Moats and AI Revisited
What business models and positioning are defensible in an AI-powered world? What moats will remain as AI grows more capable? Over the past few months I’ve put these questions to everyone who will listen, and then some, and incorporated the compelling answers into my list. It’s time for a recap.
The list remains short, though the headings are broad. I speak to many entrepreneurs and business operators and ask myself whether their offerings fall under one or more of these headings. The vast majority do not. That says to me that they will be overtaken by AI developments. Maybe they will last 12 or 18 months, but it won’t be long.
Many people made suggestions that I rejected. The rejections and the reasons for them are a topics for another time. But if you believe I’ve overloooked something that belongs on the list, do let me know!
Brand: Trust #
A recognized brand represents an economy of knowledge that applies whether decisions are made by people or AIs. You know how the burger will taste, or that you will receive good advice, or that you will find all of the tools you need. It isn’t necessary to comb through all of the details. Making decisions is costly. Reliance on brand will continue to reduce that cost.
Network Effects: Connections #
Network effects are robustly defensible because leaving an established network imposes a high opportunity cost. This is why nobody can build a Twitter killer, but someone built TikTok adjacent to Twitter. Thus network-based plays tend toward winner-take-all. However, the ability of AI to provide fast-follow work-alike implementations means that the competition to reach that winner status will become utterly savage.
How many distinct networks are supportable? I don’t know. My guess is that we will see a short-term proliferation of niche networks followed by dramatic consolidation. Will we see numerous agent networks (as opposed to human networks) spring up, expanding the supportable number? Again I don’t know, though I would guess yes.
Physical Work: Dexterity #
Businesses that integrate with the physical world and require adaptable dexterity, such as home repair, auto mechanics, and construction need capabilities beyond those of existing robotics. Advances in robotics will allow automation of an increasing number of these activities, but the ability of AI to manipulate the physical world will trail far behind its digital impact. In the meantime, AI will affect general areas of business like finance, project tracking, monitoring, etc, and in areas where adaptable dexterity is not as critical, like autonomous vehicles.
Sports: Human Performance #
Who will care whether the Robo Dallas Cowboys beat the Robo New York Giants? Some may. But sports have at their core a question of human performance. Even if robotic teams and leagues appear, they cannot replace that fundamentally human element. While everything about business operations may change with AI, I believe people will still want to see what humans can do.
Status: Human Touch #
Luxury goods signal status today. You wear a high-end watch to tell everyone what a special person you are. What happens when more of our world is created robotically? Wearing the hand-crafted watch or dress or shoes can signal your special nature. Or paying a premium to have a human massage rather than a robotic massage. Or matriculating at a university that holds in-person classes.
Genuinely Proprietary Data: Ownership #
Trade secrets will be reinvented by increasingly sophisticated AI. Sophisticated analysis of publicly available data is a variation on that theme. Those will quickly become indefensible. Data created in the course of business may be different. When I make a phone call, my carrier and I jointly produce metadata that they have contractual rights to use. That is genuinely proprietary data.
Monopoly: Exclusion #
A patent is one common form of monopoly. Utilities with exclusive right to serve a region is another. These grants of monopoly are defensible insofar as the grantor allows it. They let you exclude competition.