Summary of Key Points
This speech outlines the core perspectives of Garry Tan, the head of YC (one of the world's leading startup incubators), on entrepreneurship in the AI era: The real benefit of AI does not lie in fine-tuning large models, but in reengineering work processes and transforming organizational structures into “skill manuals” and “enterprise brains.” This enables teams as small as 15 to 40 people to generate annual revenues in the tens of millions to hundreds of millions of dollars, completely revolutionizing the traditional business logic of relying on a large number of employees. Garry also emphasizes the need to distinguish between tasks that AI is well-suited for (such as intuitive judgment) and those that require clear rules (such as code calculations), and to transform successful practices into reusable “skills” to avoid repetitive labor, ultimately creating “AI-native companies” rather than merely using AI as a tool.
Detailed Explanation
1. The True Value of AI: Not in the Power of Models, but in How Workflows Are Connected
Many people believe that the value of AI lies in more powerful models (with more parameters), but Garry disagrees. He claims that his efficiency has increased by 400 times after adopting AI, not because he switched to a more advanced model, but because he reorganized and streamlined his work processes. For example, at YC, 95% of the code in some companies is generated by AI, and small teams (of 15 people) can generate annual revenues of 15 million dollars, while larger teams (of 40 people) can generate 60 million dollars. This is achieved by optimizing workflows and using AI to handle most repetitive or complex tasks, freeing up humans to focus on tasks that AI cannot perform.
Plain Language Translation: It’s like using a food delivery app; the value doesn’t come from the app itself, but from how it connects the processes of finding a restaurant, placing an order, and delivering the food—making it unnecessary for you to do all that work manually. The same principle applies to AI: by reorganizing tasks, we can significantly increase efficiency.
2. New Approaches to Organizational Structures: Skill Manuals as Virtual Employees
In traditional companies, many employees are hired to perform different roles (sales, finance, customer service), but in the AI era, this is no longer necessary. Garry suggests that a “skill manual” can function as a virtual employee. For instance, detailed step-by-step guides for tasks like tax filing or handling customer complaints can be created, and AI can then follow these instructions. The organizational structure can be represented by “parsers,” which allocate tasks accordingly. For example, the Emergence team invested in by YC achieved annual revenues of 15 million dollars with just 15 people because they turned sales and operational tasks into skill manuals, allowing AI to handle most of the work. Even YC’s non-programmer employees can use AI to integrate 100 Excel files into a single application, effectively acting as “AI managers.”
Plain Language Translation: Instead of hiring multiple customer service representatives, you can create skill manuals for each task, and AI can perform those tasks without any additional cost or time wasted.
3. Clearly Defining the Roles of AI
Garry emphasizes that AI has two distinct “workspaces”: one for handling ambiguous tasks (such as making aesthetic judgments or understanding hidden meanings in speech), and another for tasks with clear rules (such as writing code, calculating coordinates, storing data). It’s important not to mix these roles.
Examples: When arranging seating for 800 people, AI can be used to determine who should sit together (an ambiguous task based on intuition), while the specific calculations and data storage are handled by traditional software. If AI were responsible for calculating coordinates, errors might occur.
Plain Language Translation: Think of AI as a creative director for handling intuitive tasks; let computers handle tasks with clear rules, like math calculations, to ensure accuracy.
4. The Enterprise Brain: Storing All Knowledge
The human brain can only hold about 7 pieces of information at a time, but AI can store thousands of pages of data (e.g., three copies of “Harry Potter”). An “enterprise brain” is a system that stores all company documents (emails, meeting notes, customer conversations, decision-making processes), allowing employees to quickly retrieve the information they need.
Example: Garry’s personal AI, “G-Brain,” can immediately retrieve previous conversations with founders and similar solutions when needed, 100 times faster than humans.
Plain Language Translation: In the past, company knowledge was stored in individual employees’ minds, which would be lost when they left. Now, all information is stored in a central “enterprise brain” that everyone can access at any time.
5. Turning Success into Reusable Skills
Garry advises against doing repetitive work. For example, after using AI to create a satisfactory report, the process should be documented as a skill (e.g., saved as a template or step-by-step guide) so it can be used again without having to teach AI from scratch each time. Otherwise, you’ll end up repeating the same efforts every time.
Plain Language Translation: Just like cooking a dish like braised pork for the first time and then writing down the steps, you should use those steps next time to avoid reinventing the process. The same principle applies to AI: by documenting successful methods, companies can become more efficient and avoid unnecessary work.
In Conclusion
In the AI era, entrepreneurship is about creating processes that are more efficient and using technology to transform organizations into combinations of “virtual employees” and “enterprise brains.” This is not a future vision; it’s something that can be achieved right now.