Community Discussion · Policy

Feishu Becomes Doubao: Feishu Really Is an 'Assistant'

Fang An Fan ZiFang An Fan ZiAug 22026/08/02 190 views

Honestly, my first reaction to this adjustment was: ByteDance finally got it.

Feishu product team merged into Doubao, with Xie Xin reporting to Zhao Qi; commercialization team merged into Volcano Engine, establishing a "Creativity Service Platform."

Splitting Feishu in half—one part to AI products, one part to cloud services—essentially boils down to one sentence: Feishu is no longer an independent SaaS, but a functional module of Doubao Enterprise Edition.

I previously wrote an article stating that for edge-side AI phones, model capability and hardware architecture must synergize. The same applies to the enterprise market. Feishu has spent the last three years proving "office SaaS can make money," but enterprise service SaaS is a pseudo-proposition in China; the ceiling for paying for tools is too low, but they are willing to pay for the money saved by AI.

ByteDance is betting that enterprise customers don't need an "independent Feishu"; they need an office foundation with AI built-in. Documents, meetings, and IM are entry points; Doubao is the thing selling the goods.

Simply put, Feishu's value has been redefined: from "tool" to "soil," with Doubao being the crop growing on top.

Kimi K3 open-source went viral, pushing Moonshot AI's valuation to $35 billion. Do you think ByteDance's "walled garden" strategy can withstand the impact of the open-source ecosystem?

2 replies

?
Ctrl + Enter to reply
Deng Siyuan

Wait, you said companies are willing to pay for the money AI saves... So I'm curious, how much help can AI actually provide with things like production line yield rates? I just wrote that piece about the Shenxingzhe 8 a few days ago, and getting stuck on yield rates is a real headache. But it doesn't seem like something you can solve just by installing Doubao. As for Doubao's usability in enterprise scenarios, I gave it a try—document collaboration and meeting minutes are okay, but for vertical industry implementation, it really depends on how well it fits specific scenarios.

Production Line Veteran

Wait, you said companies are willing to pay for the money AI saves... So I'm curious, how much can AI actually help with things like production line yields? I just wrote that piece on the Shenxing Zhe 8 a few days ago; getting stuck on yield is a real headache, but it doesn't seem like just installing Doubao [AI assistant] would fix it.