Bills of lading, commercial invoices and packing lists — extracted into CSV, XLSX or JSON, with every low-confidence field checked by a human before it ships.
One to three of your messiest PDFs. We reply with structured output and a per-page quote.
Every B/L, invoice and packing list carries the same information — shipper, consignee, vessel, voyage, ports, containers, weights — and every one of them gets typed into a TMS, a spreadsheet, or an email thread by hand.
Generic AI tools get you to about 90% accuracy and stop. That last 10% is where the demurrage, the mis-declared weight and the wrong consignee live. We close it with rules, a second extraction pass, and a human on every flagged field.
You do not need a demo call, an API key, or a subscription to see whether this works.
One to three PDFs. The ones that break other tools — rotated scans, multi-page B/Ls, invoices with nine line items and no grid.
Rules pull the deterministic fields. A second pass reconciles the free-text ones. Anything ambiguous is flagged and checked by a human.
CSV for the TMS, XLSX for the team, JSON for the API. Every row carries a confidence score and a list of what was verified.
A fixed schema per document type, plus the line-item table. If your TMS expects different column names, send the template — we map to it.
| DESCRIPTION | QTY | KGS |
|---|---|---|
| LED PANEL LIGHT 36W | 1,200 | 2,640.00 |
| LED DRIVER 24V | 800 | 960.00 |
| ALUMINIUM PROFILE 2M | 2,400 | 3,120.00 |
| … 6 more rows |
1 row flagged: HS code missing on row 4
No subscription to forget about. The first 10 pages are free so you can test us on your worst files.
Billed per page after the free sample. No minimum.
From 2,000 pages per month. Cancel anytime.
Native-text PDFs from carriers, forwarders and ERP exports. Rotated pages, multi-page bills of lading, invoices with attached packing lists, and forms with overlapping labels are all in scope. Fully scanned paper documents need OCR, which we quote separately.
Rule-based fields (container numbers, dates, ports, weights) are deterministic and are validated — container numbers are checked against ISO 6346, dates are parsed and normalised, weights are converted to kilograms. Every field carries a confidence score. Anything below threshold is flagged, and a human reviews it before delivery. You will see the flags in the output, not behind a dashboard.
They are stored on an isolated server, used only for your extraction, and never used to train models. We can sign an NDA before you send anything, and we delete files on request.
Yes. Send one example row or your column headers with the sample and the output will match your layout, including date formats and unit conventions.
Standard turnaround is 48 hours from upload; 500+ pages ships same-day. The list rate is $39 per 100 pages. Recurring volume is quoted per shipment — tell us the monthly page count in the sample form.
CSV, XLSX and JSON are available today. A REST endpoint and webhook delivery are available on request for recurring volume.