The question keeps coming up: AI can now spin up the kind of tool a SaaS product used to charge a monthly fee for. So is the whole category dead? Mostly no. But the answer depends entirely on what the product was actually selling.
What’s on the chopping block
The basic stuff. Single-purpose tools and wrappers that may make one technical task easier: a UTM link builder, a one-off segmentation script, a cURL wrapper for testing APIs, a form that dumps into a spreadsheet. Low complexity, low stakes, low switching cost. If that’s the whole product, AI is a real threat, because anyone can vibe one up in an afternoon now. And honestly, they should. Vibe away.
What survives
The high stakes stuff. Complex, multi-dimensional data orchestration. Deep industry knowledge baked into the product over years. The places where being wrong is expensive. If the average person doesn’t know the ins and outs of bookkeeping and taxes, they have no business rebuilding QuickBooks on a Saturday. Regulation, edge cases, and decades of domain logic are a moat AI doesn’t dissolve on contact.
But there’s a catch
Internal builds have a shelf life. Say a marketer vibe-codes a landing-page builder on their personal laptop with Claude Code. Great. Where does it get hosted? How is it secured? Who owns the updates when that person leaves? Corporate AI initiatives are full of these orphans waiting to happen. The build was the cheap part. Hosting, security, maintenance, and ownership are the expensive part, and none of them show up in the demo.
Just because you can, should you?
Imagine there was a time when the only way to get a burger was to go to a restaurant. Then the grocery store started carrying the ingredients to make them yourself. Would that automatically make you into a restaurateur? Do you even want to cook every night? A lot of AI initiatives right now are a hammer looking for a nail. AI is a real tool with real capabilities. Whether building the thing is the best use of your time is a different question, and it usually goes unasked.
Time to pay the piper
Six months ago, token-maxxing was the whole strategy. Now the Anthropic and OpenAI invoices are landing in the CFO’s inbox, and plenty of teams have nothing substantial to show. That’s not a verdict on AI. It’s a verdict on pointing it at everything without a foundation underneath. Checking your email faster and generating slide decks can feel like a 10x unlock while moving none of the outcomes that actually matter.
So, is it dead?
Not quite. But “we built a wrapper” was never a business, and “we can build it ourselves” is not a plan. The SaaS that survives solves problems worth the complexity. The internal tools that survive get treated like real products: owned, hosted, secured, maintained. Everything else has a shelf life, whether you bought it or built it. The question was never whether you can build it. It’s whether it compounds, and who’s still running it a year from now.