported revenue exceeding 3 billion RMB in 2025 and year-over-year revenue growth of over 100% in the second quarter of 2026. Why would a high-growth, star business be reorganized at its peak? The answer lies within ByteDance's internal strategic focus on AI and increasing external competitive pressure.

Why 'Dismantle' Feishu: From a Standalone SaaS Narrative to an AI Ecosystem Puzzle

Feishu's creation was itself a product of ByteDance's strategic cycle. Around 2016, ByteDance was in a high-growth phase, and its existing, scattered office tools were no longer suitable for the company's operational rhythm. Then-Vice President of Human Resources Xie Xin proposed developing an in-house office tool. In 2019, Feishu was officially launched externally, transitioning from an internal tool to a market product. In 2021, Feishu was elevated to an independent business unit, becoming a key lever in ByteDance's ToB strategy.

After a decade of development, Feishu posted impressive results. According to public reports, Feishu's ARR exceeded 3 billion RMB in 2025 and grew over 100% year-on-year in the second quarter of 2026. Industry insiders speculated that if this growth momentum continued, Feishu could potentially become the top revenue earner in China's collaborative office market within 2026. By the second quarter of 2026, over 90% of Feishu's new customers were also purchasing Feishu's AI products.

However, beneath these bright numbers lay hidden concerns. Feishu's team of approximately 4,000 people carried an annual labor cost of 3.2 to 4 billion RMB, with Xie Xin once extending the profitability target to 2030. Against the backdrop of ByteDance's planned AI capital expenditure exceeding 200 billion RMB in 2026, Feishu's "independent narrative" appeared excessively extravagant. A business generating billions in annual revenue but with a distant path to profitability was inevitably going to face a strategic reassessment under the group's laser focus on AI.

A deeper reason lies in how large language models have rewritten the logic of enterprise software. Before the explosion of LLMs, Feishu's core value was enhancing organizational efficiency through advanced tools—documents capturing knowledge, meetings forming decisions, knowledge bases accumulating experience, and multi-dimensional tables organizing data. The core of enterprise software competition revolved around "organizational efficiency." But with the advent of generative AI and agents, this logic has been upended. The ceiling for office productivity products is no longer determined by UI experience or feature modules, but by the cognition and execution capabilities of the underlying large model. The most acute pain point for enterprises is no longer "how information flows," but "can AI read my enterprise knowledge, understand organizational permissions, invoke business tools, and ultimately get the job done?"

Model, computing power, data, tools, and interface—these five layers are all indispensable for an Agent that can truly work inside an enterprise. Volcano Engine provides the model and computing power, Feishu provides the scenarios and permissions, and Doubao provides the interactive interface. By integrating the three, ByteDance can place the "brain, infrastructure, work environment, and interactive interface" into a single unified system.

In recent years, both Feishu and Doubao have been exploring enterprise AI directions. Feishu launched the AI application development platform Aily, while Doubao also continued to invest in enterprise-level model services. However, the two business lines remained relatively independent, leading to resource dispersion. More critically, the professional version of Doubao has been moving increasingly deeper into office scenarios, incorporating built-in Office suites, supporting Agent tasks, and possessing document processing and data analysis capabilities, blurring the boundaries with Feishu itself. The enterprise version of Doubao, co-developed with the Feishu product team, has already begun internal testing with some Feishu customers, natively integrating with Feishu's documents, spreadsheets, meetings, and group chats.

On the commercial front, Feishu sells SaaS, while Volcano Engine sells cloud computing and MaaS. Facing the same large enterprise clients, two separate sales teams created overlaps, increasing communication costs and preventing the formation of a consolidated force. Such organizational friction became increasingly unacceptable in the AI era. ByteDance CEO Liang Rubo stated clearly at the Volcano Engine conference in June 2026: "Climbing the peak of AI is ByteDance's most important task right now... climbing requires focus. So in the past few years, ByteDance has been focusing and shrinking the breadth of non-core businesses, concentrating efforts on AI, and within AI, focusing on improving model capabilities."

As of June 2026, the daily token call volume for the Doubao large model had surpassed 180 trillion, a more than tenfold increase from the previous year. Doubao's monthly active users reached 382 million in June, exceeding the combined total of the second and third-place competitors. Volcano Engine ranked first in the IDC China Public Cloud MaaS Service Market with a 49.5% market share. Based on the average consumption in July, ByteDance's large model business ARR has reached $4 billion, surpassing the combined ARR of all other domestic model companies. When Doubao has become the largest C-end AI application in China, and Volcano Engine is already the largest MaaS platform, Feishu's role must be redefined: from an independent SaaS product, downgraded to a capability component within the AI ecosystem.

