What US freight and logistics companies disclose about AI
Of 30 US freight companies, 18 name an AI initiative, and 11 of the 13 above $2B in revenue do. Two credit a measured cash figure directly to AI.
What did the study find?
Most large US freight companies publicly name a specific AI initiative, and most smaller ones don't. Of the 13 freight companies with $2 billion to $10 billion of 2025 revenue, 11 named one; of the 17 with $50 million to $2 billion, 7 did. Across all 30, two companies credit a measured cash figure directly to AI: ArcBest and Dorian LPG.
AI research agents (Anthropic's Claude) did the research: they searched the public record for each company, recorded every AI initiative they found with a quote and a link, and graded what each quote claims. Every quote was then read again on its source page, by script or by a separate checking agent, and a checking agent re-read every cash and hours figure. Luke Wilcox, Roiva's founder, reviewed the grade and use-case mapping of all 111 quotes from that first pass. A second set of agents then listed every earnings call, conference and annual-meeting transcript in the window, read the 393 they could find (four could not be), and added 52 quotes the first pass had missed; each was checked on its page and graded under the same rules, and Luke reviewed those too.
The study is the research behind the Roiva AI in Practice Index, US Freight & Logistics 2026, which names four leaders.
How do freight companies compare with other industries?
Smaller freight companies name AI about as often as banks, credit unions and manufacturers of the same size, and far less often than insurers. The other industries were studied only at $50 million to $2 billion of revenue, so the larger freight companies have no peer column yet.
| Freight, $2B-$10B (n = 13) | Freight, $50M-$2B (n = 17) | Banks (n = 50) | Credit unions (n = 50) | Insurance, excl. AI-native insurtechs (n = 42) | Manufacturing (n = 49) | |
|---|---|---|---|---|---|---|
| Name at least one specific AI initiative | 11 (85%) | 7 (41%) | 26 (52%) | 25 (50%) | 33 (79%) | 25 (51%) |
| Give any figure for one (cash, hours or operational) | 6 (46%) | 1 (6%) | 6 (12%) | 12 (24%) | 11 (26%) | 8 (16%) |
| Counting only figures the company or the press published | 6 (46%) | 1 (6%) | 5 (10%) | 4 (8%) | 10 (24%) | 7 (14%) |
| Give a cash figure of any kind | 1 (8%) | 1 (6%) | 3 (6%) | 3 (6%) | 2 (5%) | 5 (10%) |
| Give a cash figure they credit directly to AI | 1 (8%) | 1 (6%) | 2 (4%) | 2 (4%), both forecasts | 1 (2%) | 2 (4%), both forecasts |
| Nothing found at all | 1 (8%) | 9 (53%) | 18 (36%) | 25 (50%) | 6 (14%) | 12 (24%) |
A figure that only counts people using a tool, such as the share of staff using AI, isn't counted as a figure here, in any industry; that rule, adopted during this review, moved one bank, one manufacturer and one trucker out of "gave a figure". The freight samples are every company that met the rules, not draws: 13 and 17 companies, so one company is 8 or 6 points. The insurance sample counts AI in products an insurer sells; the others don't.
Do larger freight companies disclose more?
Yes, and the break comes at about $2 billion of revenue. None of the three companies with $1 billion to $2 billion of revenue named an AI initiative, against 11 of the 13 above $2 billion.
| Revenue | Companies | Named an initiative | Gave a figure | Nothing found |
|---|---|---|---|---|
| $50M-$500M | 7 | 3 | 1 | 3 |
| $500M-$1B | 7 | 4 | 0 | 3 |
| $1B-$2B | 3 | 0 | 0 | 3 |
| $2B-$5B | 7 | 5 | 2 | 1 |
| $5B-$10B | 6 | 6 | 4 | 0 |
Below $2 billion, three of the seven smallest companies and four of the seven mid-sized ones named an AI initiative; the groups are seven companies or fewer.
Why do trucking companies look so different above and below $2 billion?
