Keep and productionize
Retain sound components and close testing, configuration, deployment, and operational gaps.
Solution
AI built my app. What should I do next?
ChatGPT, Claude, Codex, Cursor, Replit, Lovable, Bolt, and similar workflows can move a product surprisingly far. Screens and workflows may function while the architecture, failure paths, release process, and ownership remain uncertain. The next step is to evaluate the system as a whole—not assume the code is either good or bad because AI helped create it.
Request a Technical AssessmentThe demo works, but important questions remain
Core screens and workflows exist, but nobody has reviewed the architecture.
The database or integrations work only in the expected path.
Large generated changes keep introducing unrelated failures.
Testing, logging, and error handling are incomplete.
Deployment or store release has never been repeated reliably.
Documentation and technical ownership are missing.
A working demo is not automatically a production system
Generated code can be useful and maintainable. The risk comes from unverified assumptions across dependencies, configuration, authentication, data handling, APIs, deployment, and edge cases.
The business needs to know what should be kept, what should be repaired, and what must change before customers or operations depend on it.
Avoid making the uncertainty larger
- Do not keep requesting broad rewrites without preserving a working state.
- Do not share production secrets through prompts or unsecured channels.
- Do not assume a successful happy-path demo covers failure handling, security, or release.
A controlled next step
- 1
Stop large blind changes and preserve the current working state.
- 2
Verify ownership of source control, domains, cloud, databases, and platform accounts.
- 3
Assess the repository, architecture, dependencies, configuration, and build path.
- 4
Identify production, security-sensitive, data, testing, and operational risks.
- 5
Decide what to keep, repair, replace, or simplify.
- 6
Create a controlled implementation and release path.
Possible paths after evaluation
Focused rescue
Repair unstable architecture, integrations, state, build, or release paths.
Replace selected parts
Keep valuable working behavior while replacing layers whose risk exceeds their value.
Relevant evidence
Case studies
AI and software workflows
Review existing architecture and workflow examples without treating them as customer outcome claims.
View context →Relevant HerbDev services
Find out what the AI-built application actually needs.
Preserve the current state, then assess what works, what is fragile, and what must happen before production.
Request a Technical Assessment