FreighAI
FreighAI Playground · Learn
Lesson 5 of 6 · 15 min read

What are other forwarders actually doing with AI?

Not what the market says. What twelve named companies have published. The publisher and the date sit beside each one.

The result is quoted in the source's own words. Where the number sits on a supplier's website, this page says so.

See the twelve, desk by deskGet your own number
The short answer

Twelve public examples, checked and dated. Most run one AI on one job: reading email, checking a photo, posting an invoice. Time saved is the usual claim, not fewer people. The strongest numbers sit on vendors' websites, so read those with care. Almost nothing is public from forwarders your size.

Lesson 5 of 6 · 15 min read

By the end of this lesson you can

  1. Name three companies and what they use AI for.
  2. Tell a company's own number from a vendor's number.
  3. Say why nothing public exists from firms your size, and what that means for you.
  4. Take the patterns into your own plan.

What kinds of company are doing this?

Three kinds of company show up in the public record. Big global forwarders: C.H. Robinson, Flexport, Kuehne+Nagel, DHL Supply Chain and Forto. Carriers and ports: Maersk, Hapag-Lloyd and DP World. Then smaller and regional firms: Aramex, Scan Global Logistics, Navia Freight and Masterpiece International. Now the honest part. Almost nothing is public from forwarders of 10 to 200 people. The two smallest names here are only visible because their software supplier wrote them up. So nobody can tell you what a firm your size gets. You have to measure that yourself.

Recommendation

Read this list for the shape of the job, not the size of the number. A global forwarder quoting in 32 seconds is not a promise to you. What travels is the kind of work they handed over. Reading email. Checking a photo. Posting an invoice.

What are they using it for, desk by desk?

Twelve examples, sorted by the desk they sit on. Every result is quoted word for word from the source. Where the source is the supplier rather than the company, the table says so.

The pricing desk

The pricing desk. Every result is quoted from the source.
  1. C.H. Robinson
    What it does
    More than 30 AI agents read emailed quote and capacity requests, price the quote, process the order, classify the freight and set the appointment.
    What they say came of it
    "delivers customer-specific price quotes in just 32 seconds"; "5,500 truckload orders per day", "saving 600 hours of labor daily"; "Instead of waiting up to four hours for a person to get to their shipment in an email queue, over 5,200 customers are getting their loads accepted in under 90 seconds".
    Source and date
    C.H. Robinson newsroom, 12 June 2025
  2. Flexport
    What it does
    A real-time door-to-door pricing engine, voice agents that call carriers, and an agent that audits past customs entries.
    What they say came of it
    "Early adopters of our Control Tower saw freight cost savings of 10% on average" (a company statement about its own customers).
    Source and date
    PR Newswire, Flexport release, 24 February 2025

The booking desk

The booking desk. Every result is quoted from the source.
  1. Forto
    What it does
    An assistant called Flash scans email for booking requests, pulls out the details, picks a vessel and files the documents to the shipment.
    What they say came of it
    No number stated ("often within minutes").
    Source and date
    Forto, undated, read 7 September 2026

The documents desk

The documents desk. Every result is quoted from the source.
  1. Kuehne+Nagel
    What it does
    AI checks proof-of-delivery photos from road carriers, rejects unreadable or invalid ones and asks for a new one.
    What they say came of it
    No number stated.
    Source and date
    Kuehne+Nagel newsroom, 26 March 2024
  2. Masterpiece International
    What it does
    AI processes supplier invoices, pre-alerts and customs paperwork across 12 branches.
    What they say came of it
    "save just under 1,000 hours of labor" monthly; "automating up to 93% of documents each month with zero to minimal user interaction" (the vendor's case study). The number comes from the vendor, not the company.
    Source and date
    Raft (vendor case study), 1 September 2026