Because the smaller truckload carriers say almost nothing about AI. Of the six trucking companies below $2 billion, one named an AI initiative; all nine above it did.
| Segment | Companies | Named an initiative | Gave a figure |
|---|---|---|---|
| Trucking, $50M-$2B | 6 | 1 | 0 |
| Trucking, $2B-$10B (LTL, truckload, Landstar) | 9 | 9 | 5 |
| Water freight (both bands) | 8 | 5 | 1 |
| Brokers, forwarders and 3PLs (both bands) | 3 | 2 | 1 |
| Air freight, rail and warehousing ($50M-$2B) | 4 | 1 | 0 |
Heartland Express's 2025 10-K says it outright: "We do not currently use AI in any material capacity." Several smaller carriers describe safety cameras, event recorders, telematics or dispatch technology without saying AI or naming the vendor, and the study doesn't count those: many modern dash cams use AI, but there is no way to tell from the filing which these are. Where a company named a product its maker describes as AI, such as Netradyne's AI dash cams or Sofar's Wayfinder, it counts. No smaller trucker's latest 10-K names one.
Above $2 billion, the less-than-truckload carriers and the large truckload carriers and brokers talk about AI on nearly every earnings call. Earnings calls and conference transcripts carry 75 of the 105 quotes from the larger companies.
Which companies credit a cash figure to AI?
Two companies credit a measured cash figure directly to AI. Each figure below is the company's claim, as published; none was checked against the company's accounts, and none is a benchmark.
| Company | Figure | Credit to AI | Source |
|---|---|---|---|
| ArcBest | City route optimization, "which uses AI to reduce manual tasks, improve route planning, and maximize asset utilization", brought "the total savings from the project to $15 million in 2025"; in 2024 it was saving "around $1 million each month" | Direct; the $15 million is the project's total to date | Q4 2025 and Q1 2024 earnings calls |
| ArcBest | "AI-powered process improvements" in truckload buying are "delivering $2.5 million in operating income benefit last year" | Direct | Q4 2025 earnings call, January 2026 |
| Dorian LPG | "Dorian LPG's use of Wayfinder resulted in average cost savings of $26,000 per voyage, a 9% average reduction in fuel consumption per voyage" | Direct | Dorian press release, February 2025 |
ArcBest's $15 million is the route-optimization project's total savings to date, of which $2 million came in 2025. Wayfinder is Sofar Ocean's voyage-optimization platform; Dorian's release pairs it with its own ship models "enhanced with machine learning tools", and Sofar describes Wayfinder as combining "machine learning models with experienced human routing experts." Dorian gives no count of voyages or ships.
Cash figures set beside AI without being credited to it, or credited to AI in general rather than to a named use of AI, count as general statements about AI, not as figures for an AI initiative. Among them are Air T's segment earnings, credited to "revenue growth" and "lower labor costs as AI-driven tools improved efficiency", Werner's back-office expense, "lowered by 40% over the last two years through modernization and AI automation", RXO's costs taken out since its spin-off, "through targeted initiatives, including AI investment, real estate optimization, and productivity", its brokerage cost per load, and Schneider's cost and Logistics margin targets.
What do the other figures look like?
Mostly productivity, and mostly at brokers. The larger companies measure AI in calls automated, hours saved and productivity rather than dollars.
- RXO says agentic AI for carrier inquiries is "reducing manual effort by tens of thousands of hours," and that its AI agents "automated more than 500,000 phone calls in the quarter" (Q1 2026).
- Schneider's 2025 annual report credits AI and generative tools with a "61% productivity improvement in carrier-broker matching."
- FedEx Freight says machine-learning tools for its dock leaders have "enabled a 10% reduction in planned versus actual handles" (Investor Day, April 2026).
- ArcBest says "more than 23,000 carriers used our AI phone agent to cover over 7,000 shipments" in 2025.
What do freight companies use AI for?
