Unifying 4 invoice formats into a single structured data lake

The procurement team received invoices from 200+ vendors in at least 4 distinct formats — ranging from simple single-page invoices to multi-page documents with line-item tables, tax breakdowns, and shipping details. Manual data entry into their ERP was slow, error-prone, and couldn't keep up with volume. Existing OCR tools failed on the format variety and couldn't reliably extract line items from complex table layouts.

The team set up a DocumentIQ project with fields matching their data lake schema: vendor name, invoice number, date, PO reference, line item descriptions, quantities, unit prices, tax amounts, and totals. Multi-row extraction handled line items automatically — even when tables spanned multiple pages or used inconsistent column headers. Annotations on 10-15 representative invoices per format taught the AI the structural differences. The feedback loop let AP clerks correct the occasional miss, and re-extraction with corrections brought accuracy up quickly.

• 4 distinct invoice formats processed with a single field configuration
• 15,000+ invoices digitized in the first 3 months
• Line item extraction accuracy reached 95% after initial annotation round
• Average processing time dropped from 8 minutes per invoice (manual) to under 30 seconds
• Direct integration with data lake via CSV/Excel export eliminated manual data entry entirely