EXIO Research Institute | June 24, 2026
On June 20, 2026, WeChat revealed an AI trump card to a portion of its user base — the native AI assistant “Xiao Wei” began a limited rollout. Users who received the internal test access found a green robot-eye-shaped icon in the upper left corner of the main WeChat interface. Clicking on it entered a dialogue interface marked “Beta,” supporting both text and voice interaction[1].
You might think this is just another chatbot. But if we zoom out, the significance of this event is far greater than it appears on the surface—it may be the most crucial piece of the puzzle in the grand narrative of the internet transitioning from “universal internet access” to “all assets on chain.”
I. The Four Tiers of “Xiao Wei”: Not Just Chat, But “What You Say Is What You Get”
WeChat's team has positioned Xiao Wei with restraint: “Users can converse with Xiao Wei via text or voice, leveraging AI technology to better use WeChat, communicate with friends, or activate Mini Programs to provide better services.”[2] However, the capability system revealed by user feedback from the limited rollout hints at a clear ambition.
From a technical standpoint, Xiao Wei employs WeLM (the WeChat team's self-developed Chinese large language model) as its primary model, calls upon DeepSeek as a fallback for complex reasoning scenarios, and is also testing third-party models like Zhipu and Alibaba to maintain technical flexibility[3]. This pragmatic “hybrid model” approach itself speaks volumes—WeChat's goal has never been to show off technical prowess, but to make AI truly functional.
In terms of practical experience, Xiao Wei demonstrates four progressive tiers[4]. The first tier is native “What You Say Is What You Get” control—you can say, “Send a message to Old Zhang saying tomorrow's meeting is moved to 3 PM,” and Xiao Wei can accurately recognize the social relationship chain and complete the operation. The second tier is ecosystem invocation, automatically launching Mini Programs to complete full-process life services like hospital appointments, shopping, and ride-hailing. The third tier is content comprehension; send it a PDF, and it generates a structured summary within seconds. The fourth tier is the most impressive—generating small tools using natural language. “Help me make a running check-in Mini Program,” and it can produce a prototype in seconds.
This is no longer just a “smarter Siri.” It is the world's largest super-app by user volume, for the first time truly embedding an AI Agent into the ordinary person's daily scenarios of communication, payment, social interaction, and work.
II. The “D-Day” for AI Agents: From Geek Toy to Utility
Why is the limited rollout of “Xiao Wei” a landmark event? Because before this, for most Chinese people, AI Agents remained a concept in the news—ChatGPT required VPN access, Manus needed an invitation code, and various domestic Agents mostly stayed at the demo stage. WeChat did one thing: it lowered the barrier to entry for AI Agents to zero.
1.4 billion monthly active users don't need to download a new app, register a new account, or learn a new interface. They just need to open WeChat as usual and speak to a “little green eye” to experience the magic of “generate a Mini Program with one sentence”[5]. The explosive potential of this “seamless access” reminds us of how the App Store in 2008 made it easy for everyone to download apps, or how WeChat Mini Programs in 2017 made the “touch-and-go” service revolution possible.
From a broader perspective, the global AI Agent market is on the eve of an explosion. According to Precedence Research data, the global AI Agent market size in 2025 is approximately $7.92 billion, projected to soar to $294.66 billion by 2035, with a compound annual growth rate (CAGR) of 43.57%[6]. Among role segments, “Productivity & Personal Assistant” type Agents lead all categories with a CAGR of 29.50%. And WeChat's Xiao Wei is precisely the one that first cracked open the hardest nut in this high-growth sector—the mass consumer market.
III. From “People Seeking Functions” to “Functions Finding People”: A Silent Productivity Revolution
Let's imagine daily life after “Xiao Wei” matures.
You wake up and say to your phone, “Xiao Wei, see if there are any important messages today, then order me an Americano to the office.” It will not only summarize the key points of your overnight unread WeChat messages but also open the Starbucks Mini Program to complete the order. At the office, you throw a 30-page research report PDF at it, and three minutes later, you receive a structured summary with key highlights. During your lunch break, you suddenly think, “I want to take my parents to Yunnan next month, budget 10,000 RMB, help me plan it.” Xiao Wei will generate a travel itinerary, hotel comparisons within minutes, and even create a “Family Trip” group chat and post the guide inside. In an afternoon meeting, you tell it, “Create a running check-in Mini Program, run three times a week, three kilometers each time.” It actually generates a usable Mini Program prototype for you in seconds[7].
In this scenario, you no longer need to remember which function is in which menu of which app. You speak, and things get done. This isn't a simple “voice assistant upgrade”; it's a fundamental shift in the interaction paradigm—from GUI (Graphical User Interface) to LUI (Language User Interface), from “people seeking functions” to “functions finding people.”
A research report from Tianfeng Securities proposes a core formula: AITAM = Total Salary × Replacement Rate × AI Efficiency Optimization[8]. They estimate that the total target market for AI Agents in China alone is as high as 3.61 trillion yuan. When AI Agents evolve from “white-collar assistants” to “versatile managers by everyone's side,” they replace not only repetitive labor but also the friction cost in human interaction with the digital world.
IV. From “Universal Internet Access” to “All Assets on Chain”: An Invisible Undercurrent
The development history of the Chinese internet can be summarized with three keywords: connection, services, and assetization.