Matching freight to trucks, routing it, and pricing it. Across both groups, the research recorded 72 specific initiatives; the most common uses were:
| Use | Initiatives | Example |
|---|---|---|
| Freight quoting and pricing (the sell side) | 8 | RXO's spot-quote agent, ArcBest's dynamic pricing, XPO's LTL pricing |
| Route, linehaul and dispatch optimization | 6 | XPO's linehaul and pickup-and-delivery models, ArcBest's city route optimization |
| Customer care and assistants | 6 | ArcBest's virtual agent, Werner's conversational AI |
| Load matching and carrier sourcing (the buy side) | 6 | RXO's freight-matching model and AI voice agents for carrier calls |
| Freight operations agents (documents, booking, appointments, payments) | 4 | ArcBest's 30-plus AI agents, Landstar's agentic workflows |
| Carrier vetting and cargo-theft prevention | 4 | AI fraud screening at Landstar, Werner and RXO |
| Vessel voyage and fuel optimization | 4 | Dorian with Sofar's Wayfinder, Pangaea's machine-learning weather routing |
| Employee AI assistants | 3 | Microsoft Copilot at Landstar and Pangaea, RXO Assist |
One company reports autonomous trucks on its lanes: Schneider says it is "running a few lanes today" with Aurora and Torc.
How was the study done?
Each group is every US freight company that met the rules, not a sample. The research window covers disclosures dated 2024-01-01 to 2026-11-09.
- Population: US companies filing 10-Ks under the SEC's freight codes (trucking, rail, water freight, air courier, freight arrangement, warehousing, and freight companies coded as transportation services), with 2025 revenue of $50 million to $2 billion (17 companies) or above $2 billion and up to $10 billion (13). Foreign filers, passenger airlines and cruise lines, subsidiaries of other listed companies, and companies that don't move freight (railcar leasing, travel management, cash handling) were left out. Larger companies, such as C.H. Robinson, J.B. Hunt and the Class I railroads, are not in it yet.
- Sources searched: 10-Ks, 10-Qs, annual reports, proxies, 8-K exhibits, earnings calls, press releases, investor decks, company websites, vendor case studies and trade press, plus a search of the companies' SEC filings for 53 AI vendors' names and web searches pairing each company with likely vendors.
- Who did it: six research agents (Claude Opus), three per revenue group, and two checking agents; then a set of agents that re-read the earnings calls in the window (393 of 397 listed; four could not be found).
- What counts: a specific AI initiative named in a public source. The source must say AI, or name a product its maker describes as AI. AI in something a company sells to others is left out, and so is selling to AI customers, such as hauling data-center equipment.
What are the study's limits?
The groups are small: 13 and 17 companies, so one company is 8 or 6 points, and RXO alone holds 25 of the 105 quotes from the larger companies.
- Earnings calls. Every call in the window was listed, and all but four were read; Universal Logistics' calls after mid-2025 couldn't be found. Calls after November 9, 2026, such as Proficient Auto Logistics' third-quarter call, are outside the window.
- What carriers don't say. Smaller carriers that describe cameras or telematics without naming the vendor or saying AI count as naming nothing, even if the tools use AI.
- Dates. A source with no publication date, such as many vendor customer stories, isn't counted.
- Judgment. Mapping and grades were made by research agents under one written rubric. A person then reviewed every quote, including the 52 the call re-check added. Where a company described what might be one program on several calls, a person decided whether to merge them; two pairs were merged and 13 kept apart.
Has this study been corrected?
Fourteen times: on November 15, 2026, nine times on November 16, and four times on November 17. Four changed the manufacturing column, one the insurance column, five the freight figures, and four both the freight figures and the insurance column.
- A general remark about AI isn't an initiative (November 15). On its second-quarter 2026 call, Alamo Group set "early efforts to apply artificial intelligence" beside a company-wide SG&A ratio. That names no specific use of AI, so it now counts as a general statement about AI, not as an initiative with a cash figure. Manufacturers that name an initiative went from 27 (55%) to 26 (53%), any figure from 11 (22%) to 10 (20%), figures the company or the press published from 10 (20%) to 9 (18%), and cash figures of any kind from 7 (14%) to 6 (12%).
