AI in logistics and transport

Why AI in transport starts in the forwarding office, not in the cab

AI is sold to transport companies from the fleet side: route optimisation, delay prediction, driving-style analysis. Those projects can be valuable, but they need different data from office work. Ready-made route-planning tools need current data on orders, vehicles and time windows, while a custom delay-prediction model needs journey history joined to the orders. Where that history does not exist, the project starts with data collection, and when the first predictions arrive depends on how much data has to be collected.

Meanwhile, in the same company, a stream of work flows every day that AI can take over now: transport orders arriving by email and from freight exchanges in a dozen formats, retyped by hand into the TMS (transport management system); CMR scans and delivery documents collected from drivers and attached to orders before invoicing; delivery notifications and status queries answered by hand from the same templates; subcontractor invoices checked against rates one by one. These are document- and text-based processes — the data exists, the rules can be written down, and a mistake is detectable and reversible.

That is why a good first AI project in a transport company looks unspectacular: it does not touch the vehicles, it touches the flow of information around them. In return, the first results appear within weeks, which helps fund the next steps.

What we actually automate in transport companies

• Transport order handling. Reading orders arriving by email, as PDFs and from freight exchanges — in layouts nobody has described before — and preparing the TMS entry for a forwarder to approve. This is the most common first process: it recurs daily, and order confirmation adds a second checkpoint. In the pilot we measure both the accuracy of the entries and the time needed to check them.
• Consignment documents and CMR. Attaching CMR scans and photos, pallet notes and delivery documents to the right orders, checking completeness before invoicing, and a list of gaps with reminders drafted. It shortens the stretch where today the longest wait is not for the haul but for the paperwork after it.
• Delivery notifications and status queries. Drafting notifications and replies to status questions from the order data — for a forwarder to approve before anything goes out to the customer. Repetitive work that today pulls dispatchers away from more important tasks.
• Subcontractor settlements. Checking carriers’ invoices against orders and agreed rates, with a list of discrepancies to explain — instead of comparing two lists by hand. Ambiguous cases go to a person, not straight to posting.
• Driver communication. Consolidating reports and documents sent in by drivers into one register, with reminders for what is missing — and an assistant that answers recurring procedural questions from the internal instructions, citing its source.

The common denominator: a person approves the output, the system prepares it. Which processes qualify for this kind of automation at all — the five conditions a good candidate must meet — is set out on our AI automation page; in a forwarding office those conditions are checked in exactly the same way.

What we do not promise

• A model trained on the fleet’s journey history, without suitable data. If routes are to be planned by a model trained on your fleet’s operations, it learns from journeys that have already happened. If the fleet does not record times, routes and stops in a form that can be joined to the orders, we start with data collection — and say plainly that the result comes later. We assess other ways of planning routes, using current data and agreed constraints, separately.
• Delay prediction “on the side”. A custom model that predicts delays from your fleet’s operations needs suitable journey history. Before it starts, we assess data availability, how to obtain the data and the cost — it is a separate project with its own budget, not an add-on to document automation. Ready-made tools that estimate arrival times (ETAs) have different requirements, depending on the solution.
• Unattended handling from day one. Everything that goes out to customers and subcontractors first passes through a stage in which a person approves every result. The decision to loosen oversight is made on pilot data, not by assumption.

How we work

As in every project: we pick one process and measure one indicator before we start, for example minutes from an order arriving to its TMS entry. Then we run the smallest possible pilot. Before switching over, the system runs alongside the current way of working for a while. The system prepares, a person approves every output. The full course — four phases and what you receive at the end of each — is set out in our implementation methodology, and the full scope of the service, from process audit through to upkeep, on our AI implementation page.

A separate subject that comes up in every transport conversation is data: rates, customer data, driver data — including location data where telematics is involved. Where data goes when we integrate an AI model, and what to ask any supplier, is set out on our security and GDPR page.

What it costs

Three things drive the cost: the number of systems to connect, the state of the data, and how reliable the results must be, that is, how rarely they may be wrong. Fleet size barely matters — what matters is whether the orders live in one TMS or in the dispatchers’ mailboxes and spreadsheets beside it. All the components, upkeep included, are broken down in our note on what an AI implementation costs.

Frequently asked questions

Where should a transport company start with AI?

In the office, not in the fleet. Handling transport orders, assembling consignment documents, delivery notifications and subcontractor settlements are document- and text-based processes — the data already exists in mailboxes and the TMS, a mistake is reversible, and results show in weeks. A custom model that predicts delays from your fleet’s operations needs suitable journey history. Its availability and the work needed to prepare it affect the project’s scope, cost and timetable; we assess these before starting. Starting with such projects is the most common way transport companies spend their budget before seeing a first result.

Will AI optimise our routes and predict delays?

It depends on what you need. Route planning usually works from addresses, distances, vehicle capacities and delivery time windows, and arrival times can be estimated from the traffic data that map services hold. That does not always require your fleet’s journey history — the requirements depend on the chosen tool, and we check whether a ready-made solution is enough and whether it can be connected to your TMS. A custom model that predicts delays from your own fleet’s operations is a separate piece of work: it needs suitable journey history — times, routes, stops — joined to the orders. If the fleet does not collect such data, or the data cannot be exported, we start by putting data collection in order and say plainly that the predictions come later.

Will this work with our TMS?

The audit at the start of a project settles that. We check whether the API (the interface other software uses to connect to the system) or file import gives access to the data and actions the chosen process needs. On that basis we confirm whether the integration is feasible, what it covers and what it costs — and whether an existing system feature would be enough. The TMS systems in use vary widely here, and that difference affects the cost — and so whether the project pays off. More important than the system itself is how orders reach it: email, freight exchanges, customer portals — because that stretch of manual retyping is what we automate first.

What does an AI implementation cost in a transport company?

Cost is driven by the number of systems to connect — TMS, email, freight exchanges, customer systems — the state of the data, and how reliable the results must be, that is, how rarely they may be wrong. Fleet size barely matters: a forwarding office with one TMS and orders arriving through one channel is a matter of weeks; the same work scattered across dispatchers’ mailboxes and spreadsheets is a different project and a different quote. Upkeep is a separate line: per-operation fees and periodic retesting after model version changes. We break the components down in our note on what an AI implementation costs.

What about driver and customer data?

Orders and consignment documents contain personal data — of drivers and of contact people at customers and consignees — and on-board systems additionally produce employee location data, which is subject to stricter rules of its own. That has to be settled before the integration, not after it: which data leaves the company at all, where it goes and on what basis — from the choice of provider and processing region to limiting the data to the minimum the step needs. The principles, and the questions worth asking any supplier, are set out on our security and GDPR page.