AI startup is the new 'we have an app too'

AI startup is the new "we have an app too". Sounds like the future. Often it's just packaging. For months I've been seeing pitches, websites, and LinkedIn profiles where "AI" does roughly the same job the word "app" did ten years ago: a sticker you slap on the product so it looks like progress.

I remember the app wave well. Restaurants wanted an app. Hairdressers wanted an app. Car dealerships selling two cars a week wanted an app. In 95 percent of cases a decent website would have been completely enough. But people wanted to be part of it. The packaging became a substitute for the product. And exactly that is happening again — just this time with a model in the background.

What I consider a real AI product

For me, a real AI product is one where the AI carries the core of the value creation. If I take the AI out, the product collapses, because the alternative would be either much more expensive, much slower, or simply not feasible. A CRM with a small prompt field isn't an AI product. A classification engine that cleanly sorts millions of inbound documents and is measurably better at it than the rules the company has written over fifteen years — that is one.

That's a hard distinction. I know. But it means most of the "AI startups" I'm seeing right now are in truth SaaS companies that have built in an OpenAI API feature. What they're selling isn't the AI. What they're selling is a previous tool with a text generator on top. That can still be useful. But it doesn't justify an "AI-first" positioning. And it certainly doesn't justify valuations that exceed their competitors just because two letters are in the name.

Why packaging in this market still works, for now

I understand why so many jump on it. AI is an extremely well-scaling promise. "We automate your inbox with AI" opens budgets that "We're building you a better rule parser" never would. At the same time, many buyers don't currently have solid criteria for distinguishing real from superficial AI. If it looks nice in a demo screen, that's often enough for a pilot.

That's the phase we're in right now. And it won't last forever. I'll point at the app wave again. Four years on, nobody got money any more for "we have an app too" because the market had learned what an app really has to deliver. The same will happen with the AI label. The only question is how much budget will have flowed into packaging instead of product before that point.

My takeaway for founders and buyers

If you're building a product right now, ask yourself a sober question: what happens when I delete the word "AI" from my pitch? Does an interesting product remain that solves a real pain? If yes, all good. Then you can put the label back on. If no, you've just realised your product has no substance. That's painful but honest.

If you're buying, ask yourself a similar question: what business outcome do I expect from this tool in six months, regardless of which model runs behind it? If the answer is "less work in process X", let them show you how much. In numbers. With references. Without demo data.

I'm saying all this not because I think AI is hype. On the contrary. I work a lot with AI agents myself, and I believe they'll be some of our most important tools in the coming years. That's exactly why the current label inflation annoys me. It burns trust. It makes life harder for serious products, because they disappear among 50 copies.

"AI startup" is a useful word when it's true. It's a warning word when it mostly sits in the title of the deck. And as with the app wave: in the end what stays isn't the label — it's the product. Everything else is packaging, and packaging eventually ends up in the bin.