How AI Is Changing Startup Building
This is a time full of excitement around artificial intelligence, yet we are only at the beginning. Just look at your phone’s home screen: How many apps do you have there that are built entirely on AI? And how many of them were created with the help of artificial intelligence tools? For most people, that number is close to zero, excluding obvious examples such as ChatGPT or Grok. Yet one day, every spot on the screen could be occupied by such apps. Where is the AI-powered calendar or the social network built on artificial intelligence? During the shift from the web to mobile, the way we use messaging, social networks, and email changed rapidly—everything moved into apps. Why is that not happening as quickly now? This situation points to an enormous opportunity, because artificial intelligence has not yet changed the way we work. So far, we simply search less on Google and use prompts more.
Questions About Building Startups with AI
Artificial intelligence is changing not only products, but also how those products are created. We are at the beginning, and many things remain unclear. For example: Will the startups of the future need fewer or more employees? The argument is that AI provides a thousandfold leverage. What once required an entire company can now be handled by an individual. Imagine one person overseeing a thousand agents that spend all day coding, and that becomes a billion-dollar company. On the other hand, if an AI startup grows quickly but some tasks still have to be performed by people because AI cannot yet handle them, then many people will still be hired. Perhaps taste will turn out to be the key, and a large number of designers will be needed. Or it may be possible to prototype many popular products, only to eventually run into human limitations.
Another question concerns defensibility against fast-moving AI competitors. If a product can be copied instantly, how can you maintain an advantage when AI commoditizes everything? The last decade in consumer apps has shown that in an environment with little technical differentiation, user growth and network effects determine the outcome. Perhaps the pace of software creation will accelerate so much that the differentiator will be the ability to innovate continuously, launch new features, and create new products. One idea is that it will only be worthwhile to build products with multi-year horizons and high equipment costs. For example, space technology or enterprise hardware—areas where you need deep knowledge of today’s markets for the bet to pay off. Any software that AI can build within a few years will quickly lose value. The advantages will be similar to those in direct-to-consumer businesses, where brand and temporary distribution expertise are decisive. But building a great company this way is much harder.
Costs, Organization, and Geographic Changes
Will building startups with AI be cheaper or more expensive? Some infrastructure, such as foundation models, requires enormous investment in equipment. In theory, some applications should be easy to produce. But growth and distribution cost a great deal of money if products are to reach customers. The last decade has taught us that even if building a web application is inexpensive, acquiring users can cost millions. Failure is common because competition is intense, so that may remain the limiting factor.
How will team organization change? Does it make sense to have separate functions such as engineering, product, or design? When product creation becomes fully multimodal thanks to AI that can build software from a description or sketch, perhaps all disciplines will merge into one. The structure of work has changed for centuries. Through industrialization, we moved from home workshops, where families made things by hand, to factories and corporations.
Should San Francisco remain the center of technology? It used to hold that position because of network effects, venture capital, and expertise. Its advantage has weakened because talent is moving to New York and knowledge is spreading through podcasts and blogs. If creating products becomes as easy as creating content, entrepreneurs will be everywhere, just like content creators. Without the need to scale through employees or capital, these centers will become dispersed.
Venture Capital and History
How will venture capital work in a world of fragmented startups? For a long time, it has been about investing in spin-offs. What if it becomes incredibly easy to build and test a new product? If people are willing to pay, it can become profitable immediately, especially when it is built by one or two people at low cost. Such products are being created around the world, so venture capital will become dispersed like growth capital, available globally.
A related question: Do stages such as pre-seed, seed, or Series A/B/C still make sense? These categories group together different stages. Perhaps more products will jump directly from zero to Series A. And does it make sense to fund people to run experiments, or will that simply be a side project?
These questions are fascinating because the modern tech industry is only decades old, not centuries old. Things such as venture capital, the mobile internet, or startup hubs such as San Francisco emerged recently and can change quickly.
History confirms this. In pre-industrial Britain in the 18th century, a metalworker worked in a workshop next to the house, the entire family made things by hand and sold them at the market in small batches. In the 19th century, industrialization brought factories with masses of workers, corporations to organize them, shareholders to finance them, and layers of managers. Burnham’s The Managerial Revolution describes this well. Similarly, in the 17th century, the development of naval power and trade networks led to the invention of limited liability companies. The East India Company needed both technological and commercial innovation to become the world’s largest corporation, with an army of 260 thousand soldiers.
It therefore seems inevitable that current structures will not be sufficient to organize AI-driven production. If we moved from family-based organization of work to factory-based organization, what will a world of agents, computing power, and models bring?
The Most Optimistic View
In the best-case scenario, AI will allow fewer people to produce more. AI startups should need far fewer employees. Defensibility should come from amazing features and technologies, not merely from distribution and monopoly. We want more new things, not for established large companies to dominate the next generation. These technologies should make building startups less expensive.
But the opposite can also be argued. The biggest winners could be those with enormous data centers, data, and computing power—which would lead to centralization, where the big players become even bigger. Or AI may simply be a great set of features that does not help with marketing and distribution. Established companies will slowly convert their products to AI and overpower startups.
The question is: Will established companies get innovation first? Or will startups get distribution first? Perhaps the incumbents will win.
The coming years will set the direction. Recent years have brought waves of teams focused on AI research and foundation models. Now that models have absorbed all the world’s data, their efficiency is approaching its limit. The next few years will be about creating business logic on top of these models. These companies will not conduct research or train their own models. They will be model-agnostic and create compelling interfaces on top. These products are entering every sector, from tools to the integration of entire industries with technology.
It will be a wild ride. And thanks to automation, which increases productivity by as much as 40 percent, startups are becoming more agile and disruptive in industries such as healthcare, finance, and logistics. Smaller teams can accomplish more because AI takes on marketing, customer service, and coding. Generative AI accelerates ideation, analyzes trends, and helps identify gaps in the market. Barriers are falling, so even an individual can run a valuable company, increasing the number of startups and changing entrepreneurial culture.
Source: andrewchen.substack.com
