Growhold
Apps made with AI app builders arrive in demand asking to be finished, fixed and stabilised three times as often as the average request. The trigger is the first outside check: patient data rules, paying customers, a second tenant, search. Build for that week.
The app took a weekend. It worked on the first demo.
It is a healthcare app, and it runs on Replit. To hold patient data it needs a hosting partner that will sign a health data agreement. Replit will not. So the app moves to Azure.
That is the moment this post is about. Not the build. The first time someone from outside checks the work.
Almost done is the most common state
Apps built in Lovable, Replit or Base44 arrive in demand with a recognisable request attached. Finish it. Fix it. Stabilise it.
Requests that name one of these builders talk about finishing, fixing or stabilising about three times as often as the average request. They talk about production and scale half again as often. Ownership and compliance follow close behind.
Requests that name Lovable, Replit, Base44, Bolt or v0, against every request in the market.
The pattern is the same everywhere. The prompt gets an app most of the way. The last stretch is where the outside world starts asking questions.
The first auditor is rarely an auditor. It is the first person whose trust the app has to earn.
Four auditors
The first outside check comes in different clothes.
- The compliance officer. A medical billing tool built in Base44 has to meet the rules for patient data and connect to the insurers' billing network. The backend moves off the builder to a self-hosted, compliant system.
- The paying customer. A platform with eleven paying customers has to live in a real code repository and a real cloud account before the next feature ships. Eleven paying customers is a reason to stop improvising.
- The second customer. A CRM made in Base44 serves several companies now, and is most of the way through its move to its own Postgres stack. Each company's data has to stay apart, and the prompt never asked for that.
- The search engine. A Lovable site carries twenty pages and a hundred blog posts. It has to rank, and every URL has to survive the move to WordPress.
None of them care how the app was made. Each one tests something the builder was never asked to provide.
What to build for the week after the weekend
These businesses already have a product and users. What they lack is everything the first audit checks.
Someone could start on any of these today:
- A production check for AI-built apps. Point it at a Lovable or Replit project and get back what an auditor would find: exposed keys, missing row-level security in Supabase, no backups, no tenant separation.
- Compliant hosting with an export. A home for prompt-built apps that signs health data agreements, keeps the code in the customer's own repository, and lets them leave on day one.
- The finishing work, packaged. Tests, error tracking and a staging copy, generated for the exact stack these builders produce.
The weekend still matters. It is how most new software starts now.
Build for the Monday after it.
