What is an AI Transport Management System?

An AI Transport Management System is a TMS that uses AI agents to automate the document-heavy, repetitive parts of forwarder operations, reading booking emails, extracting Bills of Lading and CMR waybills, preparing customs declarations, pushing proactive track-and-trace updates, without requiring operators to re-key data between inbox, portal and TMS.

Do you need to replace your existing TMS to get AI benefits?

No. The modern pattern is an AI agent that wraps around the existing TMS (CargoWise, Softpak, Descartes, Boltrics 3PL Dynamics) rather than replacing it. Migrating a TMS is an 18-month project with real operational risk; layering an AI agent takes 14 days with no data migration.

How is AI different from RPA or OCR in a TMS context?

RPA automates clicks on fixed UIs, it breaks when the UI changes. OCR reads text from an image, it fails when layouts change. AI agents understand the meaning of a document or email. They work across carrier formats, handle exceptions, adapt to new layouts, and decide which action to take. For a forwarder backoffice that deals with 50+ carrier formats and constant layout changes, AI is the only approach that scales.

What does an AI TMS cost compared to a traditional TMS?

A traditional TMS like CargoWise or Softpak typically costs €500-2,000 per user per month, with implementation projects running €50,000-€500,000. An AI agent layer sits on top and is priced on volume (documents or emails processed), typically €300-1,500 per month for a small-to-mid-size forwarder, with a 14-day setup and no migration.

Learn · TMS & AI

AI Transport Management System

A practical guide to what an "AI TMS" actually means in 2026, how agentic AI layers on top of CargoWise, Softpak, Descartes and Boltrics, what it automates, what it does not replace, and how to evaluate vendors without falling for buzzwords.

Traditional TMS: what it is and what it does not do

A Transport Management System is the operational system of record for a freight forwarder or carrier. It stores shipments, bookings, rates, carrier partners, documents and invoices. The dominant products in ocean and multimodal freight are WiseTech CargoWise (global market leader), Softpak (Dutch, customs-strong), Descartes (global, with strong customs and rate products) and Boltrics 3PL Dynamics (Microsoft Dynamics-based, strong in Benelux 3PL).

What a traditional TMS does well: structured data, workflow enforcement, financial integration. What it does not do well: read unstructured inputs. Every booking email that arrives still needs a human to open it, interpret it and type the result into the TMS. Every Bill of Lading, CMR, customs document, carrier invoice or pre-alert has to be looked at by a person. The TMS is the destination, not the processor.

What "AI TMS" actually means

There is no single definition on the market. In practice three different things get labelled as AI TMS:

Built-in AI features inside a traditional TMS, rate prediction, ETA calculation, anomaly detection on the TMS's own data. Incremental.
AI-native TMS replacements, new products (Flexport, Gofreight) that build an entire TMS from scratch with AI at the core. Big architectural commitment, full migration.
Agentic AI layers, AI agents that sit on top of an existing TMS, reading emails and documents on one side and writing structured data into the TMS on the other. No migration.

For an established forwarder with 500k+ historical shipments in a TMS, the agentic-layer pattern is usually the only economically sane option. Migrating a TMS is an 18-month programme; adding an AI agent is 14 days.

What agentic AI actually automates in a forwarder stack

Booking emails → TMS shipments

Read incoming booking emails, extract shipment data, create the shipment in the TMS, reply with booking confirmation. Email automation.

B/L & CMR extraction

Read Bills of Lading and CMR waybills regardless of carrier format, extract containers, weights, consignees, write into TMS.

Customs preparation

Prepare T1, MRN, ICS2, EMCS from booking and B/L data. Push to Portbase and national customs systems.

Track & trace and notifications

Poll carrier APIs, compute real ETAs, push proactive updates to customers via email or WhatsApp.

Carrier invoice three-way matching

Match incoming carrier invoices against TMS bookings and agreed tariffs, flag overcharges and unexpected detention. Invoice automation.

Quote generation

Draft quotes automatically from RFQ emails against carrier rate cards, push to sender in minutes.

Why AI ≠ RPA ≠ OCR

The three approaches are often conflated in vendor marketing. They are not the same thing:

RPA (Robotic Process Automation), scripts that automate clicks on fixed user interfaces. Works if the UI never changes. Breaks when carrier portals update, new layouts ship, or any variation is introduced.
OCR (Optical Character Recognition), extracts text from images. Needs per-layout templates. Fails on handwritten reservations, variable layouts and new carriers.
AI agents, understand the meaning of a document or email. Work across carrier formats, handle exceptions, adapt to new layouts, and decide which action to take. This is the only approach that scales to the 50+ carrier formats a typical forwarder sees in production.

Evaluating an AI TMS vendor, 7 questions to ask

Does it replace my TMS or sit on top? The answer determines whether this is a 14-day project or an 18-month one.
Which TMS do you integrate with natively? If your TMS is not on the list, implementation cost rises sharply.
How do exceptions get to humans? Every AI agent produces exceptions. The quality of the exception-handling UI is where real operational time is spent.
Where does the data live? EU-resident processing matters for GDPR and AEO audits. Non-EU processing creates paperwork you may not want.
Can I get a named AEO / FENEX-compliant DPA? If no, walk.
What happens when you change your model? AI agents improve. You want a change log and the ability to pin versions for critical flows.
What is the pricing unit, per user, per document, per shipment? Forwarder economics tilt heavily depending on which unit the vendor bills.