The customer service desk

The customer service desk. Every result is quoted from the source.
  1. DHL Supply Chain
    What it does
    Voice and email agents schedule appointments, take status calls and coordinate with warehouses.
    What they say came of it
    Volume only: "hundreds of thousands of emails and millions of voice minutes annually". No outcome number stated.
    Source and date
    Truck News, 25 November 2025
  2. Hapag-Lloyd
    What it does
    AI reads customer feedback, sorts it by mood and theme, and writes a report every two weeks.
    What they say came of it
    "over 15,000 feedback items per month with an accuracy of 95% for sentiment classification"; analysis cut from "hours, sometimes days" to "seconds". The number comes from the vendor, not the company.
    Source and date
    AWS Machine Learning Blog, co-written by Hapag-Lloyd engineers, 5 May 2026
  3. Aramex
    What it does
    Conversational AI on WhatsApp and chat handles routine tracking and support questions in more than 65 countries.
    What they say came of it
    "Aramex has automated 90% of cases and saved over a million agent hours annually" (the vendor's release). The number comes from the vendor, not the company.
    Source and date
    Sprinklr newsroom (vendor release), 13 May 2025

The accounts desk

The accounts desk. Every result is quoted from the source.
  1. Navia Freight
    What it does
    AI reads supplier invoices and posts them in accounts payable.
    What they say came of it
    "Approximately 75% of invoices were processed automatically"; "35% of invoices required no human touch, while another 40% needed only minimal input" (the vendor's case study). The number comes from the vendor, not the company.
    Source and date
    Raft (vendor case study), 1 September 2026

The operations desk

The operations desk. Every result is quoted from the source.
  1. Maersk
    What it does
    On-board machine learning forecasts fuel burn and wind resistance so crews can adjust course and speed.
    What they say came of it
    "over 2.5 billion IoT data points in real time"; adjustments "within minutes". No efficiency number stated.
    Source and date
    Maersk, 10 September 2025
  2. DP World
    What it does
    An AI tool decides how containers are stacked at its Busan centre, so cranes sit idle less.
    What they say came of it
    No number stated.
    Source and date
    DP World, undated, rollout March 2024
  3. Scan Global Logistics
    What it does
    An AI monitor tracks vessels, predicts arrival times and flags trouble from weather and port congestion.
    What they say came of it
    Forecasts "15% more accurate"; predictions "up to 12 days in advance" (published by the majority owner).
    Source and date
    CVC, the company's majority owner, 13 January 2026

Look at how narrow each one is. One AI, one job, one desk. Only C.H. Robinson publishes a chain that runs from quote to order to appointment.

Four of the twelve state no number at all. That is worth noticing. A big company announces the tool long before it publishes what the tool did.

Which numbers can I trust?

A number is only as good as whoever published it. There are four kinds on this page, and they are not equal.

  • The company's own announcement. The firm says what its own software did. C.H. Robinson, Flexport, Kuehne+Nagel, Forto, Maersk and DP World sit here. Nobody knows the job better. Nobody has more reason to sound good.
  • Trade press. A reporter writes it up. DHL Supply Chain is the one here. A reporter has no product to sell, but the detail is usually thinner.
  • The owner's publication. The firm's parent company writes about the firm it owns. Scan Global Logistics is here. Better than a supplier. Softer than a reporter.
  • A vendor case study. The software seller writes up its own customer. Aramex, Navia Freight and Masterpiece International sit here. So does Hapag-Lloyd, whose post was written with its supplier. The number comes from the vendor, not the company.

Here is the awkward part. The biggest numbers on this page sit on suppliers' websites. 90% of cases. 75% of invoices. 93% of documents. Most of the companies' own announcements give no number at all.

That does not make the vendor numbers false. It means nobody outside the deal has checked them. Read them as a claim, and go and check.

Recommendation

Trust a number as far as you can trace it. Open the link. See who published it and when. Then ask that vendor for a customer you can telephone. One name you can call is worth three numbers on a website.

What does the one signed forwarder case we can name look like?

