Guide
You built a prototype with an AI app builder. What changes before it goes live?
A prototype from an AI app builder or an AI coding tool is a good way to show what you mean and test interest. Before it takes payments, holds customer data or connects to your operations, someone has to own the business logic, the security, the data and life after launch. First, though, someone should decide whether it deserves a production build at all.
What a prototype proves, and what it doesn't
| A prototype shows | It doesn't yet show |
|---|---|
| What the product does and how it looks | Who pays, and whether the numbers hold |
| That a flow can work for one user | That it works for many users at once, with permissions |
| That the idea can be explained | That data is protected and stored where it should be |
| That you can move fast | Who fixes it at 2 a.m. after launch |
The gap between a demo and a product you can run
Before real users and money arrive, these usually need an owner:
- sign-in, roles and permissions;
- payments, refunds and invoices;
- personal data: what is stored, where, for how long, and under which agreements;
- integrations with your CRM, accounting or operations;
- separate environments for testing and release;
- monitoring, backups and recovery;
- support, updates and security fixes;
- the monthly cost of running it;
- accounts and ownership: code repositories, cloud accounts and domains held where you can reach them.
Independent research shows why this matters. Veracode's 2026 GenAI Code Security Report puts the average security pass rate of AI-generated code at 56% across the models it has tracked, without security-specific prompting. In the 2025 Stack Overflow Developer Survey, developers' trust in the accuracy of AI output fell to 29% from 40%.
Three ways forward
- Keep and harden. The prototype's structure holds up, and the work is security, data, integrations and operations around it.
- Rebuild, using the prototype as the specification. The prototype shows exactly what you mean, and a production version is built alongside it.
- Stop and test demand first. If you don't yet know whether people will pay, the prototype is the right tool for now. Come back when you know.
Which one fits depends on the product, not on the tool it was built with.
How we approach it
Send us the link: a working prototype shows what you mean faster than any brief. We use the same AI tools in our own engineering; what we take on is responsibility for the result. For a new product that fits, the Build Decision comes first: proceed, reshape or decline, in writing, with reasons, within seven working days. If we proceed, the offer states what we keep from the prototype, what we rebuild and why, and the first version starts from $50,000, excluding VAT.
If your product is already live with paying users and growth has stalled, Existing Product Review may be the right route instead.
Questions
Tell us what you're working on.
New product, business process, live product, grant or introduction: a few lines are enough to find the next step.
We reply within one working day. The form opens with this route already selected.