AI in forwarding is getting paid first for checking the bills

AI’s first real payday in freight is the boring one: checking the bills
Evos, a start-up that builds AI software “operators” for small and midsized industrial businesses, including freight brokers, has raised $9 million. It is a pre-seed round, meaning very early money. Uncork Capital and Field Ventures co-led it, with Crosslink Capital, Utopia Capital Management, Sandwith Ventures and angel investors joining, according to FreightWaves on 16 October. Evos announced the round on Wednesday 14 October. TechCrunch’s Mobility newsletter listed the deal on 18 October.
Evos says its annual recurring revenue (ARR, the yearly value of its subscriptions) passed $1 million four weeks after launch. It has not disclosed its valuation or how many customers it has.
Our view is simple. AI in freight is not getting paid first for big ideas like self-running supply chains. It is getting paid for dull money work: checking invoices, catching charges nobody billed, and spotting rates that do not match. That is good news for forwarders, as long as they treat the vendors’ numbers as claims, not facts.

What Evos says it does

Evos’s software learns how staff make decisions inside the systems a company already uses. Once its decisions keep matching those of experienced workers, it takes over that task. Founder and CEO Urav Shah said that much of the know-how in these firms sits with long-serving staff and “has never been written down.” In his words: “The most capable system in many of these businesses is the person who has been there 20 years”.
Evos says it can deploy an operator in 24 hours. It compares that with 9 to 18 months for work led by consultants. That is Evos’s own comparison.
The main proof point is MyFreightWorld, a freight broker. There, an Evos operator audits LTL carrier invoices against the broker’s transportation management system (TMS, the software that holds bookings and agreed rates). LTL means less-than-truckload: freight that shares a truck with other shippers’ goods.
Evos says the operator made nearly 3,500 audit decisions over two months, with 97% accuracy. FreightWaves points out that Evos did not explain how it worked out that rate, or what sample it used. Evos also says the operator flagged more than $600,000 in billing discrepancies over the most recent five weeks. FreightWaves notes that this is not money recovered or savings realised. The company has not said how much it has actually clawed back, and its accuracy claim has not been independently checked.
So the honest reading is this. A machine looked at a lot of bills and raised a lot of flags. Whether those flags turned into cash is still unknown.

Why invoice audit is the obvious first job

Freight bills are full of small, rule-based errors. That makes them a good fit for software that checks one line against another.
FreightWaves cites a 2024 analysis by AFS Logistics. It found 11,881 errors in nearly 250,000 invoices. Inaccurate accessorial charges made up 42.8% of those errors. Accessorials are the extras on a freight bill, such as a liftgate, a residential delivery or waiting time. Fuel surcharges made up 14.4% of errors and freight misclassification 12.3%. Misclassification means billing goods under the wrong freight class, which changes the price.
None of this is new. In October 2023, Loop, a company that uses AI to audit and pay freight bills, raised a $35 million Series B led by J.P. Morgan Growth Equity Partners and Index Ventures. Its press release claimed that “on average, 20% of freight invoices have rate errors, which in turn increase costs.” That is Loop’s figure, from its own announcement. Still, it shows investors have backed this idea for at least three years.
What has changed is the pitch. Older audit tools checked bills against rules. The new pitch is software that copies how a skilled clerk thinks, then does the job alone.

Forwarders are saying the same thing about their own books

The leaks are not only in what carriers charge you. They are also in what you forget to charge your customers.
In The Loadstar on 16 October, Metro Shipping and Portcast argued that forwarders must fix unbilled charges, manual rework and bad data before AI can protect margins. Ian Powell, tech and solutions director at Metro, said margins slip through wrong buy and sell rates, costs not passed on, and repeated admin. “If we’re constantly reworking activity” was how he began the point. He called re-invoicing and raising credits against jobs “not optimal” from a cost point of view.
Emma Crumpton, enterprise solutions director at Portcast, pointed to detention and demurrage. These are the fees for keeping a container too long, at the port or outside it. “When I think about detention and demurrage, it’s a complexity issue,” she said. Free-time allowances, charging start points and contract terms all vary. Forwarders sometimes swallow the cost to keep a customer happy. “You don’t need to apply AI for the sake of it,” she added.
Neolink, an Australian forwarder, gives a rare hard number from the forwarder side. It says about half its shipment workflow is now automated. Its staff handle 80% to 90% more shipment volume than three years ago, without extra headcount. “Automation, to be honest, has probably been even bigger for us than AI,” said Sean Crook of Neolink. He also credited a model that gives each customer one contact from start to finish, and getting purchase-order data two or more weeks ahead.
That last point matters. Neolink’s gain is not one clever model. It is plain automation on clean, early data.

The wider evidence is still thin

Big-company results do not yet back the hype. Boston Consulting Group surveyed executives at 30 leading global logistics firms and published the results on 24 August. It found 97% rank AI as a strategic priority. About 70% have an AI strategy and 67% a dedicated AI budget. But only 13% say AI is delivering measurable financial impact.
BCG’s own diagnosis: “The gap between input investment and concrete output benefits is a deployment problem, not a technological one.” It named fragmented data, standalone tools that do not connect, and too little spent on people and change. In the survey, 80% named a single standalone tool as their most impactful AI solution. 37% named a lack of in-house AI expertise as their main barrier.
Writing in The Loadstar on 28 August, Adam Clermont cited the BCG figures and drew his own conclusion from that week’s AI deployments: “the winners aren’t chasing grand visions, they’re automating the boring stuff.” Evos is a clean example of that.

What this means for forwarders

First, start where the money is easiest to count. Unbilled detention and demurrage, missed accessorials, and buy-sell rate mismatches are all measurable. If a tool cannot show you cash collected against those, it is not ready.
Second, ask for recovered dollars, not flagged ones. A $600,000 flag total tells you how many questions the software raised. It does not tell you how many were right, or how many your team had time to chase. Ask any vendor how it measures accuracy, on what sample, and who checked it.
Third, fix the data before you buy. Metro needed a three-month pilot just to connect systems for airfreight tracking, The Loadstar reported. If your TMS rates are stale, an AI auditor will flag noise.
Fourth, expect the same tools to be pointed at you. Shippers and brokers can run audit software on your invoices just as easily. That likely means more line-by-line disputes over surcharges and accessorials. Clean, well-documented bills will become a sales point.
Fifth, watch the lanes with the messiest charges. LTL, drayage (short truck moves to and from ports) and container late fees carry the most extras. Those are where audit tools will land first, and where margins will feel it first.

Our call

Our prediction: by the end of March 2027, buyers of AI audit tools in freight will demand recovered-dollar figures, and at least one vendor in this space will publish one alongside a new funding round. Vendors that keep quoting only flagged totals will find the next round harder to raise. We also expect at least one mid-sized forwarder to announce an AI billing-check tool aimed at its own unbilled charges, not at carrier invoices, within the next six months.

Sources

• \
• FreightWaves — Evos raises $9M to build AI operators for industrial businesses, 16 Oct 2026 \
• TechCrunch — TechCrunch Mobility: Rivian’s next act, 18 Oct 2026 \
• The Loadstar — AI won’t plug forwarder margin leaks without better data and processes, 16 Oct 2026 \
• BCG — AI Hasn’t Fully Paid Off Yet in Logistics. But It Will., 24 Aug 2026 \
• The Loadstar — Most in logistics still can’t make AI pay – the few that can think smaller, 28 Aug 2026 \
• Business Wire — Breakthrough Transportation Logistics Startup Loop Announces $35 Million Series B, 3 Oct 2023