- A figure for a whole program isn't a figure for AI (November 16). At its June 2026 investor day, Graham said it expects "returns exceeding 20% ROI" from its IT roadmap. The roadmap covers ERP, Office 365, a defense cybersecurity certification, cloud, and "AI, automation, and analytics" as one item, and the sentence with the figure doesn't name AI. A figure for a whole program, with AI as one of several levers and no share given to AI, now counts as a general statement, not as a cash figure for an AI initiative. Graham still names an AI initiative. Manufacturers that give any figure went from 10 (20%) to 9 (18%), figures the company or the press published from 9 (18%) to 8 (16%), and cash figures of any kind from 6 (12%) to 5 (10%).
- A figure has to be tied to the AI in the quote (November 16). On its August 2026 earnings call, Azenta said it had completed "the first customer deployment of an AI-enabled biorepository inventory solution," and three sentences later that over time it expects "to double productivity." Under a stricter rule, applied so far to the manufacturing sample, a figure in a sentence that doesn't name AI counts as AI's only when the sentence just before it names the AI work, the figure refers to that work alone, and the quote includes both sentences. Here two sentences come between them, so the productivity figure no longer counts as a figure for an AI initiative. Azenta still names an AI initiative. Manufacturers that give any figure went from 9 (18%) to 8 (16%), and figures the company or the press published from 8 (16%) to 7 (14%).
- The insurance column brought up to date (November 16). The insurance column was last updated on November 10. Since then the insurers' search was extended back to January 2024, the start of this study's window, and to more kinds of source, such as investor conference talks and releases published on other sites; reviews removed rows whose source doesn't say the work came from AI, such as NMI's, whose uses are called only "tools"; and the stricter rule above, now applied to the insurers too, removed figures whose source doesn't tie them to a named use of AI or whose quote doesn't contain them, among them Crawford & Company's assessment "up to 97% faster", listed in a release without naming AI. Insurers that name an initiative went from 36 (86%) to 35 (83%), any figure from 16 (38%) to 14 (33%), figures the company or the press published from 16 (38%) to 13 (31%), and cash figures of any kind from 4 (10%) to 7 (17%). The column now matches our insurers note.
- A freight figure has to be credited to a named use of AI (November 16). The stricter rule above, now applied to the freight companies, took out two figures. At its April 2026 investor day, FedEx Freight said support cost per gross-profit dollar would improve to about 60%, "supported by technology and AI-driven investments." That names no AI program, and AI is one of several levers behind a company-wide ratio, so it no longer counts as a cash figure or an initiative. Werner's "nearly 15%" fall in Logistics operating expense, from a February 2026 conference, isn't credited to AI in the quote, and the passage names no AI use, so it was dropped; Werner's 40% figure stands. FedEx Freight's "data analytics and AI to improve decision-making," which introduces a list whose items don't name AI, and Pangaea's AI team, which names no use, now count as general statements about AI. Companies above $2 billion that give a cash figure of any kind went from 7 (54%) to 6 (46%), and companies that credit AI as one of several causes from six to five. FedEx Freight still gives an operational figure. Specific initiatives went from 116 to 112, freight operations agents from 7 to 6, and quotes from 163 to 161 (from the larger companies, 142 to 140, of which 98 come from calls and conference transcripts, down from 100). No company moved in or out of naming an initiative or giving a figure.
- AI in a service a company sells isn't its own use (November 16). Azenta's AI-enabled biorepository inventory solution is part of a service it sells: its chief executive called the deployment its "first customer deployment," which strengthens its "biorepository services offering." AI in something a company sells counts as part of the product, not as an initiative in its own operations, so Azenta now counts as naming AI only in a product or a general statement. Manufacturers that name an initiative went from 26 (53%) to 25 (51%).
