The hype machine for Artificial Intelligence is in high production mode. Every week, I get pitched by someone whose startup will revolutionize healthcare, finance, education, or even how you order a burrito at 2 AM. Investors are buying anything with AI in the name, and founders are jumping on the bandwagon. But the majority of them will not last. Some will quietly disappear, and others will fall apart publicly. A couple will blow through hundreds of millions before somebody acknowledges that the emperor has no clothes. That is why, before I invest a single dollar in any AI firm, I ask one question. Just one. It’s not the question everybody anticipates. 3 Common Questions I’ve Noticed That Cut Through the Hype I don’t ask “How sophisticated is your model?” I don’t ask “What is your addressable market?” I don’t even ask “Who is on your founding team?” My question is very simple: “If the AI part isn’t a differentiator anymore, what will still be valuable in your product?” Since that day will come, when “AI magic” that is rare today will be ubiquitous tomorrow. Models get commoditized. APIs for a fortune this year are low-cost or even free next year. You ask me, what do you do when the moat that you are confident exists disappears overnight? I’ve seen it. In the early 2010s, facial recognition was the thing. I was following one startup whose software had the ability to identify a person from a crowd in seconds. They signed multiple deals with security firms, and it was so groundbreaking that the founders were even featured in the media. Three years later, all the major cloud providers offered a faster and cheaper API that did the same thing. The startup’s customers left, and the company ended up selling its old code for almost free. They were smart and talented, but built all of their worth around a feature that became a commodity. Why This Question Matters More Than Ever In AI, a competitive advantage can disappear in months. A breakthrough that can take years to develop can be imitated in a matter of months. What’s groundbreaking in January can start to feel antiquated by August. If the only thing that’s truly unique about a company is how its model is doing today, they’re in deep trouble. Open-source releases can eliminate that advantage nearly overnight. Compute costs decrease. Datasets even get sold, and the magic trick isn’t so magical anymore. I want to know what will still be worth something when that initial novelty has worn off. How I Apply the Question When I am listening to a pitch, I want to hear how a founder talks about their value. If they are talking non-stop about their proprietary model, I cut them off. I ask, “If next month someone develops a free open-source model with the same accuracy and speed as you, why will customers still go with you?” “If next month someone develops a free open-source model with the same accuracy and speed as you, why will customers still go with you?” The strongest founders don’t panic. They talk about network effects, unique datasets that are hard to replicate, switching costs, community loyalty, brand trust, or a product platform that can do more than one thing. Weaker founders stumble. They continue to mention “our model is best.” That is when I know they are betting on a transitory advantage. What Strong Answers Look Like Successful founders are normally able to point to one or more of the following: Data Moats - Proprietary datasets that are difficult or impossible to copy. Not just large but also distinct and extremely applicable. Integrated Workflows- AI is embedded within an overarching system that’s deeply ingrained in user workflows. Substituting it would collapse the whole process. Community and Network Effects -Platforms that become more valuable the larger the number of participants, like marketplaces or shared ecosystems. Brand and Trust - In finance, legal services, or healthcare, compliance and trust could be more valuable than raw technical ability. Distribution Advantages - Regulatory approvals, strategic alliances, or privileged positions that cannot be duplicated by any competitor. Data Moats - Proprietary datasets that are difficult or impossible to copy. Not just large but also distinct and extremely applicable. Data Moats Integrated Workflows- AI is embedded within an overarching system that’s deeply ingrained in user workflows. Substituting it would collapse the whole process. Integrated Workflows- Community and Network Effects -Platforms that become more valuable the larger the number of participants, like marketplaces or shared ecosystems. Community and Network Effects Brand and Trust - In finance, legal services, or healthcare, compliance and trust could be more valuable than raw technical ability. Brand and Trust Distribution Advantages - Regulatory approvals, strategic alliances, or privileged positions that cannot be duplicated by any competitor. Distribution Advantages None of these rely on today’s AI being better. They rest on a deeper basis. The Risk of Ignoring It Not challenging this question can be very costly. A friend of mine invested in an AI-powered travel planning app. It was fast, efficient, and could provide decent itineraries. Six months later, plugins for ChatGPT and open-source software replicated the same for free. The app had nothing else for the users to linger for. It had no special deals, no social network, even no way to save and accumulate prior itineraries meaningfully. The company closed down during the year. If my friend had inquired, “What will still be valuable when the AI is no longer new?” they may have departed. “What will still be valuable when the AI is no longer new?” AI Will Continue to Evolve — So Should Your Standards I don’t despise AI. In fact, I’ve invested in companies that use it beautifully. If you listen carefully, you’ll notice that the few ones that survive treat AI like plumbing. It matters, but it isn’t the house itself. When electricity was everywhere, investors didn’t just invest in “electric lamp companies.” They invested in firms that applied electricity to achieve something permanent: creating things, refrigerating things, and communicating. AI is heading down the same path. The capabilities will keep on costing less and replicating more easily. The real value will be in what you build on top of it. The Takeaway Before you start an AI company or invest in one, always imagine a moment when the AI component is available to everyone tomorrow for free. Ask yourself whether the business will still be worth something after that. If so, then perhaps you have found something with durability. If not, then you have been presented with a temporary advantage disguised as a moat. In an industry that’s sprinting at the rate of quarters instead of decades, that single question can make you millions. It has made me more than once. The next time someone comes along selling you “the future of AI,” don’t get blinded. Look for the part of the business that still works when the magic trick is yesterday’s news. That is where the real value is.
