Summary of Key Points
Founded by Kai-Fu Lee, ZeroOneAllThings initially focused on developing general large models with trillions of parameters. However, the company has shifted its focus to providing decision-making AI services tailored to the "key decision-makers" within enterprises—such as CEOs, sales leaders, and investors. This strategic change has led to a significant increase in orders: from 500 million yuan in 2025 to 1.5 billion yuan in 2026, with the goal of going public (IPO) by 2027. Kai-Fu Lee himself has also transformed from an AI evangelist who waited for customers to come to him to a frontline salesperson and product manager who visits companies in person. By understanding their needs firsthand, ZeroOneAllThings has launched products like "Boss AI" and "Sales Champion AI," which help business leaders overcome the core issue of distorted information that often deviates from the truth.
Detailed Analysis
1. **Kai-Fu Lee's Transformation: From a Tech Enthusiast to a Customer-Focused CEO**
Three years ago, Kai-Fu Lee believed that creating a good model would be like having a money-making machine (a "Token Factory"). Now, he personally visits companies around the world—during the Spring Festival, he went to Africa to explore mines and visit Zhengda chicken farms to understand egg production processes. Why? Because enterprise AI is not just about selling software; it's about understanding the real business needs of customers. For example, mines need to optimize their supply chains, and chicken farms need to improve efficiency, and these issues can only be addressed by seeing the situations on-site. He has even become the company's top salesperson, securing several potential orders by meeting with around 30 CEOs at a summit in Switzerland over two days. This change wasn't forced; he realized that customers are willing to pay for solutions to real problems, not just for advanced technology.
2. **Why Abandon General Models? Because the Parameter Race Isn't Profitable; Business Needs Are the Real Gold Mine**
ZeroOneAllThings also pursued general models with trillions of parameters at first, but Kai-Fu Lee realized that the competition in this area was too intense, with everyone focusing on the number of parameters. Without differentiation, it led to price wars and little profit. In contrast, the needs of key decision-makers in enterprises are clear and profitable. For instance, CEOs often struggle to see the true situation within their companies due to distorted information. AI can help them identify risks and allocate resources more effectively. By focusing on these real issues, the company's orders have increased by three times, with gross margins reaching up to 85%, which is much more rewarding than the costly competition in the parameter race.
3. **What Exactly Is Decision-Making AI for Key Decision-Makers? Helping Them See Through the Fog**
In simple terms, decision-making AI creates a central system for making decisions within enterprises. For example, "Boss AI" doesn't just provide a chat tool for CEOs; it first builds an "enterprise ontology"—a comprehensive model of the company's structure, processes, key indicators, and decision-making logic. Multiple AI systems (for risk management, compliance, valuation, etc.) work together to help CEOs solve critical problems. For instance, a mine owner can use this system to quickly analyze supply chain issues, and a sales manager can predict which customers are most likely to make a purchase based on historical data and market trends. This isn't just another AI tool; it's about reengineering the entire decision-making process, allowing CEOs to manage their companies without needing to know how to code.
4. **How Difficult Was the Transformation?**
The transformation was challenging, with some team members leaving, but those who stayed shared Kai-Fu Lee's vision. The team fell into three categories:
- Old partners who trusted him and followed his lead, becoming core members of the new direction;
- Professional managers who did their jobs diligently;
- Those who came for the general models and left when they realized the shift in focus.
Kai-Fu Lee compares this process to a ship encountering rough seas, where only those willing to take smaller boats can survive. Now, he recruits two types of people: those over 40 who understand business and AI (e.g., senior executives from large companies learning how to use AI for project management) and younger professionals (around 20) who are skilled in AI but not necessarily familiar with business. The management approach is more flexible, with direct reports to him rather than relying on traditional departmental structures, all aimed at quickly responding to customer needs.
5. **The Future of the Industry: Only a Few General Models Will Survive; Enterprise AI and General Models Will Coexist**
Kai-Fu Lee predicts that only a few general large models will remain (perhaps just a few leading companies in China), while others will be eliminated or used internally by specific organizations like Tesla. Enterprise AI and general models won't replace each other completely. Just as Excel can't replace SAP (enterprise management software) due to their different purposes, enterprise AI is a customized decision-making system. He's not worried about general model companies competing with them, as they lack the ability to build an "enterprise ontology"—a detailed understanding of customer needs that requires on-site engagement.
Conclusion
Kai-Fu Lee's transformation represents a shift from idealistic technology pursuits to practical business applications. His actions show that AI must be closer to customers and their real needs to be profitable. ZeroOneAllThings' success highlights the importance of addressing real problems in enterprises, rather than wasting resources on parameter-driven competitions. This is the true battlefield for AI implementation in the industry.