Phase 1: Project Discovery
Don't worry about the code yet. Just answer these questions in plain English about the client's problem.
Thinking...
Turning plain English into a tech architecture
Step 1: How the App Works
1. The Messy Input
Unstructured Inputs
2. The AI Reader
LLM reads and organizes the data
3. The Python Math
Arithmetic & Saving Data
Preview: AI Vibe Code Sandbox
Data Structure (For the AI)
Business Rules (For Python)
Live Test Run
Step 2: Who Builds What?
Team Lead / Human
- Approve what data the AI will extract.
- Ensure we don't break compliance rules.
AI Backend Assistant (ChatGPT/Gemini)
- Write the code connecting to legacy software.
- Organize the JSON data safely.
Frontend UI Coder (Cursor IDE)
- Build the screens the user will click on.
- Build the big red "Escalate to Human" buttons.
Data Integrity Code
- Handle data saving smoothly.
- Write standard math tests (no AI hallucination).
Step 3: Golden Rules
1
Data First, UI Last
Never let AI build buttons until it knows exactly what the data looks like.
2
No AI Math
AI is for reading and writing text. Use regular Python code for all math and money.
3
Human Checkpoint
Stop the code here.
Step 4: The Client Pitch
App Proposal
How this tool solves their headache.
Money & Time Saved
$0
Calculated Cost Avoidance / Year
Approved By: _________________
Date: _________________