In logistics, time is lost on documents, not on the road.
The truck has GPS and the warehouse has scanners. But the order, the CMR, the customs paperwork, the customer status updates and the invoice still move through email, phone and spreadsheets — and through the hands of people retyping them from one system into another. That is where AI gives back the most time.
Book a call (30 min) →Every shipment carries a dozen documents.
A single shipment passes through the customer’s order, the carrier’s quote, transport and customs documents, proof of delivery and the invoice. Each document arrives in a different format, and someone opens, reads and retypes each one.
At hundreds of shipments a month, this is work that grows faster than volume — because every new customer brings a new format to learn.
Order intake
Orders by email, PDF or spreadsheet, in each customer’s own format, translated into the transport management system by hand.
Documentation
Extracting data from transport, customs and supplier documents, and checking they agree with each other and with the order.
Customer status
Answering “where is my shipment” — the same data, looked up and sent by hand, several times a day.
Invoicing
Reconciling completed transports against carrier invoices and billing the customer, with surcharges that are easy to miss.
Measure first, then automate.
Diagnose takes two to three weeks. We follow a shipment from order to invoice, talk to dispatchers, operations and finance, and inventory the systems and formats the data passes through.
The output is a workflow map, a systems map, and three to five automation opportunities with estimated ROI. In Design we decide what to build, what to buy, and in what order.
Reading documents is a job for AI.
This is where language models genuinely add something new: documents that never arrive in the same format. The model reads the order or document as it arrived, extracts the data and writes it to the system, and a person checks only what does not match.
Every such system is evaluated on real documents from your own operation before launch — so you know how far to trust it before you hand it the work.
On the systems you already have.
Automation is built around the transport management system, ERP and email you already use — without replacing them. If data must not leave the company, the solution runs on your own infrastructure.
Frequently asked.
- How is AI used in logistics?
- Mostly in handling documents and communication: taking in orders in different formats, extracting data from transport and customs documents, shipment status updates for customers, and reconciling carrier invoices. Route optimisation and planning are a separate field; we work on the tasks people in operations and finance do by hand.
- Can AI read documents in different formats?
- Yes, and it is one of the areas where language models genuinely add something new. The model reads the document as it arrived — email, PDF, scan, spreadsheet — and extracts the data. Before launch the system is checked against real documents from your operation, and a person reviews anything that does not match.
- Which process should a logistics company automate first?
- The one that repeats most often and has the most predictable outcome. In logistics that is often order intake and reconciling carrier invoices against completed transports. But that is a measurement finding, not a rule — Diagnose shows what is most expensive in your case.
- How long does it take, and what does it cost?
- Diagnose takes two to three weeks, Design one week, and building one or two clearly defined solutions two to four weeks. Every engagement has a fixed scope and a price agreed before work starts. A Pilot on one process takes six weeks, and if the written success criterion is missed, the design and delivery fee is refunded.
Start here
Start with one document.
On a thirty-minute call we walk one shipment from order to invoice and the systems it passes through. From that we propose a scope — Diagnose or a Pilot on one process. If the call shows it will not pay off for you, we will say so.
Book a call (30 min) →or write to ai@meteorit.rs