Visual extraction
Vision language models read layouts, tables, stamps, and Arabic, English, and Turkish type without a traditional OCR pass.
Documentation
INBAT is delivered as a private system, configured to your schemas and your ERP. These are the parts every deployment shares.
Vision language models read layouts, tables, stamps, and Arabic, English, and Turkish type without a traditional OCR pass.
Each field carries a calibrated score. Low-confidence values pause for a two-click human check before they can be saved.
The engine looks for seals and signatures, then compares invoices, packing lists, and customs files for mismatches.
Verified documents are indexed for questions in Arabic, English, or Turkish, with a citation back to the source page.