Quote Copilot Procurement Agent — AI skill for Claude Code

Quote Copilot: AI Procurement Agent

Quote Copilot reads messy vendor quotations, turns them into one comparison a buyer can trust, and answers award questions in plain English.

**[Live demo](https://appapppy-cl7jquf8iannuwn4kvf32r.streamlit.app/)**: password protected. The password is shared on request.

---

The problem

When a buyer sends out a request for quotation (RFQ), the replies come back in every format a vendor can think of:

• an Excel sheet in the vendor's own layout;
• a letterhead PDF, or a scanned one with no text in it;
• a Word letter with a discount buried in a paragraph;
• a one-line email ("rest same as last year");
• a phone photo of a printed rate card with a price crossed out and a new one written in by hand;
• a revised quote that replaces an earlier one.

Prices arrive per piece, per 100, per 1,000, per kg or per strip, in rupees or dollars, with or without tax and freight. Before anyone can compare them, someone rebuilds everything in a spreadsheet by hand. That is slow, and the mistakes in it are invisible.

What it does

Three AI loops, all live calls to a language model:

1. Drafting. The buyer describes what they need in a chat. Quote Copilot drafts a structured RFQ (line items, a supplier questionnaire and commercial terms) and exports the Excel template that goes to vendors. It asks for missing essentials instead of inventing them.
2. Reading and normalising. It reads every vendor reply, whatever the format, and builds one comparison:
   • the same lines, in the same unit, on the same basis (excluding tax, delivered to the plant);
   • questionnaire answers and attached certificates next to the prices;
   • anything uncertain sent to a review queue instead of silently into the table.
3. Analyst. The buyer asks questions such as "split it, cheapest per line, only vendors who cleared the questionnaire". The answer comes with tables, charts and an Excel award summary, and every number in it is checked by code.

There