From the full functional connection to the international internet in 1994 to today, the number of Chinese netizens has exceeded 1.1 billion, with an internet penetration rate over 78%. “Universal internet access” was largely achieved in thirty years. Building on connection, WeChat, through Mini Programs and Official Accounts, accomplished “services going online”—government affairs, healthcare, finance, education, almost all life services have been moved online. Today, we stand at the threshold of the third stage: assets on chain.
The term “all assets on chain” often first brings to mind RWA (Real World Assets, tokenization of real-world assets)—moving traditional assets like real estate, bonds, and gold onto the blockchain. This understanding is correct but too narrow. The truly broad sense of “all assets on chain” should include all value that can be verified, quantified, and circulated:
Your social influence is an asset. The ten-year accumulation of Moments, private domain traffic, and social relationship chains within WeChat, once just “data,” may in the future be tokenized through SocialFi mechanisms, becoming the basis for your earnings[9].
Your attention is an asset. When you “scroll through videos” on a short-video platform, the platform sells your attention to advertisers via algorithms. In the Web3 logic, this attention rightfully belongs to you and could be captured and priced through token mechanisms[10]. Additionally, your personal credit could be an asset. For a long time, credit records have been confined within data silos like the central bank's credit system and internet giants. When AI Agents become everyone's “digital avatar,” the behavioral data, fulfillment records, and social reputation they accumulate could form a set of on-chain personal credit systems, becoming your “credit collateral” in the DeFi world.
Does this sound distant? It's closer than you think. 21Shares, in its 2026 crypto market outlook report, predicts that the total value locked (TVL) in tokenized real-world assets (RWA) will increase from $35 billion in 2025 to over $500 billion in 2026[11]. Roland Berger is even more optimistic, forecasting that the global tokenized asset market size could reach nearly $11 trillion by 2030[12]. From real estate to cultural IP, from airline miles to luxury diamonds, all assets are being reimagined.
V. When AI Agent Meets RWA: The Future Paradigm of Wealth Management
Let's superimpose these two trends: on one hand, AI Agents are taking over the entry point for human interaction with the digital world; on the other hand, global assets are moving on-chain at an unprecedented rate. What happens when these two meet?
Imagine this scenario: Your AI Agent knows you have some idle funds maturing next month. It proactively tells you, “Based on your risk preference and recent market conditions, I suggest allocating 30% to tokenized U.S. Treasury bonds (current annualized 4.2%), 20% to gold tokens to hedge against inflation, and leaving the remaining 50% in a liquidity pool for随时 use. Do you need me to execute this for you?” You nod, and three seconds later, the allocation is complete—all settled on-chain, all compliant and transparent, all backed by underlying assets.
This is not science fiction. This “AI Agent + RWA” wealth management paradigm is moving from concept to implementation. In the B2B sector, enterprise-level Agents can already automate high-workload tasks like invoice verification, SOC alert triage, and data entry, reducing manual workload by over 60%[13]. In the C2C sector, AI financial apps like Albert already offer features like “auto-save,” “tax planning,” and “financial freedom roadmap”[14]. When these capabilities combine with on-chain assets, personal wealth management will enter a completely new dimension—your Agent not only helps manage your accounts but also helps you search for, allocate, and rebalance optimal asset portfolios globally.
VI. The Next Future for AI Agents?
As a research institute long focused on the convergence of Web3 and traditional finance, EXIO Research Institute believes the limited rollout of “Xiao Wei” is a signal: AI Agents are moving from “proof of concept” to the “user education” stage. WeChat, with its 1.4 billion users, has helped the entire industry complete the most difficult cold start—enabling ordinary people to genuinely use an AI Agent for the first time, seamlessly, at zero cost, and with zero barrier to entry. Once this step is taken, there's no turning back.
And the pace of asset tokenization is also accelerating. In Hong Kong, the EXIO Group is accelerating its efforts in this field—including building a full-chain service system from RWA tokenization issuance (RWA Issuance) to secondary market liquidity management, to compliant infrastructure (RWA Launchpad)[15], laying the groundwork in advance for the future of all assets on chain and even automated trading. The underlying logic of these initiatives converges with the vision of “Xiao Wei”: lower the barriers, enable more people to participate. WeChat is lowering the barrier to using AI, while more startups are attempting to lower the investment barrier to quality assets.
Conclusion: What We Are Experiencing Is More Than Just a Technology Upgrade
In 1994, China connected to the internet via a 64K international dedicated line. Back then, few imagined that this thin line would grow into a digital world that 1.4 billion people cannot live without thirty years later. In 2026, that green little eye icon in WeChat may be a similar historical node. It marks AI Agents transitioning from “geek toys” to “utilities,” and also foreshadows the “all assets on chain” wave moving from niche to mainstream.
From universal internet access, to all services online, to all assets on chain—each step is not an isolated technology upgrade but a redefinition of the relationship between humanity and the digital world. And AI Agents are precisely the golden thread connecting all of this.
When 1.4 billion people begin to get used to “speaking” to operate everything, when every command may reach some asset on-chain, when everyone's attention, credit, and social influence are endowed with financial attributes—what we are ushering in is not just a more convenient WeChat, but possibly an entirely new wealth paradigm.
