QualityNew

Automatic feasibility check of customer drawings at RFQ intake

Added 2026-10-14

Role

quality engineer, sales engineer

The problem

Customer drawings arrive with mixed units, different tolerance conventions and finish notes buried in PDFs. Reading them by hand slows intake, and tight tolerances, special finishes or threads are sometimes missed, then turn into costly nonconformance reports after the order is won.

Step by step

1. Write your capability policy: tolerance bands per process, finishes you can and cannot do, materials, threads.
2. Collect 50 past drawings, including ones that caused NCRs, as a test set.
3. Run extraction and check accuracy on the price- and risk-driving features.
4. Encode policy checks that flag out-of-capability features automatically.
5. Route every No-Go or low-confidence drawing to a quality engineer before quoting.
6. Record NCRs from spec misreads before and after.

Tools

• Werk24 drawing-reading API: used at the injection moulder
• Others: CADDi, Paperless Parts, and LLMs with vision for a manual test

As reported by the source; most figures are vendor or self-reported and not independently verified.

Human review: cautions

• Anonymous vendor case study.
• Do not upload export-controlled drawings to tools without the right data residency.
• Extraction errors on critical tolerances are costly; keep human review on those features.

You are a quality engineer at a [process] plant. Attached is a customer drawing. Our capability: [tolerance limits per feature type, finishes, materials].
1. Extract all dimensions with tolerances, GD&T, threads, finishes and notes.
2. Compare each with our capability and mark Go / No-Go / Check.
3. List questions for the customer.
Mark anything unreadable.

Sources

• https://werk24.io/case-studies/plastics-company-feasibility-analysis