After the WeChat Open Class released the “Guidelines for Developers to Access the WeChat AI Ecosystem” on June 8, the four leading brands, Midea, Haier Smart Home, TCL, and Siemens Home Appliances, quickly announced the first batch of access, adding a national-level entrance to the smart home track.
For a time, judgments such as “AI reconstructs smart homes” and “whole-house intelligence ushered in new variables” filled industry public opinion.
But peeling away the shell of the propaganda rhetoric, a sharp problem is before us: the smart home capabilities supported by WeChat AI that are currently implemented have hardly jumped out of the existing functional boundaries of brand mini programs.
Will the so-called WeChat AI Agent be just a “replica of the mini program with a voice shell”?
Is this highly anticipated entrance change a true paradigm shift, or another “looking beautiful” stock migration?
Question: Start by replicating small programs?
Judging from the relevant situation, the first batch of access categories determined by these four companies are concentrated in their own advantageous categories, such as Midea air conditioners, Haier water heaters, TCL TVs, Siemens steam ovens, etc. At the same time, the brands will gradually complete the adaptation of all categories.
Why did WeChat issue guidelines on June 8 and several companies completed the adaptation on June 12?
Because the first batch of WeChat AI smart homes that are currently connected are essentially “AI voice encapsulation of mini program capabilities”, it is not wrong to say that it is a “replica of mini program”.
Because none of the functions officially announced by the four major brands currently exceeds the capabilities of the original mini programs.
In terms of technical implementation, WeChat AI access is divided into two modes.
One is automatic mode.
WeChat AI uses image recognition capabilities to click on the interface buttons of mini programs like a human being. The essence is “AI clicks on the screen for you.”
Developers can access it without modification, but they can only implement the standardized operations already in the mini program.
The second is the development model.
The manufacturer encapsulates the core interface into “skills” that can be called by AI. AI can directly adjust the interface without having to click through the page, making it more efficient and stable.
But what is called is still the original control, query, and reservation interfaces of the mini program.
When it comes to home appliances, whether it is switches, temperature adjustment, mode switching, status inquiry, or scheduled appointments, they are all mature functions that have existed in mini programs for five or six years.
What has changed is only the interaction method: from “find the applet → click on the device page → click the button” to “say a word → AI execution → return the result”.
Seeing this, you would say, isn’t this “old wine in new bottles”?
You are right, this is the norm in the industry, and it is also very reasonable.
Because for manufacturers, it is the lowest cost and least risky strategy to quickly move up the existing capacity and occupy the first batch of entrance positions.
For WeChat, the first phase only involves “understanding intentions and scheduling services” without interfering with business logic beyond the boundaries, which is also in line with its consistent style.
But if it stays at this stage, WeChat AI is really just a “voice remote control for mini programs”. Not only is it inferior to Hunyuan and Yuanbao, it is not even as efficient as directly clicking on mini programs – after all, clicking on a simple command is faster than speaking a sentence.
Breakthrough: What are the abilities that small programs cannot do…?
If this is just a transitional form of the 1.0 stage, then the real variables of the WeChat AI ecosystem should be hidden in the capabilities that are “not like mini programs”.
This part of the capability, let alone small programs, cannot be achieved by any other smart home entrance device, and it is very likely to become the most differentiated experience direction in the next 1-2 years.
Specifically, there are two capabilities that are unique and lethal advantages of WeChat.
The first, of course, is cross-brand neutral scheduling.
In the past, users who wanted to connect the whole house were either forced to buy a complete set of equipment of the same brand, or switch back and forth between multiple apps and multiple speakers.
Xiaomi, Huawei, Midea, and Haier are all naturally exclusive in their respective ecosystems, and no one is willing to import competing products.
But WeChat is neutral – it does not produce hardware, it does not even transport hardware, and it is not tied to a single brand.
Theoretically, as long as manufacturers are connected to the WeChat AI ecosystem, users can use natural language to schedule devices of different brands in the same conversation.
For example, in this case: “I will be home in five minutes” – WeChat AI starts to command, turn on the Midea air purifier, turn on the Haier air conditioner to cool down, and the Siemens oven starts to preheat… Cases like this have been fantasized by home appliance companies, speaker companies, and Internet giants for a long time, and now they finally see that WeChat seems to be able to run through it.
But just because I have been thinking about this case for a long time, it seems that although it can be realized, it is still not amazing enough.
As for this second ability, I think it has a bit of an AI ecological meaning: for example, someone in a family group says, “I’ll be home in half an hour, turn on the air conditioner,” and the AI in the group directly recognizes the command and executes it, without everyone having to install an APP and have permissions. Send a message to parents, “Help me turn on the air purifier at home.” There is no need to teach the elderly to operate the mini program, AI automatically recognizes and executes it; if a child reports an error with a device at home, parents can remotely troubleshoot and make an appointment for after-sales service on WeChat, without the need for step-by-step video guidance.
Therefore, WeChat AI still needs to make good use of WeChat’s most irreplaceable asset—the social relationship chain.
This is a natural advantage that all smart home entrances do not have.
If WeChat’s AI ecosystem can develop more such operating ideas, I think it will indeed be a great progress for smart homes.
Worry: Will it become a “tasteless upgrade”?
At present, there are three core challenges for smart homes to connect to the WeChat AI ecosystem, which determine whether WeChat AI smart homes will eventually become an industry variable or another tasteless function that is “applauded but not popular.”
The first challenge is that voice may not be more efficient than clicks.
For high-frequency instructions such as “power on, power off”, the click speed is much faster than voice input.
If WeChat AI can only cover basic operations, users will soon return to the old way of “pointing directly on the screen”, just like many users back then switched to APP touch operations after trying voice control for a while.
Only when linkage between complex scenes becomes the norm, can the efficiency advantages of AI interaction be truly reflected.
The second challenge is how much data and permissions companies can open up.
For leading home appliance companies such as Midea and Haier, user data and device data are core assets.
Each of them has its own self-developed AI brain and agent system, and no one is willing to completely open the core data to third-party platforms.
In the end, the industry is likely to form a balanced situation of “limited openness”: basic control capabilities are fully liberalized, and in-depth data and core capabilities are retained within the own system.
WeChat AI can only perform superficial command scheduling and cannot achieve true active intelligence and deep personalization.
The third challenge is, are users really willing to hand over decision-making power?
In other words, where is the sense of boundary of the AI ecosystem?
It is considerate to automatically turn on the air conditioner when the user says “it’s a bit hot”; but if the user just casually complains about the weather and the device starts on its own, it will become an intrusion.
The sense of proportion is a great test of product capabilities. Once it crosses the line, proactive service will turn into proactive harassment.
Another problem is that the star-level players in the WeChat AI ecosystem are platforms such as JD.com, Meituan, Ctrip, Didi, Tongcheng, and Dewu. They will further test the capabilities of WeChat AI, and the user experience will also change greatly. Smart home issues are not yet on the list.
Conclusion: Paradoxical, changing the norm
Back to the original question: Is the current WeChat AI Home a replica of the mini program?
Maybe at this stage, but hopefully not in the long run.
When the mini program was first born, it was widely questioned that it was “just a shell change of H5” and there was no possibility of replacing Yuansheng APP.
But what ultimately changed the industry was not how advanced the mini-program technology itself was, but how it lowered the threshold for service acquisition to a sufficiently low level.
The significance of WeChat AI to smart homes should probably be the same.
It will not transform the whole house’s intelligence overnight, nor will it immediately give birth to amazing black technology.
But it may allow “silent users” who have never used smart functions to control the device with just a casual sentence for the first time.
The current “plausibility” has always been a necessary stage for platform reform.
