Where freight forwarders lose hours every day, and which processes are worth automating

Very little of the office time a freight forwarder loses goes into actual planning. Most of it goes into copying information that already exists in digital form. A transport order arrives as a PDF attachment and someone keys it into the TMS. The signed CMR comes back as a phone photo in the drivers' group chat, and someone in the office works out which shipment it belongs to. Two hours of waiting at a loading dock get mentioned in a voice message and never appear on the invoice. So if you are looking at freight forwarding process automation, start by counting something dull: how many times, during a single order, does a person carry the same piece of information from one system into another?

Freight forwarding process automation starts with the hand-offs

In most forwarders an order passes through three stations: intake, dispatch and billing. Each has a person reading information from one channel and typing it into another system. There is nothing wrong with that as an organisation. It gets expensive through volume, and because a typo at intake usually surfaces two stations later, when the driver pulls up at the wrong depot or the customer sends the invoice back.

You don't need a consultant's slide deck to put a figure on it. Take the number of orders per day, multiply by the minutes it takes to key in and check each one, and scale that up to working days per year. Suppose, as an illustration, a forwarder receives 60 orders a day by email and each takes five minutes. That is five hours a day, more than half a full-time role doing nothing but data entry. Those are example numbers. Yours are sitting in your inbox and your TMS, and a week of tally marks is enough to find them.

Order intake: every customer has their own format

Large shippers often send orders via EDI or a portal, and in many companies those already flow cleanly into the system. The work comes from everyone else. A regular customer emails a spreadsheet, a manufacturer sends a PDF generated by its own ERP, a small customer describes in three sentences what needs collecting and from where. All of them contain the same details, just in different places under different labels: pickup and delivery address, time window, weight, loading metres or pallet count, the customer's reference number, notes on dangerous goods or pallet exchange.

The typing is the smaller part. The gaps take longer. A missing time window means a phone call. A delivery address given only as a company name means someone in dispatch looks up the right site. And when a quantity change arrives as a reply to the original email, somebody has to notice that the order is already in the system and must not be created twice. None of this is in anyone's job description, and it fills every morning anyway.

This is where an AI agent pays off most clearly. It reads the email and attachment, finds the details regardless of format, checks customer and addresses against your master data, and creates the order as a draft in the TMS. Anything missing or contradictory gets flagged, with a query to the customer already drafted. Dispatch reviews and approves. Retyping turns into a check that usually takes a fraction of the time. Unclear orders still need a person, which is fine: those are exactly the cases where you want someone who knows the customer.

Route planning is the wrong place to start

Ask what a forwarder should automate and route planning tends to come up first. We think it is the weakest starting point. Route optimisation software has been around for years and it is good at calculating distances. What it lacks is what the dispatcher knows without looking it up: that one customer only accepts deliveries until 2 pm even though the master record says 4 pm, that a particular dock runs an hour behind on Fridays, which driver holds an ADR certificate and which truck has a tail lift.

As long as those rules are not written down anywhere, an agent plans on wrong assumptions too, just faster. And every plan depends on the orders feeding it. An order without a time window or with the wrong weight ruins the plan whoever, or whatever, draws it up.

Where an agent can help in dispatch is the preparation: pre-sorting orders by region and time window, checking order restrictions against vehicle and driver, flagging conflicts before the route is fixed. Which route actually runs stays the dispatcher's call, and for most forwarders that is the right split.

Billing: the invoice is waiting for a photo

Most forwarders only invoice once proof of delivery is in, meaning the signed consignment note or the signature captured in the driver app. If the proof comes as a photo, someone has to match it to the order. If it doesn't come at all because the paper is still in the cab, the order stays open, and so does the invoice. Meanwhile the money sits with the customer.

Then there are the charges that were never part of the original order: waiting time at the dock, an extra stop, pallets that were or were not exchanged. That information is created on the road, by the driver, and reaches accounts by phone, by chat or not at all. What doesn't arrive doesn't get billed, and nobody ever sees it as an error.

With subcontracted hauliers the same flow runs in reverse. Their jobs have to be checked and settled by self-billing credit note, and at month end everything goes to financial accounting, which in many forwarders is when the correcting entries start. An agent can match delivery proofs to the right order, often even when the photo is crooked, suggest extra charges from driver messages as invoice lines, and hand the finished booking data to DATEV or your ERP. Whether to push the waiting time charge onto the customer remains a human decision.

Why intake comes first

All three stations run on the same data. Whatever is captured wrongly or incompletely at intake costs time in dispatch and goes missing in billing. An order that lands in the system complete and consistent helps every later step: route planning gets time windows and weights it can trust, and the invoice carries a reference number the customer recognises.

Intake has a second advantage, which is that the risk stays small. An agent that creates drafts and leaves approval to a person produces, at worst, a wrong draft that someone corrects. An agent that plans routes on its own can send a truck to the wrong dock at the wrong time.

Sophera Consulting builds this part as an agent: it takes transport orders from email and customer portals, creates them in your TMS and, after delivery, passes the billing data to accounting, at a fixed price and without a subscription. Which of the three stations costs you most is something we work out beforehand in the free automation check.

What we recommend

Keep a tally for one week before anyone talks about software. Three columns are enough: how many orders arrive through which channel, how often someone has to chase missing details, and how many days pass between delivery and invoice. If emails and queries dominate, start with order intake. If the gap between delivery and invoice regularly runs to more than a few days, billing is the more expensive problem, because that is where money arrives late or not at all. Route planning comes after that, once the data it depends on is right.