Paper CMRs and delivery notes are read by a multimodal LLM (OCR + structure extraction), automatically matched against the TMS, and checked for discrepancies — missing seals, bad signatures, damage entries. Achieves 70–90% automation on document workflows.

Structures and prepares fuel, subcontractor, highway, and repair invoices for accounting. Self-checks subcontractor invoices against the tariff agreed in the order — catching overcharges before payment. Significant loss-prevention value.

A customer raises a claim. The assistant searches for transport data, CMR scans, GPS trace, door-opening events, and photos, then compiles a structured insurance card file ready for submission — typically saving days of manual work.

Checks incoming orders for ADR classification against Decree 25/2014 (NFM). Flags incorrect UN numbers, packaging groups, and missing documentation before the vehicle departs. An incorrect classification carries a 6–7 digit fine — a second pair of AI eyes is essential.