There’s a wave of new AI-powered app builders promising live apps in under an hour. And in some cases, they deliver. You describe what you want, and out comes a basic CRUD app with forms, lists and buttons. For simple admin tools, internal dashboards, or prototypes, that’s probably enough.
We’ve tried them. Some of them are genuinely impressive.
But the more we see them used in the wild, the more we’re asking the same question: what happens when this needs to do something real?
These tools are brilliant at putting something on the screen. But they’re not great at answering the harder stuff: how does this scale, what if the logic changes, where’s the data going, how do we control who sees what?
And who actually understands how it works?
Because once something has been generated by a black box, abstracted three layers deep and glued together with hidden logic, there’s a good chance the answer is no one.
We’ve Built Enough Apps to Know Where They Break
We’re currently building a platform to analyse responses to documents. Could we have pushed it through ChatGPT and had it summarise things back to us? Probably, but we didn’t.
We put real effort into it, including steps to clean and align the data, deal with uncertainty, flag weird results, handle tone, expose contradictions. We built all that because we’ve already seen what goes wrong when it’s missing.
If your entire build happens in a no-code builder, how do you know the hard stuff has been dealt with? Most of the time, it hasn’t.
We Use AI All Over the Place… But Properly.
We’re not against any of this – we’re in it. We use AI every day, across loads of our work:
We’ve built an internal help bot trained on our own stack, so we’re not wasting time searching docs.
We use it to assist with QA and code review – spotting logic that looks off, or suggesting cleaner patterns.
It generates tests, rewrites functions, cleans things up.
In some client tools, we’ve used it to extract or summarise content – but never without a human sanity check.
You Get What You Ask For – Not Always What You Need
If I asked AI to build a bridge over the river behind my house, it’d probably give me a nice little wooden one. But if I came back the next day and said, “Now make it strong enough for a lorry,” you’d end up with the same basic bridge, just with 200 weird support legs bolted on.
It would work, but it wouldn’t be designed with that end use in mind. It wouldn’t account for real-world pressure, long-term wear, or anyone driving over it at speed. That’s what a proper engineer is for. And the same applies here.
The Code Isn’t the Point Anymore
Here’s something we talk about a lot: if your app idea can be built in an afternoon by an AI, what’s stopping anyone else doing it too?
What’s stopping us?
The code’s not the hard bit anymore. AI has made that cheap. The real value is in understanding the problem properly, designing something that solves it, and making decisions you can stand by when things get weird or busy or complicated.
You Still Need People Who Ask the Hard Questions
AI tools won’t ask:
Where does the data go?
What happens if the platform you used goes offline?
Who can see what?
How do we know the logic still works after version 4?
What happens if you succeed?
We’re the ones asking those questions, poking at the edges, and working out how to make sure what you build now doesn’t become a liability six months from now.
We’re not here to slow things down. But we’re also not here to pretend everything’s fine just because something looks finished.