Microsoft shows off how its social chat bot Xiaoice, which is in use across platforms in China, can place phone calls to humans, similar to Google's Duplex
Similar to Google Duplex, but only in China — Google demonstrated a jaw-dropping new capability in Google Assistant earlier this month …
Context & Ripple Effects
Xiaoice began as a personality play: when Microsoft introduced the chatbot in 2015 it had 20 million registered users on Chinese platforms, valued for humor and listening rather than task execution. Three years later, Microsoft is showing the same bot doing what Google demoed at I/O this month — placing voice calls to humans — but on home turf where Google Assistant doesn't operate.
The timing is pointed. Google's Duplex demo drew immediate pushback for sounding too human, and Microsoft is answering with a system that has three years of accumulated conversation history behind it. The related coverage also shows Alibaba claiming a conversational AI already handling millions of package-delivery queries a day, so voice-based task agents are becoming a contested category in China specifically.
First-order effects
- Microsoft turns Xiaoice from a text-based social companion into a voice agent that places calls on users' behalf, directly matching the headline Duplex capability while operating across Chinese platforms Google can't reach.
- Google's uncanny-valley backlash sets the disclosure standard Microsoft now has to navigate: any human-sounding bot placing calls invites the same scrutiny Duplex received.
Second-order effects
- Alibaba's claim of millions of daily customer-service conversations means Microsoft isn't racing a demo but an in-production rival — pushing both toward scale deployment rather than staged showcases.
- Chinese platform partners gain leverage: whichever bot they host gets the call-placing traffic, making Xiaoice's multi-platform distribution a distribution moat against single-platform competitors.
Third-order effects
- If the pattern holds, the competitive frontier for voice AI shifts from who demos best to who accumulates real call volume — favoring players embedded in high-frequency messaging apps over those attached to phone ecosystems.
- Human-sounding call placement is converging on a disclosure problem regulators will eventually have to settle, with China's draft AI rules already signaling that content and conduct boundaries are coming.
The trend: Assistants are graduating from text chatbots to full-duplex voice agents that place real calls, with China emerging as the deployment battleground ahead of Western markets.