Can we build a mobile app with AI: does the end result actually work?

The hardest part of software isn't writing the code; it's making sure that code doesn't blow up six months later when a new OS version drops. No agency will touch messy AI code to maintain it—they'll just tell you it needs a full rewrite.

I started with the same optimism last month. Built the frontend in 3 days, but the app got rejected by the store 4 times over target SDK version warnings and permission policies. Ended up paying 20.000 TL to an expert just to clean up the architecture.

AI code looks amazing when it's running in isolation on your screen. Until 50 people place an order at once and lock up the database. AI code that hasn't been stress-tested with real users in production is a disaster waiting to happen.

Follow these three steps: 1) Look into existing open-source templates, 2) Only use AI to generate small function-level code snippets 3) Set up your developer accounts and read the store review guidelines cover to cover.

i tried that buttons look great on screen but everything falls apart once you connect to a database. anyway you fix one bug and it breaks something else endless loop tbh.

Will your app have live map tracking and stored credit cards? If it's just a simple menu listing where orders go through via text that's doable but if it's a multi-vendor marketplace, doing it alone will be really tough.

If your budget is tight, cut live courier tracking from v1. Stick strictly to store listings, cart, and checkout. Trimming the feature set makes it way easier to manage with AI and speeds up store approval.

AI definitely speeds up prototyping, no question. I built internal tools for our company without any dev background. But once real money, customer addresses and phone numbers are involved the liability shoots up. You definitely need pro help for security and app store submission at least.

Same here.

Let me summarize the topic, since several different answers were given. Any unwritten clause becomes a point of disagreement later, as both sides remember it differently.

Thanks for writing this, that's the right way. The biggest time-waster for us was not knowing who had the final say.

Let me clarify the technical side. Access credentials should be opened in the company's name, not personal accounts.

we need to take it step by step. i mean implementing a change request process doesn't slow things down, it speeds them up.

start with a small trial; dont commit to eveything at once.. and honestly correct me if Im wrong.

Correct.

I agree with this. If acceptance criteria aren't written, when the work is done is open to debate.

Trying to do this alone is the most expensive way.

There's a trap here, let me mention it. Any unwritten clause becomes a point of disagreement later, as both sides remember it differently.

Hope this helps.

The opposite happened to me, that's why I'm writing. People defend habits, not processes. Resistance comes from there.

Correct me if I'm wrong.

I went through the same thing two years ago. The answer varies greatly by industry; there is no one-size-fits-all rule.

If you have questions, write them; I'll answer as best I can.