Logistics
How to automate bills of lading, CMRs and proof-of-delivery paperwork
Every delivery produces paperwork, and until it's matched to the right job, checked and filed, you can't invoice with confidence. This guide explains how AI handles PODs, CMRs and bills of lading, including the awkward ones, and where people still need to stay involved.
Why freight paperwork is still a bottleneck
Electronic PODs and e-CMRs are growing, but paper is still everywhere. Drivers photograph signed delivery notes on their phones, subcontractors email scans days later, and bills of lading arrive as PDFs from shipping lines and forwarders. Each document has to be matched to a job, checked and filed before the job can be invoiced or a claim defended.
Done by hand, that means someone squinting at blurry photos, searching the TMS for a reference number and chasing anything missing. It is slow, it delays invoicing, and it is easy to miss the one note in the margin that matters, such as "2 cartons crushed".
The documents, briefly
- Proof of delivery (POD): the signed delivery note or app record confirming the consignee received the goods, with any remarks.
- CMR consignment note: the standard document for international road freight, with sender, carrier and consignee details, goods description, and boxes for reservations and the consignee's signature.
- Bill of lading: the receipt and contract of carriage for sea freight, and often a document of title, issued by the carrier or forwarder.
How the automation works
- Collect. Documents are picked up wherever they arrive: the driver app, a scanning inbox, subcontractor emails or a shared folder.
- Identify. Each page is classified (POD, CMR, bill of lading, delivery note, something else) and multi-page scans are split into separate documents.
- Read. AI extracts references, consignee, date and time, quantities, whether it is signed and stamped, and any remarks or reservations, including handwritten ones.
- Match. Each document is matched to a job in the TMS using the references, consignee, date and vehicle, not just one field, because references are often missing or misread.
- Check. Rules flag problems: no signature, quantities that differ from the job, clausing, the wrong delivery date, or a document that matches no job.
- Act. Clean documents are attached to the job and the job is released for invoicing. Exceptions go to a person, and clausing can open a claim or alert the customer service team.
A worked example
The awkward cases, and how to handle them
- Poor photos. Give drivers quick feedback in the app when a photo is unreadable, rather than discovering it a week later.
- Handwritten remarks. Treat any handwriting in the remarks or reservations area as a reason for review until accuracy on your documents is proven.
- Missing references. Match on several fields together, such as consignee, postcode, date and vehicle, and send ambiguous matches to review.
- Multi-drop and part deliveries. Compare delivered quantities with each drop, not the job total.
- Signatures. Automation can check that a signature is present. It cannot verify who signed, so keep that in your claims process.
Getting started
Collect a few hundred recent documents covering your usual sources, along with the job each one belongs to. That gives you an answer sheet to measure accuracy against before going live. Start with PODs from your own drivers, which tend to be the most consistent, then add subcontractor documents and CMRs.
Frequently asked questions
Can AI read handwritten notes on PODs and CMRs?
Often, yes, but accuracy varies with handwriting and photo quality. Handwritten remarks matter for claims, so they should go to a person for confirmation until accuracy is proven on your own documents.
Does this work with photos taken on drivers' phones?
Yes, provided the photo is reasonably sharp and the whole document is in frame. Checking quality at the point of capture, in the driver app, saves the most time.
We are moving to electronic PODs. Is this still worth it?
Electronic PODs reduce the problem for your own fleet, but subcontractors, international partners and shipping lines will keep sending paper and PDFs for some time. Automation covers those.
Can it handle customs documents too?
The same approach works for commercial invoices, packing lists and other customs paperwork, but customs has its own rules and liabilities, so it is best treated as a separate project with its own checks.