Extracting bill of materials from supplier spec sheets across 200+ part families

The engineering team received spec sheets from 80+ suppliers in inconsistent formats — some as tabular PDFs, others as scanned documents with handwritten annotations. Each spec sheet contained a bill of materials with part numbers, material grades, tolerances, quantities per assembly, and unit costs. Engineers were manually re-keying this data into their PLM system, averaging 25 minutes per spec sheet. Errors in transcription led to incorrect parts reaching the production floor, causing costly rework.

The team configured DocumentIQ with fields matching their PLM schema: part number, material grade, quantity per assembly, tolerance range, unit cost, and supplier part reference. Multi-row extraction handled the BOM line items automatically. Annotations on 20 representative spec sheets taught the AI to handle both clean tabular layouts and scanned documents with irregular formatting. The feedback loop caught edge cases like merged table cells and footnoted tolerances.

• 200+ part families with spec sheets fully digitized
• Average processing time reduced from 25 minutes to under 45 seconds per document
• Part mismatch incidents on the production floor dropped by 68%
• 94% first-pass extraction accuracy, reaching 98% after annotation refinement
• Engineering team reclaimed 30+ hours per week previously spent on manual data entry