OCR for CMR and Invoices in Freight Forwarding: How AI Eliminates Re-Entry

OCR for CMR and invoices in freight forwarding is one of the easiest automations to deploy in the freight industry, and at the same time the one that pays back the subscription cost the fastest. A dispatcher manually re-enters 12 documents per day on average: CMRs, invoices from carriers, customer orders, consignment notes. AI with an OCR model fine-tuned for transport reads them in 3 seconds with 98% accuracy.

In this article we show how OCR works for CMRs and transport invoices, how it differs from a regular document scanner and how much time you can realistically recover in a freight forwarding company. The material is based on data from a BOBI pilot in 14 freight companies in Q1 2026.

What is OCR for freight forwarding documents?

OCR (Optical Character Recognition) is text recognition on images. Classic OCR from a decade ago returned raw text, without context, without fields, without verification. Modern AI-OCR understands the structure of a document: it knows that "CMR No." is the consignment number, "Forwarder" is the forwarder name and "EUR 1820" is the net amount. From this it builds a structured record ready to be inserted into the TMS.

How OCR for CMR works step by step

1. 1The driver or dispatcher takes a photo of the CMR with a phone (or uploads a PDF scan).
2. 2AI recognises the document type: international CMR, domestic consignment note, e-CMR.
3. 3It extracts fields: sender, consignee, carrier, place of loading, place of unloading, load, weight, quantity, dates, signatures.
4. 4It validates the data (VAT number consistency, field format, date consistency) and flags any gaps.
5. 5It links the CMR to an existing order in the TMS or creates a new record.
6. 6It archives the scan in PDF/A format with searchable metadata.
7. 7It only notifies the dispatcher if it detects a mismatch or missing field.

OCR for transport invoices: what does AI read?

• Seller and buyer VAT numbers: automatic VIES verification
• Invoice number and issue date
• Invoice line items (transport, transhipment, insurance)
• Net, VAT and gross amounts
• Payment due date and bank account number
• The linked transport order number (if shown on the invoice)
• Currency and exchange rate (for EUR/USD invoices)

How much time a freight forwarding company saves

Time to enter a CMR

From 8 minutes (manually) to 3 seconds (AI)

Number of CMRs/day per dispatcher

On average 12, i.e. 96 minutes saved every day

OCR accuracy for CMR

98% of fields read correctly, 2% require manual approval

Eliminating typos in VAT

94% drop (auto-validation in VIES)

Time to match invoice to order

From 4 minutes to automated: AI links documents itself

OCR vs e-CMR: are they the same thing?

The e-CMR is the electronic version of the CMR regulated by the additional protocol to the CMR Convention (2008). It is a digital document with electronic signatures; no OCR needed because the data is already structured. The catch: not all countries have adopted the e-CMR (Poland has, since 2019), and many customers still use paper or a PDF scan. OCR is the bridge between the paper world and the digital TMS.

Most common mistakes when rolling out OCR in freight forwarding

• Low-quality scans: a phone in auto mode gives a better result than an old photocopy
• No VAT validation after OCR: typos slip into the TMS unnoticed
• Skipping the e-CMR: do not OCR an e-CMR, the data is already structured
• No archiving of originals: OCR returns text, but the PDF scan must be kept for audit
• Choosing a generic OCR (e.g. plain Tesseract): without understanding the CMR context, accuracy drops to 70%

Is OCR safe for data?

OCR models can be cloud-based (faster, more accurate) or local (with data staying in the company). BOBI runs OCR in line with GDPR and the AI Act, with providers bound by a data processing agreement.