We have one customer we can name, and one signed reference behind the numbers. It is only fair to hold ours to the same test as everybody else's.

Infinity Logistics, a UAE freight forwarder and integrator key-account reseller, runs on FreighAI.

  • At Infinity Logistics, average response time fell from more than 60 minutes to about 15 minutes, and the win rate rose from 35% to 73%.
  • At Infinity Logistics, the quotation error rate fell from 67% to under 5%.
  • Infinity Logistics quoted 70% more shipments with the same five-person team, comparing Q4 2024 with Q4 2025.

Notice the shape of that last one. The team did not get smaller. It got through more work. That is the same shape as every public number on this page.

The caveat, every time

One company, one year, one signed reference. That is not a market average, and we do not present it as one. It is what one forwarder measured, signed on 15 May 2026. Our number lives under the same rule as the vendor numbers above. Ask us for a customer you can telephone.

What patterns should I take from this?

Five things run through all twelve. The patterns travel much better than the numbers do.

One process at a time, not the whole business

Almost every example runs one AI on one job. Kuehne+Nagel checks delivery photos. Hapag-Lloyd reads feedback. Navia Freight posts invoices. Only C.H. Robinson publishes a chain from end to end. Nobody bought a machine that runs the whole company.

Email and documents are where it starts

Look at where each one begins. An inbox. A PDF invoice. A photo of a delivery note. The work was already text, so software could read it. That is why the enquiry inbox is usually the first place this pays.

Time saved is the claim, not fewer people

Four hours down to 90 seconds. Hours down to seconds. 600 hours of labor a day. Not one source here claims it cut staff. Take that as the honest version of the claim, and the version you can check.

The best numbers come from vendors

Aramex at 90% of cases. Navia Freight at 75% of invoices. Masterpiece International at 93% of documents. All three sit on a supplier's site. So the loudest numbers are also the least checked.

Nothing public from small forwarders, so your own number matters more

Navia Freight and Masterpiece International are visible only because a vendor wrote them up. For a firm of 10 to 200 people, there is no public record. Nobody can hand you the number for a firm like yours. So measure one desk in your own office. That number beats everything on this page, because it is yours.

Recommendation

Copy the pattern, not the number. Pick one job that arrives as text, give it a month, and count the same thing before and after. If you quote by hand, start at the enquiry inbox.

Where do I go next?

Two of these explain. Two of them make you count. The counting ones are the ones that change anything.

The assessment asks about your own desks and gives you your number back. The benchmark exists because the market number on this page does not exist yet. Add yours and the picture gets better for everybody.

Questions

Common questions

01

Why is there no forwarder of my size on this page?

Because almost none has published anything. The two smallest firms here, Navia Freight and Masterpiece International, are visible only because their software supplier wrote them up. A firm of 10 to 200 people rarely writes a press release. The silence is not proof that nothing is happening.

02

Can I trust a number from a vendor case study?

Treat it as a claim, not a fact. The number comes from the vendor, not the company. Open the page, look at the date, then ask that vendor for a customer you can telephone. If nobody will take your call, the number is a slogan.

03

Are these companies replacing people with AI?

Not on the record. Every result quoted here is about time or volume: hours saved, seconds instead of hours, cases handled. No source on this page claims a cut in staff. Our one signed case is the same. The same five-person team quoted 70% more shipments.

04

How current is this list?

Every link was opened and checked on 7 September 2026. The dates run from March 2024 to September 2026. Each one sits beside its example. Read the date. A claim from 2024 tells you less than one from this year.

Sources
Previous
← Lesson 4 · Add-on or AI-native
Next
Lesson 6 · Map your own work →
READY WHEN YOU ARE

Copy the pattern, then go and get your own number

Nobody has published what a forwarder your size gets from this. So the number that counts most is yours. Pick one desk. Measure what it costs you today. Read the rest of the Playground while you wait.

All the Playground explainersThe AI impact assessment