Collaborative Office Enters the Agent Era: Feishu + Doubao Work Takes the Lead

Collaborative Office Enters the Agent Era: Feishu + Doubao Work Takes the Lead

The new official pairing for AI collaborative office—I'm already on board.

By Tingyu, sent from Aofei Temple

QbitAI | WeChat Official Account QbitAI

In the AI era, how should people and Agents within enterprises collaborate? What is the smoothest experience you can imagine?

And, with collaborative office having developed to this point, what exactly should the next new stage look like?

These questions have now been answered with a pioneering demonstration from Feishu and Doubao Work.

At the recently held 2026 Feishu Future Unlimited Conference and Doubao Work Launch Conference, two major new products arrived: Feishu 8.0 and Doubao Work Partner.

One is a collaboration platform built for Agents and humans, and the other is currently the hottest enterprise-grade Agent, both backed by ByteDance.

With these three elements combined, I feel this launch event may represent a major turning point in the collaborative office space.

The core of the Feishu 8.0 upgrade is a systematic "Agent-friendly transformation" of Feishu.

Creating group chats, writing documents, operating spreadsheets, routing approvals... everything a human can do, the platform has now fully equipped Agents to do as well.

Doubao Work Partner is China's first team agent (Tag) product, benchmarked against Claude Tag.

It has an independent organizational identity, can be added to group chats, sees team discussions, attends meetings, reads documents, checks code, and proactively follows up on and drives work.

More notably, ByteDance CEO Liang Rubo expressed clear views on the direction of work and productivity in the AI era during the launch event.

Liang Rubo stated that Agents do not exist in isolation; they need to collaborate with people, work environments, and other Agents. As a collaboration platform carrying enterprise context and tools, Feishu is the critical foundation for Agents to deliver value.

He emphasized that Agents, models, and collaboration environments must be tightly integrated to truly unleash their value. Based on this judgment, ByteDance completed an organizational restructuring, integrating Doubao, Feishu, and Volcano Engine together to focus on building high-quality work and productivity products.

Feishu CEO Xie Xin stated, "After years of effort, Feishu has become the collaborative office platform loved by the most people, and it is also the best work platform for Agents."

Now, if an enterprise needs to choose a collaborative office platform, Feishu is undoubtedly the best choice. Coupled with the natively integrated enterprise-grade Agent Doubao Work, this will be the optimal choice for work in the AI era.

Feishu 8.0: Systematic "Feishu for Agent" Transformation

If your company uses Feishu, you probably know how smooth the work experience can be.

Whether it's cloud documents, Minutes, or video conferencing, every collaboration tool is handy. It must be said that Feishu has a stellar reputation for user experience.

After the explosion of Agents this year, events such as Feishu becoming the first Chinese IM platform to receive official OpenClaw support and the open-sourcing of Feishu CLI have directly shown everyone that Feishu has the highest level of Agent-friendliness, openness, and product adaptability among collaborative office products.

Enterprise clients are also increasingly realizing the importance of human-Agent collaboration within organizations. In the first half of 2026, Feishu's ARR growth rate was 2.5 times that of the same period last year, with over 90% of new customers simultaneously purchasing AI products.

Among them, over 40% of new customers considered "native Agent support capabilities" as a core factor in their selection. Leading enterprises such as SAIC Motor, Starbucks, Guming, and Air China, which have used mature collaboration systems for years, are all shifting to Feishu based on new AI standards.

It seems that everyone has realized the importance of enterprise context for AI.

The core of the newly released Feishu 8.0 is the systematic "Feishu for Agent" transformation—going forward, enterprise Agents can use Feishu's enterprise data and tools just like humans do.

The underlying logic is that all collaboration software in the past was designed for humans, but under the new working paradigm of human-AI collaboration, collaboration software also needs to be designed for Agents.

An Agent entering an organization is like a new employee onboarding. If there is no desk, no system permissions, and no one tells it how to coordinate with colleagues, hitting walls everywhere is inevitable.

Based on this, the three core upgrades of Feishu 8.0 all revolve around human-Agent collaboration:

1. Human-Agent collaboration is as natural as human-human collaboration

You can add an Agent to a group chat just like adding a colleague, and get it to work with an @mention; multiple Agents can also @ each other and collaborate on tasks without humans passing messages in between:

You can also quickly invoke an Agent with slash commands; Markdown/HTML generated by AI can be edited directly within Feishu without downloading and switching to other software:

1. Agents can freely use the full range of Feishu capabilities just like humans

Messages, documents, multi-dimensional spreadsheets, calendars, approvals... almost everything a human can do, an Agent can do.

Even! Agents use them better than humans.

Taking Feishu Whiteboard as an example, compared to simple flowcharts drawn by humans, Agents can use it to draw highly complex infographics where every element is editable:

1. Enterprise full-lifecycle security management and control for Agents

Agent permissions in Feishu are always consistent with the user; if an employee cannot see certain information, the Agent cannot access it either.

Enterprise administrators can also centrally manage Agents, trace usage, set AI usage limits, and control high-consumption models and scenarios.

There are also accompanying upgrades:

Multi-dimensional spreadsheets have been upgraded to the enterprise data foundation for the AI era; a business system can be built and data analysis performed with just a natural language sentence;

Feishu Project's AI capabilities are embedded throughout the entire workflow, with AI automatically executing at fixed nodes (requirement review, iteration retrospective).

To summarize, as Xie Xin said:

In Feishu 8.0, we hope Agents can work just like humans.

Native Integration + Team Agent: Enterprises Enter a New Stage of AI Adoption

Although the internet is now full of tutorials on connecting various Agents to Feishu, if you don't want to figure out how to use Feishu CLI to connect Codex, Claude Code, or other Agent products to invoke Feishu...

