WeChat Tests AI Social Feature: Two "Xiaowei" Agents Can Communicate on Behalf of Users
nashnova research
WeChat is testing 'Xiaowei AI Social', a feature that lets two users' AI assistants talk directly to each other — coordinating meals, schedules, and errands — with humans stepping in only for key decisions. Tencent's president frames this as the starting point of WeChat's evolution into an AI-first ecosystem.
Two AIs chatting for you — how does it actually work?
WeChat's AI assistant Xiaowei (小微) can now talk directly to a friend's Xiaowei, handling dinner plans, scheduling, and other social coordination — no real-time user involvement required.
This is fundamentally different from "AI ghostwriting." Ghostwriting sends messages in your chat window; Xiaowei AI Social is a conversation between two AIs in a separate Xiaowei interface — nothing appears in your personal chat.
This means → WeChat is testing a new interaction layer: humans stay out of the conversation itself and are only prompted when a decision needs sign-off.
What can Xiaowei do, and what powers it?
Xiaowei's capabilities span three tiers: controlling native WeChat functions (messaging, audio/video calls, Moments search); triggering mini-programs for booking appointments, ordering coffee, reserving hotels, and hailing rides; and searching, generating images, and creating notes.
The core model is WeLM, built in-house by Tencent's WeChat team, with some responses routed through DeepSeek. WeLM supports a 128k-token context window — roughly 90,000–100,000 Chinese characters — and is optimized for logical reasoning and math.
In plain terms = Xiaowei is not just a chatbot. It can get things done inside WeChat — from sending a message to calling a cab, all triggered by a single command.
Jiemian News reporters who tested it called Xiaowei the biggest update in WeChat's history in terms of both interaction quality and feature breadth.
How does Tencent's leadership frame this?
At Tencent's Q2 2026 earnings call on August 12, president Martin Lau (刘炽平) said AI will transform WeChat from a communication tool into an "AI-first ecosystem" — users issue commands, the system executes.
He compared the trajectory to how WeChat delivered ten times the value of QQ in the mobile-internet era.
This means → Tencent internally treats Xiaowei not as a feature update but as a platform-level architectural shift — moving WeChat from "people chatting with people" to "people directing AI, AIs collaborating with each other."
Agent-to-agent commerce — can it scale?
Lau outlined a longer-term vision: every mini-program, merchant, and user will eventually have a dedicated Agent, and Agents will communicate with each other and complete transactions autonomously. Tencent is building the underlying infrastructure.
He said, "As we roll out prototypes, we are increasingly confident in this direction."
This reflects Tencent's ultimate ambition beyond AI assistants — an Agent-driven commercial network. Whether the agent-to-agent transaction architecture can scale will be the critical test of WeChat's AI ecosystem monetization capability.
市场有风险，内容仅供研究参考，不构成投资建议。