At a freight forwarder, the accounts team can receive thousands of supplier invoices a month: shipping lines, airlines, road hauliers, destination agents, warehouses. Each one with its own format, charges and surcharges. Automated invoice processing promises to end the job of reading them, comparing them with what was agreed and posting them by hand. This guide explains how to do it well and what results to expect.

Why freight invoices are a hard case

Automating invoices from an office-supplies vendor is simple. Freight invoices are not, for several reasons:

• Highly varied formats: every shipping line, airline or haulier invoices its own way, often across several pages and languages.
• Variable rates: freight that depends on route, chargeable weight or container type.
• Surcharges (fuel, congestion, security, THC…) that change frequently.
• Link to the shipment file: every line must be allocated to the right shipment to calculate its margin.

That's why generic OCR falls short and reconciliation ends up in an intermediate spreadsheet.

How automated processing works, step by step

1. Capture

Invoices arrive by email, supplier portal or EDI. The first step is to centralize intake: a single mailbox or folder from which the system picks up every document automatically.

2. AI extraction

A reading engine trained on logistics documents (at Kaiona, K-Scan) identifies supplier, invoice number, dates, amounts, charges, surcharges and shipment references (BL, AWB, container number), with no per-supplier templates. If you want to understand what's behind it, we cover it in AI in logistics: real use cases today.

3. Reconciliation against rate and shipment file

This is where the real savings are. Each line is automatically compared with the agreed rate and with the shipment data in the TMS. If everything matches, the invoice moves on without anyone touching it. If there's a discrepancy (an unagreed surcharge, a different amount, a shipment file that doesn't exist) it's flagged as an exception with the reason.

4. Posting to the ERP

Validated invoices are posted directly to the ERP and allocated to their shipment file. With K-Plug this happens on the ERP you already use, with no migration.

94% of invoices processed automatically. The team only reviews the exceptions.

An example: a shipping line invoice

Imagine a three-page invoice from a shipping line covering several containers across different shipment files. It includes freight, a fuel surcharge, terminal handling at origin and destination, and a congestion surcharge that wasn't in the quote.

Processed by hand, someone has to find each container, look up its shipment file in the TMS, open the agreed rate, compare charge by charge and split the amounts. If the congestion surcharge slips through, it gets paid without being passed on to the customer or disputed with the carrier.

Processed automatically, the whole invoice is read and each line is matched to its container and shipment file. Amounts are compared with the rate. Freight, fuel and terminal charges match and go through. The congestion surcharge is flagged as an exception, with the affected shipment file and the reason. Accounts decides in a minute whether to accept it, pass it on or dispute it.

What data is extracted from a freight invoice

For reconciliation to be automatic, extraction has to go beyond the header. A well-trained engine captures, at a minimum:

• Header: supplier, tax ID, invoice number and date, currency and due date.
• Shipment references: BL or AWB number, container, booking, customer or shipment file reference.
• Charge lines: freight, surcharges (fuel, security, congestion), origin and destination charges, with amount and taxes.
• Calculation bases: gross and chargeable weight, volume, number of containers or packages, route.

With that data structured, comparing against rate and shipment file stops being manual work.

How to handle exceptions

The quality of an automated invoice system is decided by how it handles exceptions. A few practices make the difference:

• Set tolerances: small differences from rounding or exchange rates that are accepted without review, so the team isn't swamped with false alerts.
• Classify by reason: an unagreed surcharge isn't the same as a missing shipment file or a duplicate invoice. Each type has its owner and its response.
• Close the loop with the supplier: repeated discrepancies with the same carrier are useful input for the next rate negotiation.
• Review the rules regularly: if an exception keeps appearing and is always accepted, the rule or the reference rate probably needs updating.

Common mistakes when automating invoices

• Stopping at extraction. Dumping data into a spreadsheet saves almost nothing if someone then has to reconcile it by hand.
• Not having rates in digital form. Without a reference to compare against, automatic reconciliation isn't possible.
• Testing only with clean invoices. The value is decided by small suppliers and odd formats.
• Not defining what an exception is. Agree tolerances and owners before you start.

Which KPIs to track

• Straight-through processing rate: invoices posted with no intervention.
• Average time per invoice, before and after.
• Discrepancies caught before payment: unagreed surcharges, duplicates, off-rate amounts.
• Monthly close time for supplier reconciliation.

Results: a real case

A Kaiona client handling 90,000 invoices a year automated their processing. The result: €141,000 in annual savings. That's the underlying logic: the team's hours go into resolving discrepancies, not typing amounts.

€141,000 saved per year on a volume of 90,000 invoices.

How to get started

Gather one month of invoices from your ten main suppliers, together with the agreed rates. That's enough to estimate a realistic automation rate and the savings before committing to anything. Billing is usually the best first step in a broader logistics digitalization project.