I’ve been watching the AI startup landscape pretty closely for the past year and a half, and I think we’re entering a fundamentally different phase for entrepreneurs. The first wave was chaotic, exciting, and — for most people who jumped in — a graveyard of good ideas that got steamrolled by the platforms they were built on.
Let me explain what I mean, and why I think the playbook is changing.
The Wrapper Problem
Since ChatGPT launched, the most accessible way to build an AI business has been what the industry calls a “wrapper” — you take OpenAI’s API (or another model provider), build a nice interface or workflow around it, and sell that to users. Resume builders, copywriting tools, coding assistants, customer service bots. You’ve seen hundreds of these.
The problem is obvious once you think about it: you have ZERO competitive moat. Your entire product is a thin layer on top of someone else’s technology. The moment OpenAI or Google or Anthropic decides to add your feature natively — and they will — you’re done. Your “product” becomes a free checkbox in someone else’s settings panel.
I’ll be honest — I’ve fallen into this trap myself. Multiple times. I’ve kicked off no fewer than seven wrapper-style projects over the past year. Some were genuinely useful. A couple had real traction. But every single time, the same nagging thought kept me up at night: what happens when the next model update makes this irrelevant?
I’m probably luckier than most in this regard. I genuinely love playing with this tech, so the time wasn’t wasted even when the business case fell apart. But not everyone has that luxury, and “I had fun” isn’t a business model.
The Other End of the Spectrum
On the flip side, you had the ambitious founders trying to build foundational AI — training their own models, creating genuinely novel architectures, pushing the boundaries of what’s possible. That’s where the REAL moats live.
But here’s the catch: that requires capital. Serious capital. We’re talking tens of millions minimum, often hundreds of millions. The compute costs alone are staggering, and you’re competing against companies with essentially unlimited resources. Unless you’ve got deep pockets or the kind of track record that opens institutional doors, this path has been effectively closed to most entrepreneurs.
So for the past year, AI entrepreneurship has been stuck between two bad options: build something with no moat, or try to build something you can’t afford. Neither is great.
The Second Wave Playbook
Here’s what I think is shifting, and it’s pretty exciting.
The smart move for this second wave of AI entrepreneurs isn’t purely an AI play at all. It’s a hybrid. The playbook looks something like this:
Build a wrapper business coupled with a traditional business that would be massively improved by the tech.
Let me unpack that because the nuance matters.
Instead of building “AI Tool X” as a standalone product, you pick a traditional industry — something with real revenue, real customers, real operational complexity — and you build an AI-enhanced version of THAT business. The AI isn’t the product. The business is the product. The AI is what lets you operate it faster, cheaper, and smarter than the incumbents who are still doing things the old way.
Think about it. A logistics company that uses AI to optimize routing isn’t a “wrapper.” A legal services firm that uses AI to do document review at 10x the speed isn’t a “wrapper.” A property management company that uses AI to handle tenant communications and maintenance scheduling isn’t a “wrapper.” These are real businesses with real competitive advantages — the AI just happens to be the engine underneath.
Why This Works
The beauty of this approach is that it solves both problems simultaneously.
You get a moat — but it’s not an AI moat. It’s a business moat. Brand, customer relationships, operational know-how, regulatory compliance, scale. All the boring stuff that actually makes traditional businesses defensible. Your competitors can’t just copy your API calls because the value isn’t in the API calls — it’s in everything around them.
You build AI expertise at someone else’s expense. While you’re running a profitable traditional business, you’re also accumulating massive domain-specific knowledge about how AI works in your particular vertical. What prompts work. What fine-tuning helps. Where the models fail. What data matters. That knowledge compounds over time, and THAT — eventually — might become your actual AI moat.
You can afford to be wrong about AI. If the models change, if the APIs shift, if the whole landscape gets disrupted again — you still have a business. The AI made it better, but the business doesn’t evaporate without it. You adapt and keep moving.
The Takeaway
I think the entrepreneurs who are going to win in AI over the next two to three years aren’t going to be the ones building the flashiest demos or the cleverest prompt chains. They’re going to be the ones who picked a real industry, used AI to operate at an unfair advantage, and built genuine scale while everyone else was chasing the next wrapper idea.
The efficiency gains from AI are real and they’re enormous. But efficiency isn’t a product — it’s an advantage. Point it at a real business, ride it to dominance in that space, and let the moat build itself through scale and brand and earned expertise.
That’s the second wave. And I think it’s going to be a LOT more interesting than the first.
