Garry Tan on the Secret to Explosive Scale with 'Small Teams x AI'
In recent years, the evolution of generative AI has dramatically changed the growth dynamics of startups. Garry Tan, President and CEO of Y Combinator (YC), discussed the new opportunities brought by AI and the mindset founders need to achieve success in an interview for Frameworks for Growth. This article summarizes the key points of that interview, incorporating quotes and specific examples, and explains them in a way that is accessible even to professional readers.
1. Startup Growth Dynamics in the AI Era
1-1. Examples of Phenomenal Rapid Growth
Garry Tan states, "We are seeing a succession of examples where entrepreneurs in recent YC batches launch with teams of fewer than 10 people and reach $10 million in annual revenue within just 12 months." This rapid growth rate is a phenomenon not seen in about 10 years, since 2009 when Airbnb achieved "10% weekly growth." According to Tan, the recent YC batches as a whole are "growing revenue at an average pace of 10% per week." The background to this is that by utilizing AI as a development resource, market validation and product improvements that previously required dozens of personnel can now be realized instantly by small teams.
1-2. The Rise of 'Vibe Coding'
While many traditional startups failed because they "could not build a product due to a lack of technology," today, a method called "vibe coding" is spreading, and technical barriers have almost ceased to exist. Tan notes, "Nowadays, technology is rarely the problem; what is important is getting inside the customer's head and understanding what they want." In fact, YC interviews report that about a quarter of startups are using AI tools to drive development through "vibe coding."
2. Factors for Startup Failure and Solutions
2-1. The Importance of 'Make Something People Want'
In the interview, Tan repeatedly emphasized the mantra traditionally passed down from YC founder Paul Graham: "Make Something People Want." He points out that the most common reason startups fail is that they "build things customers don't want," and even if they can release quickly with the power of AI, they will not be accepted by the market if they ignore customer needs. He provided a concrete image, stating that "a good product has such a strong gravitational pull that customers say they 'cannot live without it,'" and added that "the essence is not in building servers or overcoming technical hurdles, but in getting inside the customer's head."
2-2. A Real-World Pivot: Data Curve
As a specific example, the case of "Data Curve," which was originally planned to handle general AI workflows, was introduced. Before joining YC, the company was developing general-purpose AI software without a clear point of differentiation. After realizing the founding members lacked experience as software PMs (product managers) and re-examining market needs and their own backgrounds, they pivoted to "AI-powered PM support software." As a result, they reported that "revenue reached eight figures in just nine months." This case demonstrates the lesson that "the founder's own background and experience are the most important sources of insight."
3. Changes in Corporate Growth and Scaling
3-1. The End of Blitzscaling
Tan reflected that "unicorn companies like Airbnb and DoorDash in the past had teams of thousands and invested a lot of capital to achieve blitzscaling (short-term intensive growth)," but pointed out that "today, due to the introduction of AI, business models that generate high profits with small teams are becoming mainstream." He stated, "The software market has shifted to a phase where scale strategies using capital as a weapon are no longer as effective as they once were," and cited a specific example: "A certain Vertical AI company (Salient) is achieving eight-figure annual revenue with a team of six, and in the future, it will be possible to build multi-billion dollar companies with teams of 20 to 30 people."
3-2. Growth Models with Small Teams
The interview mentioned examples of "companies that achieved rapid growth with small teams," such as WhatsApp (about 30 employees) and Instagram (about 13 employees). Tan stated, "Markets that used to require thousands of people to capture can now be taken by small teams with the power of AI," and explained, "Productivity has improved dramatically because each individual collaborates with AI, maintaining teams for edge-case handling and evaluation, while leaving most routine tasks to AI."
4. Agency and Taste in Organizational Building
4-1. Cultivating 'High Agency' Individuals
Tan emphasized, "High Agency is the ability to recognize problems yourself and overcome difficulties through ingenuity, and this is a learnable skill." He cited Brian Armstrong of Coinbase as an example, explaining that "his stance of not breaking under pressure when attacked by the SEC, but clearing the issues through legal responses and lobbying, is the symbol of high agency." He also suggested a mindset for founders facing difficulties, stating, "When trouble occurs, even if the world thinks it's 'all over,' you can overcome it by continuing to search for the optimal solution."
4-2. Design Thinking and User Empathy
In product development in the AI era, Tan states that 'taste' is just as important as technical skills. 'Taste' refers to the sense for refining a product's usability, design, and the value perceived by the user. He points out that 'founders with only a technical background often fail to capture the true needs of users when they bring a product to market,' and suggests that they should 'thoroughly observe users' lives and work through ethnography to create solutions that are 10 times better.' Especially for enterprise products, he notes that the key is understanding 'how customers can get promoted' and whether you can provide that value.
