Detailed guide contents
Engineering-led guidance. This guide draws on Productionise's standards and practical engineering experience. Examples are generalised to protect confidentiality.
1. You may not need to start again
AI-generated code is not automatically unusable. If part of the app works, is understandable and suits the production need, there is no good reason to discard it merely because AI helped create it.
The opposite assumption is equally risky. A feature working in a preview does not prove it is secure, reliable or maintainable enough for customers. The developer should assess the current app before recommending what stays or changes.
2. What can potentially be kept?
Potentially useful work includes the visual design, content, page layouts, customer journeys, working components, business rules, integrations and source code.
Not everything will survive unchanged. Even a part that must be replaced can remain valuable as a clear example of the outcome you want, saving the next person from guessing.
- Designs and layouts
- Content and customer journeys
- Working components and features
- Useful business logic
- Correctly implemented integrations
- Source code
3. What does the developer need from me?
A developer needs more than the visible website. Give them enough access and business context to understand the whole system.
Lovable currently provides GitHub synchronisation and code download options for external development. Availability can depend on plan and permissions. The code does not automatically move or recreate every backend, data service or integration.
Do not send passwords or secret keys in ordinary documents, email or chat. Use appropriate account access or secure credential-transfer processes, and keep important business accounts under business control.
- Access to the source or business-controlled repository
- A clear description of what the app should do
- A list of important external services
- Domain and hosting information
- What currently works, fails or remains unfinished
- Appropriate access to relevant business-controlled accounts
4. What should the developer assess first?
The first review should establish what the app depends on, how it handles important information, which outside services it uses, how it is hosted, and what still needs work before real customers depend on it.
The result should explain the important risks and options in business terms. It should not assume that changing tools fixes every problem, or that keeping the current setup is always cheaper.
5. Keep it, strengthen it or replace it
A useful assessment gives each important part one of three outcomes. The choice should follow evidence about the app, customer journey and business risk:
- KEEP — the existing part is suitable for its production role.
- STRENGTHEN — the idea or code works but needs better controls, testing or support.
- REPLACE — rebuilding that part is the safest, clearest or simplest option.
6. Can I keep using Lovable afterwards?
Potentially, yes. Productionisation does not have to mean abandoning Lovable or the prototype. A separate prototype version can remain useful for visual experiments, layout ideas, customer-experience changes and new concepts without changing the live production service.
A designer, or several designers, can explore and compare those versions. Once an idea is approved, it becomes input to a controlled production change—not an automatic overwrite.
This keeps experimentation possible while protecting customers. The business retains a useful creative workspace, and production changes still receive review, testing and a deliberate release.
- Experiment
- Review
- Approve
- Productionise
- Test
- Release
Should I care?
This matters if…
- You need a developer to finish, launch or maintain a Lovable app that already does useful work.
- Real customers, business information, payments or important enquiries will soon depend on the app.
- You are unsure what access, context or control a developer needs before they can assess it.
It may not be urgent if…
- The project is still a private experiment and you are deliberately testing different ideas.
- You do not yet know which customer problem or journey the prototype should support.