AI document automation that shows where every value came from.
Intelligent document processing turns PDFs, scans and photos into data you can check. We also cross-check that data across every document in the operation and link each result to its page.
Extract data from PDFs and scans, then cross-check it.
- Invoice2
- Packing list3
- Bill of lading2
- Health certificate4
- Catch certificate3
- Cold store records12
- Other5
- IdentifiersMatchesInvoiceLot L-2026-0418Health certificateLot L-2026-0418
- Weights, quantities and packagesMatchesInvoiceNet weight 21,340 kgPacking listSum of lines 21,340 kg
- Date order and consistencyReviewCatch certificateDischarge 06/05/2026Bill of ladingLoaded 04/05/2026
- Page continuityReviewCold store chartPages 1 to 7Cold store chartPages 9 to 12
- 01 · Invoice and packing list
- 02 · Certificates
- 03 · Cold chain
- 04 · Shipping
- 12/05 09:41 · M. Ortega · confirms net weight
- 12/05 10:02 · M. Ortega · closes and files no. 3
- 14/05 16:18 · L. Vidal · adds supplement 3.1
Load it as it arrives
Native PDFs, skewed scans, phone photos or handwritten forms. The original is kept untouched.
Split and classify
From a thirty-page PDF, the platform proposes where each document starts and what type it is.
Read beside the crop
Each value sits beside its crop of the scan. Confirming is one click; anything unreadable goes to manual entry.
Cross-check the documents
Lots, weights, dates, validity periods, signatures and page runs are compared. Each rule returns Matches, Review or Can’t check.
File it, frozen
On closing, one indexed file, numbered and frozen. Anything that arrives later goes into a supplement.
Plain-language rules your team signs off.
Your team describes each check in its own words. We build it, and you decide when it goes live.
Review doesn’t block. You can file with open items; the platform logs who decided and why.
A governing document per field. When documents disagree, the one you chose for that field is the reference.
Observation first. A new rule logs results without raising alerts until you switch it on.
Common documents
- Commercial invoice
- Packing list
- Bill of lading
- Health certificates
- Catch certificates
- Freezing charts
- Cold store records
- Licences and approvals
- Labels
Any other type can be added with its own fields.
What people usually ask us.
OCR turns an image into text. Intelligent document processing works out which document it is and extracts meaningful fields: lot, net weight, catch date. We add cross-checking between documents and human confirmation.
Yes, to split, classify and read, using models from external providers connected through our own layer, so the provider can be swapped without rebuilding the platform. Cross-checks run on plain-language rules, and a person confirms every result.
Yes. The platform measures each page’s legibility, asks for a rescan below the threshold and proposes handwritten values with their crop alongside. We don’t quote a generic accuracy rate: you see it in the pilot, on your own documents.
Probably not. Invoice OCR tools read supplier, amount and VAT and post them to your accounts for very little. We make sense when that invoice has to be checked against the packing list, the bill of lading and the certificates.
The review takes two to three weeks and the pilot about six. A full roll-out usually takes six to ten weeks, depending on plants, document types and buyers.
Send us three closed shipment files. We’ll show you what we would have caught.
The review takes two to three weeks, and the report is yours to keep whether or not you work with us afterwards.