AI-Assisted RFQ Sourcing Automation for Automotive Manufacturing
Finding, contacting, and qualifying a new supplier took an automotive parts maker up to six weeks — much of it spent writing emails by hand. GroupBWT built an AI-assisted pipeline that does it in nine days, inside their existing Oracle ERP.
CLIENT STORY
A European manufacturer of automotive wiring harnesses and electrical assemblies, supplying commercial-vehicle OEMs. Its sourcing team buys across four commodity groups. About half of every program is catalog parts with known distributors; the other half is custom-made — cables, PCBs, connectors, housings — sourced from scratch each time. The whole team works inside an Oracle-based in-house ERP, and management set one rule before any vendor talk: automate the work, add no separate tool.
| Service: | Data Engineering |
|---|---|
| Industry: | Automobile |
| Region: | EU |
Read summarized version with
Catalog parts are the easy half. The custom half eats my week — we search Google, then write the same certificate and NDA emails to supplier after supplier, only to learn weeks later they were never released by our OEM. — Commodity Manager, Electrical Components
Management was clear: no second platform. Whatever we add has to run inside the Oracle system we already use — don't hand my people another tool to learn. — Commodity Manager, Electrical Components
A Sourcing Cycle Rebuilt From Zero Every Program
Every new sales RFQ starts a fresh supplier hunt — one for each component a harness program needs. For catalog parts, that is routine. For the custom half, about 50% of every program, it is slow, manual, and rebuilt end to end each time. That half eats the buyer’s week.
Most of that time was not analysis — it was correspondence. Each candidate meant the same emails written by hand: requesting IATF 16949 certificates, asking for proof of OEM release, sending an NDA before drawings could be shared, then chasing the reply. The Oracle ERP the team runs on logs the RFQ and holds prices once they exist, but it has no reach into the outside supplier market. Discovery, outreach, and quote collection all happen outside it, by hand.
Qualification was the worst of it. Every candidate had to prove IATF 16949 certification (the automotive quality standard) and prior release by the OEM — Scania, MAN, Volvo — confirmed one email at a time, often only to fail weeks later.
An AI-Assisted Sourcing Pipeline Inside the Existing ERP
GroupBWT mapped the real RFQ flow first, and was honest about the one step it would not automate — choosing the supplier. The rest became a single pipeline, with an AI layer doing the mechanical reading and writing a buyer used to do by hand. It was built one commodity group at a time and wired into the Oracle ERP rather than bolted on beside it.
Supplier discovery. When an RFQ opens, the pipeline reads the component spec straight from the ERP — no second data entry — and an AI layer turns it into search criteria, then sweeps 30+ distributor and manufacturer sources in one pass, in place of the manual Google search. A new source is a setting change, not a rebuild.
AI-drafted supplier communication. This is where the manual process bled the most time, so it became the core of the build. For each shortlisted supplier, AI drafts the outreach in the buyer’s voice — the certificate request, the OEM-release question, the NDA cover note — tailored to that specific component and supplier. The buyer reviews and sends; nothing leaves without a human approving it. Replies come back into the same thread, where AI reads each one and files what matters.
Qualification and quote intake. From those replies, AI extracts the documents that decide the call — IATF 16949 status and proof of OEM release — and drops unqualified suppliers in days, not weeks. Offers arrive as emails and PDFs; AI reads them into one like-for-like comparison, retiring the manual spreadsheet. The screened shortlist and quotes land back in the Oracle ERP — no second tool — where the buyer prices and logs the RFQ.
Tech stack: Multi-source web scraping, sign-in handling for gated supplier sites, LLM-based email drafting with human-in-the-loop approval, PDF and certificate reading (OCR + LLM extraction), workflow automation, Oracle ERP integration.
The AI writes the emails and reads the replies — that's the part that ate their week. What it never does is pick the supplier; that's human judgment, and there's no database to crawl for it. Every message still goes out only after a buyer approves it.
Six Weeks to Nine Days, With Zero New Tools
By having AI draft outreach and read replies, GroupBWT cut custom-part sourcing from six weeks to nine days — the correspondence that once filled a buyer’s week now takes minutes to review and send.
Because qualification (IATF 16949 and OEM release) is read out of supplier replies as they arrive, unqualified suppliers drop out in days — roughly 60% fewer dead-end evaluations.
AI parsing offers into one structured comparison retired the manual quote spreadsheet, saving the buyer about 18 hours per custom RFQ.
One AI-assisted pipeline serves all four commodity groups and lands every result inside the existing Oracle ERP — with a human approving every supplier message and the final supplier choice.
Looking to automate a manual RFQ or procurement cycle?
We map your sourcing process, put AI on the mechanical work — drafting outreach, reading certificates and quotes — with your team approving every send, and run it inside your existing ERP. We start with a feasibility review, not a platform pitch.
You have an idea?
We handle all the rest.
How can we help you?
