Yesterday (June 2), the UK's Financial Times reported: Tencent is testing an AI agent inside WeChat, and the project has been listed as the company's top strategic priority. Swipe right on the WeChat main screen to bring up an AI chat window, which automatically calls on mini-programs to help you complete various tasks — ordering coffee, hailing a ride, checking orders.
On the same day, the Wuhan Municipal Government announced at a press conference: strive to have the software industry scale exceed 800 billion yuan during the 15th Five-Year Plan period, cultivate 150 outstanding industry-specific AI agent products in 2026, and guide 100 software companies to transform into AI application service providers.
Viewed together, these two things point to the same trend: AI agents are moving from internal enterprise tools to external service entry points. For businesses, this means the way you interact with customers may soon change.
WeChat AI Assistant: The "Super Entrance" for 1.4 Billion Users
According to reports by the Financial Times and cross-verification of multiple sources, several core pieces of information about WeChat's AI agent have become relatively clear:
The essence of this product is not 'adding a ChatGPT into WeChat,' but turning AI agents into an intelligent execution layer that connects the entire WeChat ecosystem. In the past, users actively searched for services—opening mini-programs, finding stores, and placing orders. With AI agents, users only need to say: 'Order me an Americano, less ice, within 30 yuan,' and the agent will automatically invoke the coffee mini-program to complete the selection and ordering process.
What is the direct impact on businesses? In the WeChat ecosystem, carefully designed Official Account menus, mini-program homepages, and customer service scripts may all have to make way for AI agents in the future. Users will no longer need to “browse” your mini-program; instead, they will directly “use” your services through AI agents.
Think about it from another angle: if WeChat agents truly launch on a large scale, the operational logic of enterprises within the WeChat ecosystem will need to be rewritten. SEO becomes "getting agents to recommend your services first," and conversion rate becomes "getting agents to accurately call your API."
Computing power is the only bottleneck
The FT report mentions a practical issue: large-scale full deployment faces the challenge of insufficient computing power supply. WeChat has 1.4 billion monthly active users; even if only 10% of users use an AI agent once a day, that would be 140 million calls per day. Given the current cost of large-model inference, this is no small number.
But this precisely illustrates one thing: Tencent's investment in computing infrastructure is still in a catch-up phase. It was also disclosed yesterday that the official release date of the WeChat AI assistant has not yet been determined—before the computing power issue is resolved, the product can only be launched in a small-scale gray release.
For enterprises, this is actually a window of opportunity. Before AI agents become the default entry point in WeChat, there is about 6 to 12 months to adjust service interfaces, optimize the API call chains of mini programs, and test user experience in AI agent scenarios.
Wuhan's AI Ambition: 150 Intelligent Agent Products + “Super Software Factory”
Let's also look at Wuhan's policy. At a press conference on June 2, the Wuhan Municipal Bureau of Economy and Information Technology outlined three directions:
AI cultivation for software products: Cultivate 150 industry-specific agent products, build 10 vertical large models, and guide 100 software enterprises to transform into AI service providers. AI development in software development: Establish more than 10 "super software factories", achieve full coverage of AI-assisted development among 1,000 above-designated-size enterprises, and incubate 100 AI testing tools. AI in enterprise management: In the city's software industry operations and management, the usage rate of agents exceeds 80%, covering intelligent office, approval, operations, and other scenarios. Industry goal: During the "15th Five-Year Plan" period, the scale of the software industry will exceed 800 billion yuan, and it has currently reached the 400 billion yuan mark.
There are several figures worth noting: 140,000 software companies, 150 agent products, and an 80% usage rate for management agents. This is fully consistent with the direction of MIIT's 'AI + Software' special initiative, but the granularity of implementation has been refined to specific product and company counts.
For enterprise digital service providers, the signal sent by the Wuhan plan is clear: the government is willing to invest real money to promote the implementation of AI applications. If you are working on AI upgrades for ERP/CRM/OA, cities with supporting policies like Wuhan are worth close attention.
Two clues converge: the entrance has changed, and the base is also changing
WeChat AI agents and Wuhan's AI policy may seem like two independent matters, but they point in the same direction: the interaction entry point for enterprise services is being rewritten by AI agents.
WeChat represents a consumer-facing service portal—1.4 billion users frequently use WeChat in their daily lives. If AI agents become the default interaction method for WeChat, then all businesses operating within the WeChat ecosystem will need to adjust their service reach logic.
Wuhan represents a digital foundation oriented toward enterprises’ internal operations—with policy driving the AI transformation of software products, development processes, and enterprise management, AI evolves from “nice-to-have” to “essential capability.”
When both sides exert force simultaneously, the challenge facing the enterprise is twofold:
Internally, your ERP, OA, and financial systems need to integrate AI agent capabilities. Yesterday at Build 2026, Microsoft released the full-stack solution Scout + Microsoft IQ, while domestic vendors (Kingdee Lingji, Yonyou YonClaw) are also advancing in parallel.
Externally, your customers may soon use your services through AI agents rather than by “opening your application.” Is your API AI-friendly? Can your services be accurately understood and invoked by agents?
What should companies do now
Based on these two major trends, three recommendations are proposed for enterprise digital leaders:
First, evaluate whether your WeChat ecosystem services are "callable by AI." Check whether your mini program's core functions have standard API interfaces, and whether the service descriptions are sufficiently structured. If WeChat AI agents need to first "understand" your services before they can call them on behalf of users, then your service metadata must be prepared in advance.
Second, focus on cities with AI support policies, such as Wuhan.If you are using AI to upgrade software products, the Fourth Software Innovation and Development Conference held in Wuhan from June 10 to 11 (more than 800 attendees, including CASIC, China State Shipbuilding Corporation, and Moore Threads) is worth paying attention to. The combination of policy funding and industrial clusters can significantly reduce the cost of implementing AI.
Third, do not separate internal and external AI adoption. Internal agents (management, approval, operations) and external agents (customer service, order processing) must ultimately be integrated. Planning a unified agent governance framework from now on will be far more cost-effective than waiting for two separate systems to take shape and then integrating them.
/Source: June 2 Financial Times (FT) report, Wuhan Municipal People's Government press conference, TMDM, AIBase, Sina Finance
Disclaimer: This article is compiled from publicly available information and is for reference only. It does not constitute any investment or procurement advice.
