Twelve pages of line items, one at a time
A shipment of mixed homewares arrives with a commercial invoice listing ninety lines and a packing list that splits them across forty-two cartons. The operator types the shipper, consignee, invoice number, currency, totals, and then the lines themselves into the forwarding system, checking the gross weight on the packing list matches the booking. Next job: another shipper, another layout, a scanned invoice at a slight angle.
Half the operator's afternoon is spent as a typist, and the typing is exactly where mistakes creep in.
Why the documents cannot just be imported
Shippers produce these documents from their own systems, in their own formats, and send them as PDFs or scans. There is no standard your forwarding system can import directly, so a person bridges the gap.
- Every shipper's layout is different, and some change without warning.
- Many arrive as scans or photos rather than text.
- Invoice and packing list totals do not always agree with each other or with the booking.
- Line item descriptions are inconsistent and sometimes vague.
- Operators check and type at the same time, so both suffer.
What the retyping costs
Operator time is the obvious cost. The less obvious one is errors: a weight or value typed wrongly travels into other documents and filings, and has to be corrected later with more work and sometimes an amendment fee. Discrepancies between the invoice, the packing list and the booking are found late, when they are harder to fix. And experienced operators leave jobs that are mostly typing. During peak season, when bookings stack up before a holiday shutdown at origin, the typing backlog becomes the bottleneck for the whole desk, and documents that arrive late in the day wait until tomorrow to be entered.
The document extraction we build
- Documents arriving by email or upload are recognised by type: commercial invoice, packing list, and others you choose.
- Scans are converted to text, and an extraction model pulls out the parties, references, currency, totals, and each line with description, quantity, unit, value and weight.
- Totals are cross-checked: lines add up to the invoice total, packing list weights and carton counts match the booking, and the invoice and packing list agree with each other.
- The extracted data fills the job in your forwarding system, through its API or import, as a draft.
- The operator sees the extracted fields beside the original document, with any mismatches highlighted, and confirms or corrects them.
- Corrections are used to improve extraction for that shipper's layout.
| Step | Manual | With extraction |
|---|---|---|
| Reading the document | Operator reads and types | Extracted automatically |
| Line items | Typed one by one | Filled as a draft |
| Checking totals | Sometimes | Every document, every time |
| Invoice vs packing list vs booking | Found late | Mismatches flagged on arrival |
After the change
Operators check and approve rather than type. Mismatches are raised with the shipper on the day the documents arrive, not when a filing is rejected. The same data flows into the other documents and filings that need it, typed once or not at all. Operators spend more of their time on the work that needs a forwarder's judgement.
We do not decide what goes into any customs declaration. The extraction gives your customs team or broker clean data to work from; their decisions stay theirs.
Is this your operations team?
- Operators retype invoices and packing lists every day.
- Weights and values have been typed wrongly and caused amendments.
- Invoice and packing list mismatches are found late.
- Scanned documents are common.
- Good operators leave because the work is repetitive.