Prototype AI Lead Qualifier

I want to see, quickly and in a contained environment, whether a small AI agent can take the short notes we get from new prospects and come back with something immediately useful. Your task is to build a first-pass prototype—no bells and whistles, just clean code—capable of the following:

• Accept a single inbound message (hard-coded sample, CLI prompt, or simple endpoint is fine for now).
• Pass that text to an LLM such as OpenAI GPT-4 or Claude via their public API.
• Return a structured JSON object that includes:
– Key facts you can reliably extract from the message (name, company, need, budget if mentioned, etc.).
– A one-line intent or category tag (e.g., demo request, pricing, partnership).
– A lightweight priority or lead score (low / medium / high is plenty).
– A two-sentence natural-language summary.
– A short, friendly reply draft we could send as a follow-up.

Keep it simple: either a single Python script or a tiny FastAPI app with one POST route. Hard-coded prompts are fine; just comment them clearly so we can tweak later. The goal is clarity of flow—input, model call, formatted output—rather than enterprise-grade architecture.

I will consider the job complete when I can run the script locally, drop in any lead message, and receive the JSON response with all fields populated in a logical way.