5. Concrete Methods for Deep User Understanding
5-1. The Value of Undercover Research
Tan introduced an example of a startup that actually 'went undercover' in medical office work to experience the job. He says the founder of an AI company for medical billing worked as a medical clerk via Zoom while developing software locally, and by deeply understanding the practical work, they dramatically improved product quality. This method confirms the importance of thorough on-site understanding, as it 'provides insights that cannot be obtained simply by listening in meetings.' Tan also explained that 'even if you investigate secretly, you can just discard the prompts and code created from an IP (intellectual property protection) perspective,' demonstrating a methodology for conducting undercover research within ethical boundaries.
5-2. Learning Beyond Academic Backgrounds and Specialized Fields
Regarding situations where technical talent is concerned about 'losing stable white-collar jobs,' Tan states, 'It is more important to relearn CS (computer science), master AI coding platforms, take a job as a knowledge worker, and deeply understand that work.' For example, he introduced cases where people work in fields that seem unrelated to AI, such as tractor sales or the logistics industry, to gain operational knowledge before developing automation solutions with AI tools, emphasizing that 'learning beyond the boundaries of specialized fields creates new business opportunities.'
6. Future Outlook: YC and AI Regulation
6-1. YC's Future Strategy
Tan spoke about YC's 'Tree of Prosperity' vision, stating, 'We are aiming for a society where more people can choose to start a startup.' Conventionally, the career path of 'joining a major tech company and getting a stable job' was common sense, but he says that now, 'thanks to AI and software, we are in an era where more startups of 30 to 50 people can be born and provide high-quality work.' As a result, he suggests that 'the number of 1,000-person jobs that just involve checking boxes will decrease, and jobs in a 'flow state' that actually solve customer problems will increase,' showing his outlook on the positive impact that YC-born startups will have on society as a whole.
6-2. Proposals for AI Regulation and the Significance of Open Source
As AI develops rapidly, he also touched on how regulation should be handled. Tan stated, 'The 'Human-in-the-Loop' concept used in financial regulation should be applied to AI in general,' and raised a future scenario where 'even if an AI becomes CEO, a human will bear the ultimate responsibility.' He also pointed out that 'open-source AI plays an important role in preventing some companies from monopolizing the AI market.' He strongly argued, 'The current situation where 8 to 9 companies are competing and there are plenty of choices is very desirable. If we are tied to closed APIs or closed-source, innovation by emerging companies will be hindered.'
7. Leadership and Personal Insights
7-1. Coaching Founders: Difficult Decisions and 'How to Say Goodbye'
Tan explained the act of 'firing people' in startups by citing cultural challenges that organizations often fall into, stating, 'Firing someone is a painful act that everyone wants to avoid, but doing it at the right time makes the organization healthier and gives the fired employee a chance to enter a 'flow state' in a new place.' He stated, 'The person doing the firing needs to 'clearly communicate the decision and show support afterward,' and the essence of leadership lies in 'making difficult decisions and presenting the next steps.'' These words suggest how to face challenges that are unavoidable for business owners.
7-2. Learning as a Creator
In his creator activities, such as his YouTube channel, Tan reflects that 'the most viral video was the one where I honestly talked about my worst failure.' Specifically, he gained a huge response by candidly describing his experience of turning down the co-founding of Palantir in a video titled 'My $200 Million Mistake.' From this episode, he said, 'Creators can gain true empathy by speaking their minds and showing vulnerability. For building a personal brand, 'sincere storytelling' is more important than technology or titles.'
He further pointed out that 'the hardest skill to learn to succeed is saying no,' citing the Harvard Study of Adult Development. 'If you are dragged away by the temptations of fame and wealth, you may lose yourself. By properly declaring 'no' and spending time on family and important things, you can maintain true happiness.' This is also a suggestion for the work-life balance problems faced by founders and business owners.
What emerges from Garry Tan's interview is the fact that we have entered an era where 'dramatic growth can be achieved even with a small number of people' by utilizing AI. At the same time, he emphasized that basic principles such as 'making what customers really want,' 'demonstrating high agency,' and 'having deep empathy for users and design thinking' remain important in the AI era. Furthermore, concrete and practical methods such as 'undercover research' and 'pivoting to regain a beginner's mindset' were introduced, which are packed with useful insights for future startup founders.
Finally, the vision Tan stated—that 'more small companies should create rich value, and instead of organizations that just 'feed employees' like traditional large companies, there should be more workplaces where people can feel a 'flow state''—suggests the future of the startup ecosystem that will be accelerated by AI. By incorporating these frameworks into your own business, you will likely find hints for creating the next unicorn.