By | Beyond the Layout, Author｜Huahua
This may be the most important day since Feishu's inception ten years ago.
On the morning of July 30th, ByteDance sent an internal email announcing a major reshuffle of its organizational structure:
This means that from today, Feishu will no longer operate independently and will no longer have its own independent sales and marketing teams.
An independent business that ran for ten years, ended with a single email.
I. Did Feishu Lose?
The first reaction for many might be to think so. After all, Feishu is one of ByteDance's most heavily invested innovative businesses. Liang Rubo stated in 2023 that the R&D investment in Feishu and Volcano Engine is no less than that in Douyin and TikTok.
A business with investment no less than Douyin is today completely dismantled and restructured. Isn't that losing?
But looking at the timeline reveals a different story.
Over the past few years, Feishu has been fighting a war of independence. It wanted to prove itself as China's Office, China's Slack, China's Notion, China's Microsoft 365...
In one sentence, it wanted too badly to be a super gateway.
And the result?
Google embedded Gemini into Workspace. Microsoft used Copilot to take over the entire Office experience. Tencent directly infused AI into WeCom and Tencent Docs. Alibaba let Tongyi Qianwen cover office scenarios.
Today, almost all global leading companies simultaneously made one move: abandoning the gateway logic and embedding AI directly into workflows.
This also made Feishu suddenly realize that its long-held platform logic was becoming increasingly difficult to sustain in the fight.
So what happened today is not Feishu being swallowed up; it's Feishu shrinking for the first time from an independent product to an infrastructural capability.
It's not that Feishu lost; it's that AI won. AI won against everyone, including the very thing Feishu aspired to be.
Recalling Feishu's organizational streamlining in 2024, shifting from broad coverage to focusing on key accounts. Many interpreted it as a contraction back then. Actually, it wasn't.
Perhaps at that juncture, Feishu had already realized that simply selling tools couldn't win this battle; it needed to be embedded within a larger, more fundamental organizational framework.
Today, that framework has arrived. Except it wasn't Feishu that found it; ByteDance chose it for Feishu.
II. Why Today?
The answer lies in a set of data.
As of June 2026, Doubao Large Model's daily token consumption exceeded 180 trillion. Volcano Engine is already a leading player in China's public cloud MaaS market. Calculated based on average consumption in July, ByteDance's large model business annualized revenue reached 40 billion USD, exceeding the combined annualized revenue of other domestic model companies.
And Feishu's side?
In Q2 2026, over ninety percent of Feishu's new customers simultaneously purchased Feishu AI products.
This number is the key. It indicates that enterprise customers have already voted with their feet; they are not coming to buy Feishu, they are coming to buy AI.
Feishu is merely a channel through which AI reaches enterprises.
ByteDance had to act at this juncture. The reason is simple: AI has become the sole entry point for enterprise procurement.
Enterprises don't buy office software and cloud platforms separately, nor do they only purchase large model APIs; enterprises buy a suite of AI solutions.
Whoever can bundle AI capabilities, office tools, and cloud infrastructure to serve customers well gets the enterprise orders.
Doubao, Feishu, and Volcano Engine can only compete when twisted together.
In this email, ByteDance admitted one thing for the first time: AI is not a business line, but a layer of infrastructure.
You cannot operate infrastructure as a regular department; the entire organization must be rearranged around it.
This aligns perfectly with ByteDance's top-level strategy.
In June of this year at the Volcano Engine Momentum Conference, ByteDance CEO Liang Rubo stated that scaling the AI peak is ByteDance's most important task at present; by July, Liang Rubo issued his first all-staff letter in four years, reshaping ByteDance's organizational structure and leadership principles, emphasizing the elimination of business barriers and increasing organizational flexibility around the core strategy. (Extended reading: Liang Rubo Personally Dismantles the Old ByteDance)
Looking back today, the reshaping of organizational and leadership principles in the all-staff letter was the signal flare for this major adjustment.
In less than a month, the organization rearranged itself according to the strategy.
III. Doubao Begins to Become ByteDance's Organizational Focus
Saying this sounds simple, but consider its weight carefully.
For the past ten years, ByteDance had only one organizational focus: Douyin. Resources, talent, and commercialization all revolved around Douyin.
Today, a new product is beginning to possess the same level of strength: Doubao.
Note, this is not referring to Doubao's user or revenue growth; that's a product-level matter. The core is that the company's organizational structure is beginning to be comprehensively rearranged around Doubao.
