A €50,000 freight invoice and ten days to approve it: inside nShift Audit
Days before nShift launched its AI invoice auditor, the company’s CEO and Chief Product Officer sat down to record a conversation. Much of it came back to the same thing: a freight invoice worth tens of thousands of euros that nobody can properly check before it is due.
By Thomas Bailey, Product Innovation Lead at nShift
Every retailer that ships has seen the document Johan Hellman describes. “It could easily be a PDF document of hundreds of pages, very dense pages with shipment after shipment specified in a format that’s super hard to make sense of,” says nShift’s Chief Product Officer. “And it’s maybe 50,000 euros, and the payment terms is ten days.”
He doesn’t slow down there. “In ten days, you need to decide, is this correct? Should I pay it? And if you don’t pay it, the carrier is still coming. So you have to actually have to pay it.”
The bill finance has nothing to match
Why does a cost this size escape the controls that cover nearly every other supplier? Freight is the odd one out in accounts payable.
“It’s also a category where you don’t have purchase orders,” Hellman says. “So normally you would issue a purchase order, you’ll get the invoice, you will match it, and if it matches, then you just need a person to approve it, and then you’re done with it.”
Without a purchase order to match against, many companies fall back on watching the total. “As long as it’s the same amount as last month or a reasonable variance aligned with how the business grows, they just ignore it,” he says.
nShift’s customer research in 2026 backs that up: roughly three in four companies audited freight invoices manually or not at all.
What AI actually changed
nShift Audit, AI-powered carrier invoice audit and recovery software, went live in late August.
Hellman starts with the invoice itself. “The solution is basically to then use AI to, first of all, just make sense of these formats, these invoice formats, to ingest them on the fly without having seen them before, turn them into a standardized format that we can then process.”
Once the invoice is in a form the system can work with, nShift can compare it with the shipment data it already holds.
“We’re leveraging the fact that we already have shipment data, cost data, so we know what the shipment is supposed to cost. We know how it was booked, and then we can cross-check, and then we can tell our users, ‘Okay, these are the lines on this huge invoice, like thousands of lines. These 50 lines you should check.’”
For smaller retail teams, Hellman sees a practical consequence: “You don’t have to be an expert to find issues anymore.”
He is also careful about what an issue means. “To be clear, not everything is the fault of the carrier,” he says. “It’s quite common that as a shipper, if you do something wrong, you have the wrong behaviour, you’re using the wrong service… you’re booking out of spec.”
Three years of “almost there”
CEO Jurgen Leijdekker remembers the run up. “You had been playing over the last, let’s say, three years to see how far AI was to read those complex invoices,” he says to Hellman during the conversation. “And basically, every time it was like almost there, not almost there, getting closer.”
Then the answer changed.
“At some point you said, ‘I think the models are there.’ And that’s when we started building it. So it really was at the edge of where the models were about nine months ago.”
The first working version came together in hours. Turning that prototype into a product took another six months. That gap between a fast AI prototype and production software is the subject of a separate nShift piece on the so-called SaaS apocalypse. You can read the story behind the product on nShift’s blog.
Every invoice, always
Asked whether a retailer would run audit quarterly, Hellman pushes back. “No, it’s operational. I mean, every invoice. That’s the whole point. So you check every invoice always.”
He sees value in the months when Audit finds nothing, too. “It’s also a bit of an insurance policy that, okay, having this running, you know that you will find issues if they come, even if you don’t have any right now.”
For Leijdekker, the change is also about who can use this kind of technology.
“This technology went from being a professional service, like a back office… to, now for the first time being self-service, and therefore this service being available to the smallest shippers. And that wasn’t possible before in this industry.”
nShift Audit launched on 27th August, with general availability planned for October. The full conversation is on nShift’s site.
About the author
Thomas Bailey is Product Innovation Lead at nShift, whose delivery and experience management platform supports more than one billion shipments a year across 190 countries. His background spans product, technology, and go-to-market strategy, and he hosts nShift’s product conversations, including the episode this article draws on.
