Developers can now use Prompt Driven Development to refactor existing codebases with the help of AI tools like GitHub Copilot and Visual Studio Code. A recent walkthrough demonstrated how to restructure an inventory API built with Azure Functions, FastAPI, and Cosmos DB by separating business logic from database operations. The process starts with clearly defining goals in a prompt file, specifying which logic stays in the CRUD layer and what moves to a new service layer.

This structured approach allows AI to perform multistep tasks like moving code and updating references. Developers maintain oversight by monitoring task progress and the Problems panel for errors. Rather than manual fixes, the AI uses feedback from tooling to resolve issues iteratively.

Once changes are made, the API is tested locally to verify behavior. This method brings AI into existing workflows without sacrificing quality or control.