AI automation for logistics companies: start at the quote desk

AI automation for logistics companies means putting agents on your own quote desk, your status inbox and your document pile, not buying another visibility platform and hoping the operation changes around it. It pays off when you move enough shipments that somebody spends most of a week retyping information that already exists. Start at the quote desk, because that is the delay your customers actually feel.

We have written about this from the shipper’s side before, in freight quote automation, which describes the agent our own export desk uses to send rate requests out to forwarders. This post is about the desk on the other end of that email.

The industry is piloting, not running

Boston Consulting Group, with Alpega, published a logistics AI survey on 27 March 2026 covering more than 180 experts, 84 of them at logistics service providers including 3PLs and ocean and air carriers, across Europe, North America, Asia-Pacific and the Middle East. About 40% of those providers had deployed AI beyond pilots. Only one in ten had AI embedded in core operations at scale, and 13% reported measurable value.

The regional split in that survey is wider than the headline. Providers with AI embedded in core operations: 31% in Asia-Pacific, 14% in North America, 6% in Europe.

So the competitive question is not whether your peers are talking about AI. Almost all of them are. It is whether the one in ten with it running are in your lane.

Your shippers are further along than you are

Descartes published its tenth annual transportation management study in September 2026, run with SAPIO Research across 600 senior transportation decision-makers globally, split evenly between shippers and logistics service providers. Of the providers, 15% said they use AI at scale. Of the shippers, 19% did.

Four points is not a dramatic gap. The direction is the problem. In a transaction where you are the specialist, your customer is automating their side faster than you are automating yours. That shows up as a shipper whose system can issue a structured rate request in seconds waiting two days for a human at your end to answer it.

The same study found 78% of respondents plan to increase transportation technology investment, up from 53% in 2017. Money is not the constraint. Descartes recorded the two biggest barriers as data quality at 45% and integration complexity at 38%, which are both operations problems wearing a technology costume.

Half the document problem is not yours to fix

The five UN Regional Commissions and UNCTAD published their sixth global trade facilitation report in 2025, assessing 174 countries against 62 measures. Implementation of domestic paperless trade reached 72%. Cross-border paperless trade, meaning the actual electronic exchange of documents between countries, reached 49%.

Read that as a ceiling on what any internal project can achieve. You can digitise your own side completely and still deal with a customs regime on the other end that wants a stamped original. The same report estimates full implementation of digital trade facilitation would cut average global trade costs by more than 11%, which is the prize nobody controls alone.

The practical consequence: build agents that produce a clean, complete document set fast, and accept that the last mile of some lanes stays manual. Automating up to the wall is still worth doing. Pretending the wall is not there is how a project gets written off.

Where the agents go, in order

Each piece here is a tomte, our word for one production agent with one defined job. We run this pattern across our own group’s export desk, so the sequence comes from operating it, not from a roadmap deck.

1. Inbound quote requests land in a monitored intake and become a structured shipment record: origin, destination, mode, dimensions, weight, commodity, incoterm, ready date. Missing fields are flagged as missing, never guessed.
2. Rate requests go out to your carrier and agent panel as one identical structured message, so the replies can be compared later.
3. Replies get read and normalised into one table: all-in landed cost, transit, validity, surcharges broken out. A quote missing a surcharge is marked incomplete rather than treated as cheapest.
4. Status questions get answered from the shipment record on a schedule, before the customer asks. Most “where is my shipment” email exists because nobody pushed the update.
5. The document set generates from that same shipment record instead of being retyped into four templates. We wrote the mechanics of that in export documentation automation.
6. Anything the agent cannot resolve goes to a review queue with a stated reason: unknown commodity code, no reply from two of five agents, validity expired. It is never sent to the customer.

That last item is the one operators underestimate. The exception queue becomes the actual job, and your coordinators move from typing to clearing items with reasons attached. That is a different role and a different training plan.

BCG’s March 2026 survey found nearly half of logistics service providers plan to deploy AI in pricing, quoting and customer service interactions. The quote desk being step one is not our opinion about where value hides. It is where the industry is already pointed.

	Hire another coordinator	A portal or TMS module	An agent with a review queue
Quote turnaround	Faster until they are busy	Faster only for lanes already in the rate table	Same speed at 10 or 200 requests
”Where is my shipment”	Answered when asked	Customer logs in, if they will	Pushed before the question
Panel replies	Compared by eye	Not handled at all	Normalised into one table
Volume spike	Overtime, then errors	Unchanged	Queue lengthens, nothing is dropped
What breaks first	The person	The integration	The exception queue, visibly

What nobody can honestly tell you

We went looking for a credible figure on freight invoice error rates, because it is the number every vendor in this category quotes at you. The claims in circulation run from 5% to 18% of invoices carrying a discrepancy, plus a popular line about losing 3% to 5% of freight spend to billing errors.

Every instance we could trace sits on a vendor blog or a search-engine landing page. The attributions are vague, no methodology is published, the ranges contradict each other, and no primary study exists behind any of them. The same is true of the widely repeated claim that only 23% of freight forwarders have digitised more than 75% of their processes.

If someone is selling you audit software on those numbers, they are quoting nothing. Ask for the sample size and the fieldwork date, and watch what comes back.

The labour argument, stated accurately

The IRU’s global driver shortage report, released in June 2026, counted 2.9 million unfilled truck driver positions across 18 markets, about 11% of the workforce. In Europe the rate is 13%, roughly 502,000 positions, 65% of European operators rank it their most pressing concern, and around 20% of Europe’s drivers retire by 2030.

Those are drivers. They are not your back office, and we are not going to pretend the figure transfers. No comparable published study measures vacancy rates for forwarder coordinators or customs clerks, and we looked.

Why it still matters at your desk: when capacity is structurally scarce, the margin moves to whoever books it fastest and documents it correctly the first time. Coordination speed becomes the product. That is an argument for automating the office even though the shortage is on the road.

What stays with a person

Rate negotiation stays. So does any customs classification where the treatment is arguable, every claim conversation, anything that creates a binding commitment, and the call where the news is bad.

Reliability at this scale is a design decision rather than a hope. Our own group runs a registered fleet: every agent has an owner, a defined fence it operates inside, and a run log, with a governance layer that retries failures and reports health daily. In one measured 13-day stretch that fleet ran 5,450 times with zero failures across 41 agents, which is the discipline described on our agent fleet proof page. An agent that emails a customer a quote it was not sure about is worse than a coordinator who asks.

Start async

If you run a forwarding or 3PL operation and want to know which of this is worth building against your actual lane mix and volume, that is what an AI Operations Audit is for: $1,900 fixed, 10 working days, a map of how work moves through your business today, and a ranked build list with effort and payback estimates. Full refund if we find nothing worth building. Build engagements start from $9,500.

Send us the shape of your desk: shipments a month, how quote requests arrive today, how many agents or carriers you fan a rate request out to, and who currently answers the status email. Start async and a written reply comes back within one business day, with scope and price. No call to book.