From Vibe Coding to Production
Raw material: takeaways from the planIQ case study
From “No replacement of the existing system” — Polypoint PEP stayed the system of record; replacing it was an explicit non-goal.
For key systems, look for the gaps in your processes and tools instead of reinventing the system from the ground up.
From “The prototype is the concept” — prototyping with AI from day one, so functionality discussions happened against something real within a month.
No concept document conveys what a change will feel like to its users. Prototyping early makes for better discussions than a specification.
From “Hosting a prototype is not hosting the real thing” — built in Lovable, then migrated to University IT infrastructure once the concept solidified.
A vibe-coded app tells you what the software can do. Writing, running, hosting and securing it — and the data in it — takes additional thought and belongs in the plan from the start.
From “Data on a need-to-know basis” — sensitive HR fields never entered the app; roles limited what each user could see.
Just because data is available doesn’t mean the app should use it.
From “Handover to internal IT” — the project carries its own context: original brief, design patterns, decision log.
When you build with coding agents, context matters as much as code. If you don’t plan for it from the beginning, change requests get out of hand quickly.
Share
About the Author
