Guide / Industrial AI
AI for automotive suppliers: From RFQ to a reviewable quotation
AI can help automotive suppliers prepare a request for quotation (RFQ) by turning scattered requirements into a reviewable record. The first useful output is a set of source-linked requirements, contradictions and unanswered questions. Pricing and technical commitments remain separate decisions.
A customer request can contain enough information to start a conversation and still be too ambiguous to price. Volumes may differ between an email and a spreadsheet. A drawing revision may not match the supporting specification.
AI can help prepare that material for review. Its useful role is to organise requirements, preserve their sources and make unresolved questions visible. Engineering feasibility, pricing and the final commitment remain within the company's approval process.
An illustrative request
Consider a fictional request for a machined metal component. The email and attached spreadsheet state different annual volumes. A drawing carries a revision identifier that the spreadsheet does not reference.
A useful application would retain those differences rather than merge them into one apparently complete specification.
| Field | What the draft should show | Responsible next step |
|---|---|---|
| Annual volume | Conflicting values with their locations | Sales obtains confirmation |
| Drawing revision | Identifier and original file | Engineering confirms applicability |
| Material | Source designation without an invented substitute | Technical and supply review |
| Delivery date | Requested date | Planning checks feasibility |
| Price | Not yet approved | Calculate against confirmed requirements |
This is an illustrative workflow, not a report of a Welf customer deployment. It makes no claim about measured savings or accuracy.
Keep the evidence beside the field
Each extracted requirement needs a traceable source, such as a table cell or document passage. The reviewer should not have to repeat the entire search to establish where a number came from.
Some drawings or poor scans may exceed the application's reliable capabilities. Leave those fields unresolved or allocate them to a specialist. A first release can deliberately cover text and spreadsheet requirements while retaining manual drawing checks.
Historical quotations can provide context. They must not silently replace the customer's current requirement or establish today's price.
Include incomplete requests in the trial
Choose one recurring request type and agree its required fields. Include missing drawings, conflicting quantities and an unknown revision alongside complete requests.
Score extraction and contradiction detection separately. A completed form can be worse than a draft that correctly identifies gaps. Reviewers need the original request as well as the proposed record.
Connect preparation to the existing commercial process
Product mappings and pricing rules should come from their authorised systems. The application can retrieve relevant records and prepare a proposed entry. Before anything is written, a responsible person checks the confirmed requirements and the intended change.
Record which approved version was transferred. If the customer later changes a requirement, the process needs a new review rather than continuing with an outdated approval.
Integration should also handle an uncertain response from the target system. A retry must not create duplicate records simply because the application failed to receive a confirmation.
Measure until the request is ready to progress
Document extraction time is only part of the outcome. Measure the work needed to reach a technically clarified request, including questions, checks and corrections. Compare cases of similar complexity.
A useful pilot can reveal that the main obstacle is missing customer information rather than document processing. That finding should change the design, even if the extraction demonstration looks impressive.
Welf works with sales, engineering and IT on the quote-to-order workflow. Our automotive page explains the wider industry context.