Services
Solutions
Use cases
Industries
Tools
Pricing
Book a call
Industry — Logistics & freight forwarding

AI in logistics and freight forwarding

Freight papers, requests, shipment tracking and damage handling — the admin layer of logistics is the real lever for . Five concrete , with a clear dividing line to classic OR tools that mostly remain better in dispatch.

The daily reality AI has to fit into

Dispatch at 6:30 in the morning, goods receipt after goods receipt, a driver calls in sick at seven, the tour is reshuffled. Freight notes as photos from the driver, CMR sections with stamp and signature in the gallery, a damage report comes in parallel in the catch-all inbox. The phone rings because a customer wants to know where their shipment is right now.

in logistics in 2026 is above all a tool for the admin layer: receipts, requests, shipment tracking, damage handling, correspondence. Real dispatch keeps running on classic tour-planning algorithms — adds a briefing layer, doesn't replace them.

Prerequisite: plugs into the forwarding software, the WMS and the telematics. Customs and dangerous-goods duties remain unchanged in human hands — can prepare, it doesn't carry liability.

Five places where AI in freight forwarding really makes sense

For each case: the situation today, where plugs in, a pointer to the matching setup tiers, and an honest trade-off.

01. Freight-paper OCR and booking

Situation today

Delivery note, CMR freight note, customs declaration, dangerous-goods transport document — three to six documents per shipment, often hand-annotated, often as a photo from the driver via messenger. Billing waits because the inbox is still being sorted.

Where AI helps

extracts recipient, sender, volume, weight, dangerous-goods class from receipts, matches against master data in the forwarding software and creates a booking suggestion. Dispatcher or billing signs off.

What it can't do

ATLAS correctness, ADR mandatory fields, customs tariffing remain in human hands. A wrong tariff number in ATLAS has consequences an confidence score doesn't offset.

02. Shipment-tracking bot for customer requests

Situation today

“Where is my shipment?” is the most common request in customer service. The phone rings, the dispatcher interrupts real dispatch, searches in the TMS and names an estimate. With twenty calls a day, both sides lose time.

Where AI helps

A bot answers standard requests based on TMS statuses (in transit, at the hub, out for delivery, delivered) — with a concrete time window when available. On exceptions or anomalies, it immediately escalates to the dispatcher.

What it can't do

Delivery commitments beyond the current status aren't given by the bot. When a dispatcher commits, it counts — when the bot does, it's a hope.

03. Incoming mail routing across sales, dispatch, claims

Situation today

Through the day, RFQs, advices, complaints, damage reports, driver queries and billing questions land in the catch-all inbox. An assistant spends four hours a week sorting just the intake — and the urgent damage case from Thursday gets noticed on Monday.

Where AI helps

classifies incoming mails (RFQ, advice, complaint, damage, billing, driver query), routes them on and suggests a preparation. For damage cases with photo attachments, the claim file is automatically prepared.

What it can't do

No auto-commitments to customers, no auto-delivery promises. The bot classifies and prepares, the dispatcher or claims clerk decides.

04. Claims and complaint handling with photo and waybill

Situation today

Damage reported: photo of the dented pallet, waybill as a PDF, recipient mail with the back story. By the time the claims clerk has gathered everything and writes a report for the insurer, an hour passes — plus the TMS research.

Where AI helps

pulls photo, waybill and mail into a structured claim file — with shipment history from the TMS, damage classification and a draft report for the insurer. The clerk reviews, supplements, signs off.

What it can't do

Insurance approval and claim assessment remain in human hands. prepares the file, it doesn't settle it.

05. Tour planning as a briefing, not auto-dispatch

Situation today

Dispatch with twenty tours per day, classic algorithms in the forwarding software deliver route suggestions, but the dispatcher knows the driver who prefers the north-east, and the customer who only accepts on Thursdays. Pure algorithm results get manually overridden four times a day in practice.

Where AI helps

combines algorithm suggestions with briefings from tour history and customer notes — “Note: customer X only accepts from 2 pm, driver Y prefers this region”. The dispatcher gets a justified suggestion, not a black-box tour.

What it can't do

Real dispatch stays with the dispatcher. as a briefing layer on top of existing OR tools — not as a replacement.

What in logistics isn't (yet) working

Four promises that usually turn more expensive in freight forwarding than their benefit justifies:

Fully automatic dispatch via LLM

Classic operations-research algorithms (VRP, MILP-based tour planning) beat in route optimisation across practically every measurable discipline. as a plan generator causes more problems here than it solves — as a briefing layer on top of OR tools is more sensible.

