Inside WeChat's new AI agent: a hands-free experiment in China's super-app

For anyone who has ever watched a Chinese friend settle a dinner bill, book a taxi and buy movie tickets without once leaving a single app, WeChat has long seemed less like software and more like a nervous system for daily life. Now Tencent is trying to give that nervous system a brain of its own. The company has quietly begun embedding an artificial intelligence agent called Xiaowei into the platform, and for a brief window I got to let it run my digital routine for nearly 24 hours. The result is a fascinating, sometimes maddening glimpse into where the world's most important super-app is headed.

Tencent first flagged Xiaowei in its latest earnings announcement on Wednesday, framing it as a privacy-conscious, inference-efficient addition to the ecosystem. Visually, it appears as two green dots at the top of the main WeChat screen. But unlike the chatbots that have become ubiquitous across tech platforms, Xiaowei is not designed merely to answer questions. It is built to act. Where a conventional assistant might suggest a hotpot restaurant, Xiaowei is meant to book the table and place the order. That distinction between recommending and doing is the entire point of the new wave of AI agents, and Tencent is betting that WeChat's scale gives it an unmatched testing ground.

For the uninitiated, it is hard to overstate what WeChat means in China. More than a billion people use it not just to chat but to pay merchants, split bills, hail rides, book hospitals, file taxes and run small businesses. It is the digital infrastructure of a society. That is precisely why Xiaowei matters. An agent embedded inside WeChat is not a novelty feature; it is a potential operating layer over the daily transactions of hundreds of millions of people. Tencent has spent years turning WeChat into the default gateway for Chinese digital life, and now it is trying to make that gateway proactive rather than merely responsive.

My own test was revealing. There were moments when Xiaowei felt genuinely ahead of its time, executing multi-step tasks with a fluency that suggested the future has arrived. But there were also stumbles, moments where the agent's confidence exceeded its competence. That tension is worth noting for anyone tracking capital flows in Asia. WeChat is not just a consumer product; it is a platform on which entire business models depend. If Tencent can make Xiaowei reliable, it could deepen the moat around its ecosystem further, making it even harder for rivals to peel away users. If it stumbles, the embarrassment will be public and the stakes enormous.

The broader signal is about where Chinese tech is heading. After years of regulatory crackdowns and cautious spending, the country's largest platforms are again pushing the boundaries of what their products can do, this time with AI at the core. Tencent's decision to highlight Xiaowei in an earnings report, rather than at a flashy launch event, suggests a deliberate strategy: prove the technology works before promising the world. That is a characteristically Tencent approach, incremental and ecosystem-first, but the ambition is unmistakable.

What happens next will depend on execution. AI agents are only as good as their ability to understand context, respect privacy and recover from errors without causing chaos. In a super-app like WeChat, the margin for error is thin. A bot that books the wrong flight or sends money to the wrong contact is not a minor inconvenience; it is a breach of trust in the most trusted digital environment in China. Tencent knows this, which is why it is moving carefully. For those of us watching from the outside, the next few quarters will reveal whether Xiaowei becomes a quiet revolution or a cautionary tale. Either way, the experiment has begun, and it is happening inside the app that already runs China.