AI in logistics starts with the handoffs

Follow one load through a carrier or 3PL and count how many times someone types the same information. The order arrives by email and goes into the TMS. Dispatch copies it into a plan. The driver's paperwork comes back and someone keys it into billing. Then the same weights and BOL numbers go into payroll, and again into the fuel tax return.

Gary Graham Transport, a carrier in Ontario, was keying the same data three to four times across payroll, invoicing, fuel tax and accounting, and preparing the fuel tax return took three days ¹. The office team had absorbed all of it as part of the job, the way most teams do.

That pattern runs through the 208 published logistics and transportation case studies we reviewed. Most of the work companies chose to redesign sat at the handoffs, where information moves between customers, carriers, systems and people faster than the process was built to handle.

Where the time goes in a logistics day

The day starts in the inbox. Quote requests arrive as emails and PDFs, and a rep copies lanes, weights and dates into another screen. At ShipNova, building one load took 20 minutes of manual TMS entry, and replying to a quote took 35 minutes ². Sunset Transportation's reps quoted across two screens, pasting each email into the rating engine by hand ³.

By mid-morning the phones take over. Customers want to know where their freight is, so staff call carriers and drivers, then call the customer back. When Circle Logistics' business was doubling year over year, its team was spending too much of the day on these check calls ⁴.

Then the paperwork arrives: BOLs, PODs, carrier invoices and driver documents, each one opened, identified, keyed in and checked. SafeRide Health handled about 15,000 pages a month across 16 document types, at roughly five minutes a document ⁵. Tigris Fulfilment Partners keyed vendor documents from email into Excel and NetSuite, and manual accuracy sat around 70% ⁶.

At month-end, the cash trails the freight. North Atlantic International Logistics in Ontario spent a full day each week on an accounts receivable spreadsheet, and about 3% of its emailed invoices were reported as never received ⁷. Southeastern Freight Lines had 13 people applying payments from paper checks, with posting delays of up to two days ⁸.

Each of these moments is a handoff. Information leaves one person or system, and the next one has to rebuild it before the work can move.

Why logistics teams hire around the problem

Logistics operations grow around systems that were never designed to talk to each other: the TMS, the WMS, the accounting package, a carrier portal, and the spreadsheet that fills the gaps. 18% of the 208 logistics case studies we reviewed name Excel or spreadsheets as part of the old way of working.

Job postings show the same picture from the hiring side. We read 872 Canadian logistics and transportation job postings, most of them posted between July and October 2026, and counted what they ask people to do.

In the same postings, 13% include keeping customers updated on shipment status. Each of those lines describes a person paid to carry information from one place to another. When volume grows, the usual response is another coordinator, and for a while that holds. Over time the company's headcount starts to track its load count, because the workflow itself never changed.

In most logistics companies, information moves between customers, carriers, systems and people faster than the processes were designed to handle.

Sonia Sarrami, founder of Noesis AI

What changed when companies redesigned the work

The largest results in these case studies came from removing handoffs, and the technology involved was often simple.

Enter it once. Hughes Oilfield Transportation replaced QuickBooks and a set of spreadsheets with one system, so data entered once flows to billing, driver pay and reporting. Invoicing went from four people to two, and the company grew from about 50 to 150 trucks with the same 12 office staff ⁹. At Gary Graham Transport, once order entry captured weights, BOL numbers and pay a single time, the fuel tax return dropped from three days to 15 minutes, and the company avoided a hire ¹.

Read the quote request where it lands. ShipNova began extracting load details from emails and PDFs straight into the TMS, and pricing from its own lane history. Load building fell from 20 minutes to 3, quote replies fell from 35 minutes to 12, and its quote win rate rose 18% ².

Swap the check call for an early warning. Choptank Transport moved its carriers onto GPS tracking with document upload. Tracking accuracy rose from about 40% to over 97%, and the team stopped making check calls. Now they tell customers about a late truck before the customer asks ¹⁰.

Let AI read the documents, and let people check the doubtful ones. Tigris Fulfilment Partners routed vendor documents through classification and extraction, and flagged low-confidence fields for one-click human review. Accuracy rose from about 70% to over 95%, and the team got back 40 hours a week ⁶.

People stayed in every one of these stories, and their work moved toward the parts that need judgment. North Atlantic's team put the time into customer relationships ⁷. Circle Logistics' team spent it booking more loads ⁴. At Tigris, people review the fields the system is unsure about ⁶. About half of the 208 logistics case studies we reviewed (52%) describe what people still do after the change.

The pattern behind the results

The companies that got results decided what the workflow should be before adding any technology. Then every step got one of four labels: eliminate it, automate it, delegate it to AI, or keep it human. Redesigned that way, the quote workflow from ShipNova and Sunset Transportation looks like this.

The step that disappears often saves more time than the step that gets smarter. The steps that stay human are the ones where a mistake is expensive: pricing an unusual lane, deciding what to do about a late truck, settling a disputed charge.

How a workflow handles its exceptions decides how much of the work is automated in practice.

Sonia Sarrami, founder of Noesis AI

What the companies reported

Across the 208 logistics case studies, the most common target is the core operation: dispatch, freight and warehouse work. Finance work such as billing, payables and cash application comes second. Most companies report less manual work, and a large share also report more capacity, meaning the same team handles more loads, orders or customers.

Fewer than half of these implementations used AI beyond rule-based automation. Most of the gains came from fixing how work flows, with AI added where the work needs judgment or the documents are messy. We recommend the same order.

Where to start

Start with one workflow. In logistics it is usually one of these six, and the right one is where your team spends the most time rebuilding information someone else already had.

Map the process your team runs today, workarounds included, and measure it before anything is built: minutes per quote, check calls per day, days from delivery to invoice. That baseline tells you later whether the redesign worked. Then label each step, and build the new workflow inside the TMS, WMS and inboxes your team already uses.

Our masterclass walks through the whole method on a worked example of a 40-person logistics company, from understanding the business to finding roughly $100,000 a year in potential value.

Industry associations have only started to cover the subject. Of the 186 logistics and transportation associations whose public event calendars we could read, 4% listed any AI-related event between October 2024 and the end of 2027.