Summary of Key Points
WeChat’s internal testing of its AI-powered assistant, “Xiaowei,” has sparked discussions in the AI startup community about the potential threat it poses to existing entry points for these businesses. However, entrepreneurs and investors believe that the entrance itself is not the entire focus of AI products. The real danger lies in the enhanced capabilities of model companies like OpenAI and Google, which could overshadow superficial applications. Independent AI products have a chance of succeeding by focusing on “unconventional scenarios,” leveraging specialized expertise, and providing clear, tangible results to users. When collaborating with large corporations, it’s crucial to maintain control over the core aspects of one’s business. The logic of the AI business is shifting from focusing on user traffic (Daily Active Users, DAU) to emphasizing the value of each individual task completed (represented by the “Token premium”).
1. WeChat’s Agent Is Not as Destructive as Many Think
Many worry that WeChat’s use of its user base to drive AI startups might have a negative impact. However, the experts argue:
- Entrance Does Not Equal Retention: WeChat’s high frequency of usage may be advantageous for short-term engagement with mini-programs (which require fewer resources and are easier to adopt), but it may not attract users who are loyal to dedicated, independent apps. For example, Tencent’s previous AI product, “Yuanbao,” had an entrance but lacked a clear purpose, leading to low user retention.
- Privacy Is a Major Concern: WeChat’s social context (chat records and relationships) is a significant advantage, but users are wary of their data being used by AI. The risk of a privacy breach is much greater than the potential failure of an AI assistant.
- Charging for Everyday Services Is Challenging: WeChat is primarily a social and lifestyle tool, not a productivity platform. Making users pay regularly for services like ordering food delivery is difficult in the Chinese market.
In conclusion, while WeChat can become one of the entry points for AI products, it will not replace all specialized applications.
2. The Real Threat Comes from Model Companies
The biggest threat to startups comes from the improved capabilities of model companies like OpenAI and Google. As these models evolve, their capabilities expand, making existing superficial applications obsolete. For example, ListenHub emerged a week after ByteDance’s similar service, but it survived because it addressed a specific problem rather than merely replicating its features.
The key question is: Are you creating a “real product” or just a temporary interface for an AI model?
3. How Can Independent AI Products Survive?
Experts agree that the opportunity lies in three areas where independent products can stand out:
- Unconventional Scenarios: Areas that large companies overlook, such as 3D reconstruction services for family reunions or graduation ceremonies. These may not be seen as profitable by big firms, but they meet real user needs and can attract a significant audience.
- Specialized Expertise: Transforming industry expertise into scalable systems. For instance, Mizzen uses AI to conduct user research in a way that goes beyond simple question-and-answer interactions; the underlying methodology and quality standards are unique and difficult for large models to replicate.
- Clear Results: Providing tangible outcomes that solve real problems for users. For example, Genspark converts PPTs and recordings into usable formats, providing immediate value to users.
4. Collaborating with Big Companies Is Possible, but Don’t Give Up These Three Key Elements
AI startups can cooperate with large companies, but they must protect their independence:
- Model Selection: Don’t rely on a single model (e.g., GPT) as its capabilities change rapidly. Locking in to one model could slow down development.
- User Relationships: Keep control of user data and interaction records.
- Data Accumulation: Build a competitive advantage by collecting valuable internal company data and user history, which models cannot access.
If you give up these elements, you’ll essentially become a contractor for large companies, losing your core value as an independent developer.
5. The Value of AI Lies in Task Performance, Not Just Traffic
In the traditional internet world, metrics like DAU are important, but in the AI realm, efficiency is key:
- A good AI tool should provide immediate results and encourage quick user turnover.
- Two indicators of commercial success are:
- Context Collection: The ability to gather unique user data (e.g., company documents or personal work processes).
- Token Premium: If your AI can solve complex problems, you can charge more for its services.
ListenHub, with a small daily user base, relies on word-of-mouth and provides valuable results, making it more profitable than products with high DAU but lower revenue.
Conclusion
WeChat’s AI assistant is worth watching, but the real challenge for AI startups is to create products that are truly irreplaceable. The ultimate question is: “For whom can you provide a specific service, and to what extent?” With this answer, you’ll be better prepared for changes in entry points and advancements in AI models.