AI-Powered RFQ Automation & Supplier Sourcing
A Singapore-based supply chain company serving the oil and gas industry receives a constant stream of inbound requests for quotation (RFQs). Each RFQ carries several line items, and different items are sourced from different suppliers. The purchase team reads every RFQ by hand, works out which supplier can supply which item, and drafts each outbound supplier email manually — slow, repetitive work that depends entirely on each person's memory of who supplies what. On top of that, a catalogue of 10,000+ historical product records held inconsistent, free-text descriptions that were almost impossible to search or match against suppliers.
~20 hrs
Saved per week
5x
Faster RFQ drafting
90%
Sourcing match accuracy
100%
Supplier knowledge captured
Automated Workflow
How the automation handles the entire process end-to-end
This workflow is for demonstration purposes only. We custom-build automation architectures tailored to each client's specific requirements, tools, and processes.
How It Works
Each step in the automation pipeline
Historical Data Standardisation
Claude reads 10,000+ free-text product descriptions and generates clean, standardised generic product names, creating a reliable supplier-product baseline in Supabase.
Vector Embeddings for Every Product
Each standardised product is embedded and stored in a Supabase pgvector column, so line items can be matched by meaning, not just exact text.
Mailbox Reading & Intake Validation
The system reads the procurement desk Outlook mailboxes, picks up inbound RFQs by their reference ID, and validates format and duplicates before anything is processed.
Line-Item Split & AI Sourcing
Each RFQ is broken into line items; exact part numbers are matched first, otherwise pgvector similarity search finds the most likely suppliers from the master table.
Confidence-Scored Matching
Every supplier-to-product match gets a High / Medium / Low score; confident matches are drafted automatically, uncertain ones are held for a human.
Grouped Draft Preparation
Line items are grouped by supplier and a clear outbound RFQ draft is prepared per supplier in the desk mailbox, carrying the correct company-entity signature.
Human-in-the-Loop Review
The purchase team reviews every draft and sends it themselves; nothing leaves automatically. Status moves from "drafted" to "sent".
Self-Improving Database
Corrections, new suppliers and manually added products flow back into Supabase and are re-embedded as vectors, so matching gets steadily more accurate over time.
Before & After
The measurable impact of automation
Tech Stack
Tools and technologies powering this automation
Want this for your procurement team?
Let's discuss how we can automate your workflows and save your team hours every week.
Get Started