XONTAINERLogistics ecosystem
What does a TMS do, and what does AI add?
The core of a TMS is creating trips, assigning vehicles and drivers, following arrival at stops, collecting documents and preparing for invoicing. If that data model is weak, adding AI makes the problem larger. On top of the core, AI creates value in four places:
- Arrival forecast: a likely arrival time that takes driving and rest rules into account.
- Document intake: linking the documents a driver sends by WhatsApp or email to the trip file.
- Language: translation between the driver and the person on the other side, and one-click translation of incoming email in the office.
- Warnings: early flags for situations such as cabotage risk or a long wait at customs.
How is a realistic arrival time calculated?
A forecast that looks only at distance and traffic ignores legal driving times and gives the wrong answer for a truck. For a truck trip, arrival depends on how much further the driver may legally drive under tachograph rules. See the guide to arrival time under tachograph rules for detail.
The customs and border stop is part of the trip
On an international trip the biggest uncertainty is often at the border. Following the customs stop and the waiting time at customs inside the trip directly improves the arrival forecast and customer updates. For the document side see the e-CMR guide and the eFTI guide.
What to check in an AI-assisted TMS
These questions separate one system from another when you buy or set one up:
- What is the forecast based on? Tachograph rules, position and documents, or only an average duration?
- Does a suggestion go to a person, or is it applied directly?
- Is there driver or person scoring? A trustworthy system does not score people; it follows status and process.
- Where does the data sit, and can another company see it?
- Does it connect to existing ERP, accounting, GPS and tachograph systems without removing them?
From the driver’s side
A mobile app that does not make the driver’s work easier will not be used in the field. Sending a photo of a document, reporting arrival at a stop and being able to speak with the person on the other side in their own language are the driver’s real needs. In translation mode, speech is turned into text on the phone, the AI translates the text and the voice is produced on the phone; audio is not sent to the server.
Common mistakes
- Giving the customer an ETA based only on distance and traffic as if it were certain; it usually fails because the truck’s legal rest is not counted.
- Adding AI before putting trip data in order: missing or late data misleads even a strong model.
- Choosing an app that does not make the driver’s work easier; if the field does not use it, no data arrives.
- Taking the planner out and applying suggestions directly; suggestions first, then partial automation at a measured accuracy, is healthier.