I Think Some of You Should Close Your AI for a Moment
We’re currently producing a lot of SaaS junk that you then sell to one another.
And I mean that seriously.
You build a SaaS in a few days with Cursor, Claude, or Lovable. Landing page, dashboard, Stripe integration, and you’re done.
Three days later, you ask how to acquire customers. Two weeks later, you ask which features are still missing. And eventually, you want to sell the thing.
But what have you actually built?
I’ve been working in digital product development for 13 years and have founded several companies myself. I use AI every day, and I think it’s incredible what you can do with it.
But a lot of what I see here seems completely disconnected from the actual point.
We once built a company for business addresses and digital mail management. Not because we thought, “Let’s build a SaaS.”
We first talked to people. We understood the problem. We looked at providers, pricing, and legal requirements. Then we checked whether we could operate the business reliably in the first place.
That became a real company with physical locations, mail handling, and now more than 600 paying customers.
We later built the customer portal with AI as well.
But the AI didn’t know how our business worked. Which letters had to go where, which documents we needed, or who was allowed to see which data.
We had to understand and verify all of that ourselves.
A letter sent to the wrong customer isn’t a minor bug. Real people and confidential documents are involved.
And that’s what matters more to customers than whether you built your frontend with Claude or Cursor.
A lot of people currently care more about building software than about the business behind it.
Today, you can ask an AI to build you a SaaS, including the target audience, pricing, and business model. A few hours later, everything can look pretty finished.
But how do you know that anyone actually has the problem?
Because the AI said so?
And how do you know that the data is properly separated, the architecture is sound, and the system will still be maintainable later?
Because the AI confirmed to itself that everything was fine?
Do you see the problem?
Then the product keeps getting built without customers, money gets burned on tools, and eventually it’s put up for sale:
SaaS for $20. No customers, but huge potential.
Someone else buys it, adds a few more AI features, and carries on. The old architecture will probably be fine. The idea must have been good.
And if it doesn’t work again, it simply gets sold once more.
Eventually, this creates a market full of half-finished products where nobody even knows why they were built in the first place.
Of course, not every small SaaS is bad. But buying code and an interface doesn’t mean you understand the business.
So here’s my advice:
Stop developing for a few days.
Go back to the beginning.
What problem are you solving? For whom? How do people solve it today? Have you talked to them? Would they actually pay? And how will you reach them?
If you’re processing real customer data, please have the code reviewed by someone who knows what they’re doing. Don’t just ask the AI whether everything is secure.
I don’t think every founder needs to know how to program. But you do need to understand what you’re selling and what you’re responsible for.
AI can take a lot of work off your plate. It can’t take responsibility off your hands.
By all means, build with AI. Build quickly. Try things out.
But don’t confuse speed with progress.
A login isn’t a business model. Stripe isn’t proof of willingness to pay. And a polished interface doesn’t mean the product works in everyday use.
I’d rather see a founder with five paying customers and an ugly interface than someone with 30 AI features who has no idea who to sell them to.
Has anyone here ever taken over a SaaS that was built mostly with AI?
What was really under the hood? And did it become a business, or did you simply inherit the previous owner’s problems?
I’d genuinely be interested to hear.