Predictive maintenance on vehicles without telematics history

Without a stable telematics data base, predicting vehicle failures is tea-leaf reading. Manufacturer service or established fleet telematics deliver more reliable results here than an layer on top.

Auto-replies on customer delivery commitments

If the bot says “your shipment will arrive tomorrow before 10 am” and it doesn't happen, the statement still counts under liability law. Delivery commitments are dispatcher responsibility, not bot output.

AI tariffing without human review

Customs tariff numbers and ADR classes are strictly regulated. A wrong tariff number costs penalty tariffs and reputation with customs — auto-handover to ATLAS without sight check is excluded here.

What needs to be thought through for AI in freight forwarding

Four pillars against which every forwarder setup is checked:

Customs code and ATLAS

Electronic customs declaration via ATLAS has its own formats and mandatory fields. workflows can handle the preparation — the submission to ATLAS belongs in user hands with clear responsibility.

ADR and dangerous-goods law

For dangerous goods, ADR, IMDG, IATA apply with clear labelling and documentation duties. can extract dangerous-goods classes from receipts, but the mandatory-field check remains human responsibility with a training record (dangerous-goods officer).

GoBD and BAG requirements

Receipts and toll registers must be archived audit-proof and accessibly. AI-supported booking workflows need audit trail and versioning — otherwise a receipt pile emerges that gets expensive at the next audit.

GDPR on driver and customer data

Telematics data of drivers, route profiles, customer master data — all personal data. DPA with the vendor, EU region and separation of data buckets are mandatory setup, especially in telematics analysis.

Tools that already run in freight forwarding

doesn't replace these systems — it plugs in. Where the interface typically sits:

Forwarding and TMS software

CarLo (Soloplan), Active Logistics, ondot, TIMOCOM Smart Cockpit, Catkin — typical plug-in point for is the reporting interface or the email gateway

Warehouse management (WMS)

SAP EWM, AEB, vlogic, viadat — typically plugs in for classification of incoming receipts or advices, not into real-time warehouse logic

Telematics and fleet

TomTom Telematics, MAN TeleMatics, idem telematics, FleetBoard — data source for briefings, not for control

Customs and special interfaces

ATLAS software (e.g. dakosy, AEB ASSIST4), freight exchanges (TIMOCOM, Trans.eu) — specialised interfaces with clear responsibilities

How forwarders typically get started

Anyone starting without experience has two clear candidates — and an area where the temptation of auto-dispatch should be ignored for a long time.

Typical entry point 1 — freight-paper OCR

Immediate relief in billing and dispatch, clear success metric (hours per week, lead time). Risk limited by human approval before booking.

Typical entry point 2 — shipment-tracking bot

Big effect on the dispatcher's day (fewer phone calls), risk low if the bot only reports status and escalates on anomalies.

Don't start with dispatch

Auto-dispatch is tempting and expensive. Classic OR tools with an briefing layer are more results-oriented than AI-driven plan generation. First clean up receipts and requests, then see what dispatch really needs.

Funding in logistics and freight forwarding

consulting funding applies to freight forwarding and logistics SMBs as in other industries. Alongside, there are BAG programmes (Federal Office for Logistics and Mobility) for investment measures, “go-digital” and in some states their own digital-bonus programmes. For innovative projects, ZIM or industry-related BMDV funding comes into play.

→ Details on BAFA funding
FAQ

Frequently asked questions about AI in logistics

Proven: freight document and delivery note capture (CMR, PDF, photo), route optimization based on existing orders, pre-classification of customer inquiries, automatic shipment tracking. For real-time dispatch we recommend augmented intelligence (proposal, dispatcher decides) over full automation.
Yes. CarLo has a REST , SAP TM OData, COMLOGIS offers data export, WinSped has CSV and EDI interfaces. We build pre-processing steps that feed structured data into the TMS — the system remains the leading system.
Pre-classification of customs tariff numbers (HS code suggestion from product description) works well. The ATLAS submission itself remains a regulated activity — we prepare data, the EDI module or your customs broker submits. does not file a customs declaration standalone.
Telematics APIs (TomTom, Trimble, Fleetboard) and WMS events can be used for anomaly detection — late tour, unusual stand times, route deviations. We build alerting on business logic, not generic ML black boxes. The dispatcher sees "why, what to do" instead of just "anomaly detected".
Often: TMS good, telematics partial, CRM thin, ESG/reporting missing. We recommend starting with incoming documents and dispatch assistance — that is where data quality is cleanest and the lever is immediately tangible. Scaling toward customer self-service and ESG reporting comes later.

Ready for the next step?

Free intro call, no strings attached. In 30 minutes you'll know whether and how AI can help your business.

Book a callCheck eligibility
Book a callBAFA funding