Just use Doubao Work directly.

Before expanding on this section, it is necessary to review the progress of Doubao Work.

When Doubao Work was previously released, it was the first to achieve account-level integration with Feishu—logging into a Feishu account allows direct use without additional configuration.

Now, in every corner of Feishu, you can invoke Doubao Work: while writing documents, in meetings, or processing spreadsheets, it can step in as an assistant at any time.

It can be said that after this period of running-in, Doubao Work has formed a deep integration with Feishu: sharing the account system, deeply embedded in core scenarios such as documents, meetings, and spreadsheets, and naturally inheriting Feishu's organizational permissions and security capabilities.

At this conference, it was also officially clarified that Doubao Work and Feishu are natively deeply adapted at the underlying level.

What is native adaptation? It means Doubao Work is naturally familiar with the information distribution of the entire Feishu workspace; logging into a Feishu account allows direct use without additional installation or debugging.

It brings three major advantages: a smoother experience, with outputs remaining directly in Feishu for circulation; higher work efficiency, with capabilities like Feishu knowledge Q&A integrated into Doubao Work to complete the same tasks better; and greater peace of mind on security, natively equipped with compliance capabilities such as permission control and watermark encryption.

In contrast, even if external Agents connect via Feishu CLI, it is difficult to achieve this level of underlying integration.

Now, they have taken a step further by launching Doubao Work Partner.

Speaking of which, some might be confused: what exactly is the relationship between Doubao, Doubao Work, and Doubao Work Partner?

Having carefully sorted it out, the differences are:

Doubao is a general AI assistant for individuals, while Doubao Work targets productivity scenarios but retains its personal assistant nature.

Doubao Work Partner, however, is a team agent (Tag) product, expanding the service target of AI from individuals to entire teams.

In essence, it is the same thing as Claude Tag, which Anthropic first launched in the instant messaging tool Slack in June of this year.

The core capabilities of Doubao Work Partner mainly include the following four points:

1. Joining group chats, receiving tasks from multiple people, and processing in parallel

Product can ask it to organize requirements, R&D can ask it to check code, and operations can ask it to draft weekly reports—it can do them all simultaneously;

And there's no need for each person to open a separate chat window and explain the background from scratch; it remembers everything said in the group.

It can also automatically compile a weekly report from over a dozen group chats and meeting minutes; after multiple groups discuss a plan, it summarizes it itself, breaks out action items, and @mentions the corresponding owners.

1. Entering documents and meetings, bridging across groups

If a question comes up during a meeting that requires confirmation from a product team in another group, just tell it, and it will go ask on its own and bring the conclusion back;

Information in documents, discussions in group chats, and conclusions from meetings can all be strung together and utilized, without manually copying and pasting as a message relay.

1. Online 24/7, proactively working

If someone reports a bug in the group in the early hours of the weekend, before R&D is even online, it reads the code repository itself, locates the problem, and provides a fix direction; by the time people wake up, the root cause analysis is already in the group.

1. Retaining team experience, manageable and controllable for the enterprise

If one colleague teaches it a new skill, everyone else can use it; when an employee leaves, the accumulated practices remain. At the same time, it has independent permissions, making it manageable and controllable for the enterprise.

Seeing this, you can probably understand that Doubao Work is no longer the previous Doubao.

Its native deep integration with Feishu allows it to step into real work scenarios and deliver higher productivity value.

Security, Security, and Most Importantly, Security

Another biggest keyword for this launch event is: security.

Recent frequent model security incidents in the industry have pushed this issue in front of all enterprises.

As employees begin to use AI extensively at work, a very realistic concern is: when individuals use AI tools externally, after company materials are uploaded to model companies, will the data be used for training? Will it leak out in other ways? The enterprise has absolutely no control.

For knowledge-intensive enterprises, core business data, customer information, and R&D results could all be leaked without anyone noticing.

This is why, in enterprise AI selection, security is shifting from a "bonus" to a "must-have."

Compared to the uncontrollability of personal AI usage, enterprise-grade procurement of Agent products is fundamentally different in terms of security. The security layout of Feishu + Doubao Work can be viewed from the following layers:

The first layer is the model. ByteDance insists on independent R&D in model training, holding the bottom line on data security, ensuring from the source that enterprise data will not be accidentally used for model training.

The second layer is Agent management and control. Feishu for Agent provides complete enterprise-grade management capabilities:

Enterprise administrators can centrally manage Agents, trace usage, set AI usage limits, and control high-consumption models and scenarios.

The third layer is native adaptation.

As mentioned above, Doubao Work and Feishu are natively deeply integrated at the underlying level, naturally inheriting Feishu's mature organizational permissions, watermark encryption, and other compliance capabilities.

Content you are not authorized to view, the AI cannot see either, without the need for additional complex security rule configurations.

Furthermore, regarding the newly released team agent.

Doubao Work Partner has independent permissions, not shared with anyone, and only reads and operates within the scope of authorization; the management backend clearly shows AI expenditure, configures its capability boundaries, tools, and permissions.

The deeper Agents penetrate enterprise workflows, the more important security becomes. The security layout of Feishu + Doubao Work in the enterprise space is precisely one of its core values compared to personal AI tools.

One More Thing

The new stage of collaborative office has truly arrived.

Looking at history, the evolution of collaborative office has essentially been answering the same question: how to make information and work flow more smoothly within an organization?

From the paperless standalone office era, where Word and Excel freed office work from pen and paper, to email and Lotus Notes first connecting scattered individuals into the same network.

OA systems brought approvals, attendance, and reporting.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.