You have probably seen the demos by now. Someone types a few sentences into Replit, Lovable, or Bolt, and within minutes there is a working app on screen. A booking system. A client dashboard. A job tracker. It looks polished, it functions, and the whole thing took less time than your lunch break.
I get why people are excited. Honestly, I am too. These tools are remarkable at what they do. The problem is what they don’t do - and that gap is exactly where small businesses get stuck.
The first 80% is genuinely impressive
Let me be clear - I am not here to bash AI app builders. The first 80% of what they produce is often very good. You describe what you want in plain English, and you get a functioning prototype with a decent interface, some basic logic, and enough structure to make you think “this could actually work.”
For prototyping and exploring ideas, that is fantastic. The trouble starts when you try to take that prototype and turn it into something your business actually depends on.
Then you hit the wall
The last 20% of any software project is where the real work lives. It is less glamorous, rarely makes for a good demo, and it is the part that AI app builders consistently struggle with.
Error handling. What happens when a customer enters their phone number in the wrong format? What happens when someone submits a form twice by accident? A prototype might crash or silently lose data. Production software needs to handle these gracefully.
Edge cases. Your app works perfectly with 10 test records. But what happens with 10,000? What about when a customer has a hyphenated surname, or a booking that spans midnight? Real-world data is messy, and messy data breaks fragile code.
Database migrations. Three weeks after launch you need to add a new field or change how data is stored. Doing that without losing existing data takes careful planning - not just prompting an AI and hoping for the best.
Performance. Your app loads instantly on your laptop with fast broadband. Now try it on a three-year-old phone on a patchy 4G connection in a warehouse. That is where many of your users will be.
Security. Is customer data encrypted? Are API endpoints properly authenticated? AI-generated code often has security gaps that are not obvious until someone exploits them.
Mobile responsiveness. If half your customers are using phones - and they probably are - the app needs to work properly on every screen size, not just the one you tested on.
Deployment. Getting code running on your laptop and getting it running reliably on the internet are two very different things. DNS, SSL certificates, server configuration - none of this is glamorous, but all of it matters.
Why the last 20% matters more than the first 80%
That impressive first 80% is the easy part. It is the visible part, the part that makes for a great demo. But it is not what your customers experience when something goes wrong.
Your customers experience the last 20%. They experience what happens when the app is slow, when their data does not save properly, when the layout breaks on their phone. That last 20% is your reputation.
For a small business, an app that mostly works is worse than no app at all. Because “mostly works” means lost bookings, frustrated customers, and hours of your time firefighting problems you do not know how to fix.
”It works on my screen” is not a standard
Someone builds an app, tests it on their own laptop, and declares it ready. Then a customer in Birmingham tries to load it on an older iPhone and nothing renders. Or someone with a slow connection watches a spinner for 30 seconds before giving up.
Your screen is not your customer’s screen. Your internet connection is not their internet connection. Software that is only tested by the person who built it is not tested at all.
How I handle this differently
I use AI tools every day. They are part of my workflow, and they make me faster. But I use them as a starting point, not a finish line.
The difference is what happens after the AI generates code. I review it. I test it on real devices. I add proper error handling. I think about what happens six months from now when you need to change something.
That is the difference between a developer who uses AI and an AI that replaces a developer. The tools are brilliant at generating code quickly. They are not brilliant at knowing whether that code is ready for real people to rely on.
When I build software for a small business, the app you get is the finished product - tested, secure, performant, and ready for your customers to use. Not 80% of the way there with a wall of unsolved problems waiting around the corner.
Worth a conversation?
If you have hit that wall with an AI app builder - or if you want to skip it entirely - I would be happy to chat. No pressure, no jargon, just a straightforward conversation about what you need and whether I can help.
Get in touch and tell me what you are working on.