AI in logistics and transport – where it helps hauliers and forwarders, and where to start

In Poland, the industry is often referred to as TSL – transport, freight forwarding and logistics. In a mid-sized company in this sector, time is lost mainly in three places: retyping orders, handling documents after unloading and answering calls about shipment status. That is where AI delivers results fastest.

The table below lists the uses from the simplest to the most demanding. For each one, we give the starting condition, because without it the project will stall at the testing stage.

Orders arrive by email with a PDF, from freight exchanges (online marketplaces for loads), in spreadsheets from regular customers and through the data exchange systems of large shippers. The dispatcher reads each one, enters the data into the transport management system – the program for managing orders, vehicles and billing – and confirms with the customer.

AI takes over the reading and data entry in the steps described below. The dispatcher receives a ready order with any doubts flagged and decides whether to accept it. Connecting to the transport management system requires an API – a link for exchanging data between programs – or a file import.

In many companies, the invoice for a transport job is issued only once the signed CMR consignment note comes back. The driver takes a photo, the document travels by email or in the cab, and someone in the office retypes the dates and remarks. Every day of delay means an invoice paid a day later.

This is where OCR helps – technology that turns a photo of a document into text and data. AI reads the CMR, links it to the order and checks whether the consignee has added remarks, e.g. about damaged or missing goods. Documents without remarks can go straight to invoicing, and the rest for clarification.

Purchase invoices are a different matter. Since 2026, most invoices from Polish subcontractors arrive through KSeF, the government e-invoicing system, as ready-to-use data that does not need reading. The exception is businesses whose invoices total no more than PLN 10,000 including VAT a month – they may issue invoices outside KSeF until the end of 2026. OCR is useful for invoices from foreign hauliers and for fuel and road toll bills from other countries.

Paper will gradually disappear. Poland is a party to the e-CMR protocol on the electronic consignment note, and from 9 July 2027, enforcement authorities in the EU must accept transport information from certified electronic platforms. This obligation applies to authorities, though, not to companies. A haulier can still use paper consignment notes, and many customers work exactly that way.

Take a hypothetical haulier with 60 tractor units handling 900 orders a month. Assume that processing documents after unloading takes 5 minutes per order today and 2 minutes with AI. The saving is 3 × 900 = 2,700 minutes, or 45 hours a month and 540 hours a year.

The bigger benefit is often cash rather than time. Assume the company has revenue of PLN 2 million a month, or about PLN 66,700 a day. If invoices go out 3 days earlier on average, about PLN 200,000 less is tied up in receivables.

This is an estimate, not a promise. Plug in your own data: the number of orders, the time spent on documents and the average time from unloading to invoicing. The last indicator is worth measuring regardless of AI, because many companies simply do not know it.

Good planning means combining orders to limit empty runs and waiting time. Dispatchers do this from memory and experience, often for several dozen vehicles at once. AI can review more combinations and suggest which vehicle is best placed to take a new order.

The suggestion must take into account the EU Mobility Package, the rules covering, among other things, drivers’ working and rest time and the mandatory return of drivers and vehicles to base. It must also know about constraints that are not in the system: a driver who does not go to a particular country, or a customer who only accepts deliveries in the morning.

That is why we recommend suggestion mode for at least several months. The dispatcher accepts or rejects the proposal, and the reason for rejection is recorded in the system. Only after such a period can you see whether AI really reduces empty kilometres.

A large share of calls to dispatchers are the questions “where is the truck?” and “what time will it be at the unloading point?”. AI can answer them automatically, by email or in a customer portal, based on the GPS position and the order status. The dispatcher then deals only with delays and problems.

The other side is delivery slot booking at the consignee’s warehouse. When AI sees that a vehicle will not make the booked unloading time, it can prepare a request to change the slot in advance. This will not shorten queues at every warehouse, but it will reduce the number of missed slots and extra charges.

In both cases, the condition is up-to-date data. If drivers do not update statuses in the app, or GPS is not connected to the transport management system, AI will answer customers based on outdated information.

Many management boards of transport companies receive results several days late and only for the company as a whole. Yet the key questions are: on which lanes – regular routes such as Poznań–Hamburg – do we make money, which customers cause empty runs, and which vehicles are the most expensive to run?

Power BI connected to the transport management system, GPS and accounting shows this in one report. AI can add a short explanation of changes or answer questions asked in plain language, e.g. “which lanes were least profitable last quarter?”. The condition is a single, consistent definition of vehicle cost, agreed with accounting before the report is built.

The amended Polish National Cybersecurity System Act (KSC Act), which implements NIS2, has been in force since 3 April 2026. In the road transport sector, however, it covers traffic management authorities and operators of intelligent transport systems, not hauliers and forwarders. Medium-sized and large postal and courier operators may be covered, as may companies in the rail, maritime and aviation sectors.

A haulier that is not covered by the Act may still receive similar requirements from customers. Large manufacturing and trading companies covered by NIS2 must ensure their suppliers’ security and write this into contracts. It is worth knowing these requirements before a tender, not after it.

ETS2, the EU system of charges for emissions from fuels used in road transport and buildings, will start in 2028 under the current rules, not in 2027. The Council of the EU approved the change on 5 March 2026. Until then there is no separate ETS2 cost in the price of fuel, but it is worth collecting fuel consumption data per vehicle and lane now.

The safest start is one area, a few regular customers and a measure of success agreed before the pilot. For documents, this might be the time from unloading to invoice; for orders, the share of drafts approved without corrections.

The mistakes below are the most common, and each of them can stop a project.