Post-Integration: The 'Endgame Battle' for AI Office and ByteDance's Ambition

Following this adjustment, ByteDance's ToB business has formed a clear "trinity" structure: Doubao becomes the core engine and unified interface for the AI era; Feishu steps back to provide office scenarios and enterprise knowledge; and Volcano Engine takes over the commercial front end. On the product side, the entire Feishu product team has been integrated into Doubao, managed uniformly by Zhao Qi. Feishu has been downgraded from an independent business group to a capability component of the technology middle platform, with its team moving from the front stage to the backstage. On the commercial side, the Feishu GTM team has been integrated with Volcano Engine to form the "Creativity Service Platform," which will be responsible for the unified market sales of ByteDance's MaaS and SaaS offerings.

The significance of this structure is that it allows for customer acquisition through a holistic package: "base computing power (Volcano) + large model (Doubao) + front-end application (Feishu)." When clients use Feishu, they continuously consume Volcano Engine's tokens and computing resources, breaking the traditional bottleneck of SaaS profitability. ByteDance stated that this is to strengthen the product and service collaboration of Doubao, Feishu, and Volcano Engine in enterprise productivity scenarios. Summarized in one sentence: integrating Feishu's product into Doubao is to expand ByteDance's super interface; integrating Feishu's commercial team into Volcano Engine is to maximize token consumption.

The integration of Feishu and Doubao is not an isolated event. Around the same time, Alibaba integrated its three Agent products—QoderWork, Wukong, and MuleRun—to launch "Qianwen Office." Tencent's WorkBuddy has already surpassed 20 million visits in a single month, exceeding the combined total of the second and third-place competitors. Within ten days, these three major tech giants completed organizational restructurings of their AI office product lines. This reflects a shared judgment: the "collaborative office" category as an independent product is disappearing, and the "desktop war" for AI office has fully erupted.

Entering 2026, the competitive focus in the large model arena has shifted from basic parameter scale to the deep penetration of Agents into productivity scenarios. Competition has evolved from a battle between products to a comprehensive contest involving models, applications, cloud computing, and commercialization. The prize is no longer just an AI product, but the primary daily interface for enterprise work. ByteDance's decision to "dismantle" Feishu and embed it within Doubao and Volcano Engine is precisely an attempt to gain an early advantage in this "desktop war." In the past, the "trump card" of Feishu did not help ByteDance secure the top spot in the battle for the AI office interface. So, ByteDance chose to use Doubao to integrate Feishu, redefining the office interface with AI capabilities.

ByteDance's move is not unique but reflects a global trend in the AI industry. The integration of vertical applications into AI chat applications is becoming a trend in the tech world. In April, Google integrated its personal AI research assistant NotebookLM into Gemini. On July 10, OpenAI launched a new version of its ChatGPT desktop application, integrating three major scenarios: Chat conversation, Work productivity, and Codex programming, with the previously independent Codex desktop app ceasing standalone operation. Similarly, Anthropic's Claude desktop client integrates Chat, CoWork, and Code modules. The thinking among major AI companies is highly consistent: while AI office is excellent, it is best operated within the traffic gateway of a chatbot. The chatbot remains the most intuitive and most widely used mode of AI service access—ChatGPT has 1 billion weekly active users, and Doubao has 382 million MAU. ByteDance's decision to have Doubao lead the integration with Feishu shows its clear intention: Doubao will be the primary offensive player in ByteDance's AI office strategy.

Feishu has not disappeared. ByteDance stated that after the restructuring, existing Feishu products and services will remain unchanged, and Feishu will engage in even deeper product collaboration with Doubao in productivity scenarios. However, Feishu's role has fundamentally changed. It is no longer an independent business group, no longer possessing its own product definition authority or commercial autonomy. It has become a built-in capability of the Doubao Enterprise Edition and a module within the overall Volcano Engine solution. The "collaborative office" category itself is transitioning from being a "destination" to a "tool to be invoked." The enterprise collaboration data accumulated by Feishu over a decade—documents, spreadsheets, meetings, group chats—will become the foundation for the Doubao large model to understand real-world enterprise business scenarios. Feishu's office scenarios and customer resources are being more deeply integrated into ByteDance's AI system.

From one perspective, Feishu has been "downgraded"; from another, it has been given a "rebirth." It is no longer a SaaS product fighting alone, but an indispensable capability layer within the entire ByteDance AI ecosystem. As one observer noted: "A business with a few billion RMB in ARR that cannot become an interface itself can only unleash greater power by merging with Doubao and Volcano." The merger of Feishu and Doubao represents a critical strategic decision for ByteDance in the AI era. It marks a shift from "multi-track horse racing" to "concentrated force," and from a standalone SaaS narrative to an AI ecosystem integration. This restructuring dismantles not only Feishu's independent organization but also the departmental walls between traditional enterprise software and AI.

For ByteDance, this is a strategic bet on the ToB landscape of the AI era: the fight for the AI work interface will determine the pecking order in the enterprise services market for the next decade. For Feishu, saying goodbye to its independent era may not be an endpoint. Becoming a puzzle piece in ByteDance's AI blueprint might be the optimal solution it finds in the AI age. AI is redefining everything—including what it means to be an "advanced enterprise." In this restructuring, no business unit can stand alone, not even a company generating 3 billion RMB in annual revenue like Feishu.