- A closer re-read of the figures (November 16). Before the AI in Practice Index launched, checking agents re-read every remaining figure at companies that aren't among its leaders. Knight-Swift no longer gives a figure. On its third-quarter 2026 call it said it held fixed costs flat with "AI automation" one of four levers, which names no use of AI and gives AI no share. Its "as many as 15 tows a month" is a Transport Topics writer's words, not the company's, and trade press counts only an executive's or spokesperson's own words. Above $2 billion, companies that give any figure went from 8 (62%) to 7 (54%), and figures the company or the press published from 7 (54%) to 6 (46%). Specific initiatives went from 112 to 111, and quotes from 161 to 160 (from the larger companies, 140 to 139, of which 97 come from calls and conference transcripts, down from 98). The same re-read of the insurers took out figures set in a general statement about AI, beside other causes or under a slide title: insurers that give any figure went from 14 (33%) to 12 (29%), and figures the company or the press published from 13 (31%) to 11 (26%).
- A full re-read of the initiatives (November 16). Before the AI in Practice Index launched, checking agents re-read every row naming an initiative at companies that aren't among its leaders. Knight-Swift's "Implementing AI-technology solutions" in its logistics business, from a fourth-quarter 2025 investor deck, names no AI deployment, so it now counts as a general statement about AI. Knight-Swift still names an initiative. Specific initiatives went from 111 to 110, and load matching and carrier sourcing from 10 to 9. The same re-read of the insurers took out rows that name no specific use of AI: insurers that name an initiative went from 35 (83%) to 33 (79%), and those with nothing found from 5 (12%) to 6 (14%).
- The freight figures brought up to the study's rules (November 16). On November 14, four rules were applied to the rows behind this note that it hadn't yet picked up: a figure counts only in a sentence that names the AI work or clearly refers back to it; a source with no publication date doesn't count; trade press counts only an executive's or spokesperson's own words; and a figure set beside AI that the source doesn't credit to it isn't an AI result. Landstar's vendor estimate of "$2.1M+" came out, because the RTS Labs page has no publication date, and so did Samsara's undated safety figures for Werner and ArcBest. XPO's purchased-transportation and wage-cost figures no longer count, because the sentences that hold them don't credit AI. Schneider's $40 million cost target and its 3% to 5% margin target now sit beside AI rather than being credited to it. Armlogi's picking-efficiency figure is a potential, not a result. Strata Critical Medical's "starting to employ it for real-time error checking" doesn't say AI in its own sentence, so Strata no longer names an initiative. Werner's Aurora mileage was a writer's words, not the company's. Above $2 billion, companies that give any figure went from 7 (54%) to 6 (46%), a cash figure of any kind from 6 (46%) to 4 (31%), and a cash figure credited directly to AI from 2 (15%) to 1 (8%). Below $2 billion, companies that name an initiative went from 9 (53%) to 8 (47%), any figure and figures the company or the press published from 3 (18%) to 2 (12%), and nothing found from 8 (47%) to 9 (53%). Quotes went from 160 to 129 (from the larger companies, 139 to 109, of which 78 come from calls and conference transcripts), and specific initiatives from 110 to 88.
- A re-read of the leaders' initiatives (November 16). Before the AI in Practice Index launched, checking agents re-read every row naming an initiative at the index's leaders. Rows whose quote names no specific use of AI now count as general statements. Among the freight companies, specific initiatives went from 88 to 84: route, linehaul and dispatch optimization from 9 to 8, freight operations agents from 5 to 4, and employee AI assistants from 5 to 4. No company moved in or out of naming an initiative or giving a figure, and no other column changed.
- A figure credited to AI in general isn't a figure for an AI initiative (November 17). After the AI in Practice Index launched, checking agents re-read the figures at its leaders that the earlier re-reads hadn't covered. A figure a company credits to AI in general, rather than to a named use of AI, now counts as a general statement about AI. RXO's "more than $155 million in costs" taken out "through targeted initiatives, including AI investment", its brokerage cost per load, and its "productivity is up 45%", in a passage on margin expansion "driven by artificial intelligence", no longer count. Nor do Schneider's productivity figure for "very effective AI" and its cost and Logistics margin targets, or ArcBest's "AI-driven automations" that cut emails, which name no specific use of AI. Above $2 billion, companies that give a cash figure of any kind went from 4 (31%) to 2 (15%), and companies that credit AI as one of several causes went from three to two. Specific initiatives went from 84 to 79, and load matching and carrier sourcing from 7 to 6. Among the insurers, Ategrity's and Employers Holdings' expense ratios and Baldwin's forecast program savings no longer count as cash figures for an AI initiative, so insurers that give a cash figure of any kind went from 7 (17%) to 4 (10%). No company moved in or out of naming an initiative or giving a figure.
