Beyond the user-facing features, AI (especially GitHub Copilot and ChatGPT) acted like a pair programmer throughout Mediscan’s development. It helped me move faster, reduce repetitive work, and debug tricky issues that would have otherwise slowed me down.
One of the biggest productivity boosts was avoiding repetitive boilerplate.
For example, when I needed a reusable card layout, instead of manually writing out the whole widget structure, I prompted Copilot:
"Create a stateless Flutter widget called AnalysisCard that takes a title and a child widget. It should have a Card layout with some padding and elevation."
That gave me a working AnalysisCard widget in seconds. Normally I’d spend 10–15 minutes setting up the same code. It meant I could focus more on app logic instead of copy-pasting standard Flutter patterns.
While building the backend integration, I had to create several Dart models (MedicalReport, PatientDetails, ReportContent, etc.) for a fairly nested JSON response coming from a GenKit function.
Instead of writing all the constructors and fromJson methods by hand, I pasted the JSON into Copilot and asked:
"Generate Dart classes with fromJson factory constructors for the following JSON structure:"
In seconds, it generated all the model classes with proper parsing logic. That saved me easily half an hour of typing and avoided the small mistakes that creep in with manual JSON mapping.
I also leaned on AI for writing tests. For example, I needed to test my analyzeMedicalReport GenKit flow in functions/src/index.ts. Rather than starting from scratch, I asked Copilot:
"Generate a unit test for the 'analyzeMedicalReport' GenKit flow using Jest. Mock the 'ai.generate' call to return a predefined JSON string and assert that the flow's output is the correctly parsed object."
It gave me a complete test file structure with mocks and assertions. I still had to tweak it, but having a solid starting point made writing tests much faster and less intimidating.
Sometimes, coming back to a file after a break, I’d feel lost in a big widget tree. In report_analysis_screen.dart, I highlighted the FutureBuilder and asked ChatGPT to explain how it worked and what each connection state meant. That quick refresher helped me get back into flow instantly.
AI also spotted patterns I was missing. For instance, it suggested breaking down a large, repetitive widget into the smaller, reusable AnalysisCard. That refactor made the UI code cleaner, more modular, and easier to maintain.
When I ran into a type mismatch bug in my fromJson constructor, the app kept crashing. I copied both the Dart code and the sample JSON into ChatGPT and asked what was wrong.
It immediately pointed out that one field expected a List<String> but was sometimes null in the JSON. The suggestion was simple: add a null check with a default value (?? []). That fixed the crash instantly and saved me from wasting more time manually