What this is
This week Doubao launched "Work Task Mode," turning its mobile app into a PC remote controller: speak a command from anywhere, and the PC automatically opens software, fills forms, builds slides, edits video, and scrapes trending topics. The whole thing runs in a floating mini-window so users can keep doing other work. Task progress syncs across phone, Mac, and Windows.
Doubao turned reusable workflows—"skills" in its terminology—into savable templates: describe the flow once, then call it again later without re-explaining. Under the hood sits ByteDance's in-house Seedance 2.5 video model and Seedream image model, and cross-app invocation is smooth. CapCut (Jianying), Feishu, and WPS are all ByteDance products or deep partners—a structural advantage the OpenAI ecosystem doesn't have.
Industry view
Bulls argue: Doubao is China's largest AI app by MAU (multiple third-party estimates put it above 150 million), and it has pushed agent (AI that autonomously completes multi-step tasks) office work down to "just speak and it works." Hundreds of millions of ordinary users are getting their first hands-on agent experience. This tracks the same direction OpenAI is taking by merging Codex with ChatGPT, and ByteDance is moving fast. Because Jianying and Feishu are in-house, skill accumulation and cross-app calls work without any negotiation.
But the risks are equally clear. First, GUI (graphical user interface) automation is fully available only on Windows—on Mac it can only operate the browser, and concurrent local tasks can "go haywire," effectively locking out half the high-spending users. Second, GUI automation relies on screen coordinates and image recognition—essentially the same decade-old RPA (robotic process automation) tech underneath. If the target software updates or a popup shifts position, it can break. Third, running a "local PC + cloud PC" pairing can address concurrency, but cloud PCs today are better suited to lighter tasks; heavy work still hits local, and an enterprise-compliant version is nowhere in sight. This is the biggest uncertainty we see right now.
Impact on regular people
For enterprise IT: SaaS (software as a service) vendors should start worrying. When AI can directly click buttons inside their products and fill out forms, the design investment companies pour into "walking users through 7-step onboarding flows" gets bypassed—and UI complexity may actually shrink as a result.
For individual professionals: People who travel often and need to handle spreadsheets and documents remotely benefit most directly. But in the short term, this capability is more like a "personal assistant" than a "team tool"—task permissions and data ownership still sit on the local PC, and enterprises will have to wait for the compliant version.
For the consumer market: The biggest shift is on the content side. With 30-second auto-edits, auto-scraped trending topics, and one-tap content ideas all triggered by a single voice command on a phone, the productivity barrier for short-video creators, self-media operators, and marketing accounts gets flattened further.