Feishu's product team merges into Doubao; Feishu's head reports to Doubao's head; Feishu's GTM merges into Volcano Engine's To B system.
An independently operated business segment for nearly a decade was split, restructured, and absorbed by Doubao in a single day.
This is a very significant signal. A product transcends being just a product and becomes company strategy only when it becomes the organizational center.
Comparing with Douyin makes this clearer.
Before 2018, Douyin was just a short-video app under Toutiao. In 2018, Douyin exploded in the market, becoming ByteDance's strategic focus.
After years of business integration, ByteDance gradually restructured its domestic business around Douyin, incorporating e-commerce and local life into the Douyin ecosystem; the overseas market independently developed TikTok in a dual-track approach. Douyin ultimately grew from an independent product into the underlying foundation of ByteDance's domestic business.
Today, the same thing is happening to Doubao.
Especially this year when Doubao officially launched its Pro version, marking its transition from a free, entertainment-oriented conversational tool for C-end users to a productivity engine with advanced reasoning and complex task-handling capabilities.
With the foundational capabilities of the Pro version, Doubao Enterprise entering Feishu scenarios becomes logical; it's no longer just a simple chatbot, but an AI assistant deeply embedded in enterprise workflows. Natively integrated into Feishu Docs, Sheets, Meetings, Group Chats, coupled with enterprise-level data isolation, permission controls, and security auditing.
ByteDance is transforming Doubao from a C-end chatting tool into a B-end productivity foundation.
Feishu becomes the connector in this transformation.
Is this role important? Extremely important.
If Doubao were just an API, enterprises could only call the interface to get text back. But if Doubao can directly read enterprise Feishu documents, analyze meeting minutes, answer business questions in group chats, and turn knowledge bases into live assistants, that's on a completely different scale.
Feishu gives Doubao a core asset: real office scenarios.
The hardest part of enterprise AI isn't model capability; it's finding scenarios, letting AI know when and how to intervene in workflows. The office scenario data and product experience Feishu accumulated over a decade cannot be built from scratch; it can only be inherited.
This might be a better outcome than an independent IPO.
Feishu finally doesn't have to prove itself anymore; its value is completely embedded within a larger story.
IV. ByteDance Finally Thinks Like a To B Company
Many focus on Feishu merging into Doubao, but may overlook another matter: the establishment of the Creativity Services Platform.
The strategic significance of this might be greater than the product integration.
What was ByteDance's To B state in the past?
One sales team for Feishu, one sales team for Volcano Engine. The same enterprise client might face three ByteDance representatives simultaneously: one selling Feishu, one selling cloud, one selling models.
Three sales systems, three pricing logics, three types of client relationships. This isn't To B; this is the opposite of To B.
Enterprise clients never procure according to your organizational structure. They only recognize solutions; they don't care how you are divided internally, only whether one person can get things done.
Today, ByteDance merged all windows facing enterprise clients into a single outlet.
Feishu's Head of Sales Lin Chan, Feishu's Head of Strategy and Marketing Shi Zhijuan, all report to Volcano Engine's head Tan Dai.
One client, one window, one solution.
This is what doing To B looks like.
In the To B battlefield, product integration solves what to do; unified GTM solves how to sell.
In To B, how to sell is often harder than what to do. Products can iterate quickly, but rebuilding sales systems, client relationships, and delivery processes often takes years.
ByteDance chose to do it in one step.
This shows ByteDance's goal is not just optimizing To B; it's a fundamental reconstruction of the underlying logic, no longer organizing sales around products, but organizing sales around client needs.
What do clients need? AI. Feishu, Volcano Engine, Doubao are all components to fulfill this need. Components can be developed separately, but for the client, they must be sold integrated together.
This is the first time ByteDance truly thinks like a To B company.
The principle is simple, but in the past few years, almost no Chinese internet company doing To B truly figured it out. Everyone cut sales teams according to product lines, making clients adapt to the organizational structure.
ByteDance today reversed this: making the organizational structure adapt to the client.
In the AI era, what should a company organize itself around?
ByteDance's answer is: around the model, not around products.
Words from [Beyond the Layout]:
The biggest change for Feishu today is not merging into Doubao, but it proving one thing: AI has begun to reorganize a company's organizational structure.
In the past, the company decided products. Today, models begin to decide the company.
Who reports to whom, where resources are allocated, which business becomes the center, is no longer decided by experience, but begins to be decided by AI.
ByteDance is just the first, not the last.