- General statements and governance practices read on their sources (November 17). After the AI in Practice Index launched, checking agents read on their sources the governance practices and general statements about AI that the earlier re-reads hadn't covered. A general statement now counts only when it says what the company is doing with AI now: plans, hopes, hypotheticals and AI as an industry trend no longer count, and neither do risk sentences that name no use, trade-press writers' words or quotes that don't name AI. Werner's back-office figure, "lowered by 40% over the last two years through modernization and AI automation", names no specific use of AI, so it is now a general statement, and so are two other Werner rows. Werner still names AI initiatives and gives an operational figure. Above $2 billion, companies that give a cash figure of any kind went from 2 (15%) to 1 (8%), and companies that credit AI as one of several causes from two to one. Specific initiatives went from 79 to 76: route, linehaul and dispatch optimization from 8 to 7, and customer care and assistants from 8 to 7. Quotes from the larger companies went from 109 to 107, of which those from calls and conference transcripts went from 78 to 76. No company moved in or out of naming an initiative or giving a figure, and no other column changed.
- Every leader item read a second time (November 17). After the AI in Practice Index launched, a second reader, working without the first reading, read again every item behind its leaders that could be opened. Plans and hopes were dropped, programs and training recorded as uses became general statements, and quotes, statuses and dates were brought to the sources' words. Among the freight quotes, ArcBest's statement that staff use AI tools after training had been recorded twice, from two copies of the same call, and is now counted once. Quotes from the larger companies went from 107 to 106, of which those from calls and conference transcripts went from 76 to 75. No company moved in or out of naming an initiative or giving a figure, and no other figure or column changed.
- The other companies read a second time (November 17). After the AI in Practice Index launched, a second blind sample was read again, and then a second reader, working without the first reading, read again every use and governance practice at the companies that aren't among its leaders. Writers' words, figures credited to AI in general and plans were corrected or dropped. Among the freight companies, Air T's segment earnings, credited to revenue growth and to "lower labor costs as AI-driven tools improved efficiency", name no specific use of AI, so they are now a general statement, and Air T no longer names an AI initiative. Below $2 billion, companies that name an initiative went from 8 (47%) to 7 (41%), and those that give any figure, figures the company or the press published and cash figures of any kind from 2 (12%) to 1 (6%). Specific initiatives went from 76 to 72: route, linehaul and dispatch optimization from 7 to 6, customer care and assistants from 7 to 6, and employee AI assistants from 4 to 3. Quotes from the larger companies went from 106 to 105. In the insurance column, any figure went from 12 (29%) to 11 (26%), figures the company or the press published from 11 (26%) to 10 (24%), and cash figures of any kind from 4 (10%) to 2 (5%).
No bank or credit-union figure changed.
Who ran the study?
Roiva ran this study. Roiva is AI portfolio management software that tracks every AI initiative's budget, approvals, cost and ROI, so whether companies can show a return on AI is a problem the company sells into. That is a reason to check the work.
The uses this study found most often, and the gaps it turned up, shaped four of Roiva's transportation templates: Load Matching & Carrier Sourcing, Freight Quoting & Pricing, Freight Operations AI Agents, and Carrier Vetting & Cargo Theft Prevention. They sit beside the existing AI Route Optimization & Dispatch, Driver Safety & Coaching Intelligence, Predictive Fleet Maintenance and Last-Mile Delivery Optimization templates. None of the figures above went into them.
Get the next one
New field notes by email as they're published. Or subscribe to the RSS feed.