Shipium Rethinks Freight Audit as Forensics, Not Archaeology

Freight audit has long been one of the least glamorous corners of supply chain operations. It traditionally involves boxes of paper or a team hunched over spreadsheets reconciling invoices weeks after the shipments in question have already been delivered. Shipium wants to retire that model.

The Seattle-based shipping platform recently launched Always On Audit, an AI-driven system built to replace month-old invoice reconciliation with continuous, real-time monitoring of every shipment a customer sends.

Jason Murray, Shipium’s CEO and co-founder, and Penny Allen, the company’s SVP of AI Product and Engineering, joined FreightWaves Today to explain the product and why it marks, in their view, the next era of financial controllership for shippers.

Murray built Shipium around a premise drawn from nearly two decades in supply chain organizations, where he helped build technology stacks still in use today before founding Shipium in 2019. “We think of ourselves as the rails for shipping,” Murray said. “We handle everything from promise to payment when it comes to shipping.

We’ve traditionally been focused on the parcel space, but we do all of the routing and decision-making through that whole chain.” Audit, Murray said, closes the loop: “Audit completes the final piece of this, where we’re tying the financial records to the shipment itself.”

Allen drew a distinction between the old model of auditing and the new model Shipium has introduced. “Monthly audit and continuous audit are not the same job at different speeds,” she said. “They audit different objects. Monthly audit is a batch job over one artifact: the invoice. It’s archaeology.

Continuous audit’s object is the shipment itself from the moment it comes into existence to the moment it’s delivered and an invoice occurs. It’s forensics.” That shift meant rethinking the architecture from scratch.

“What I need is a record that opens at the start and is continually updated when it’s created, rated, carrier selected, label printed,” Allen said. “Every step needs to write to the same record in real time. That’s the whole architectural bet.”

Getting there required compute power, heavy data engineering, and an AI layer trained on nuance, according to Allen. “It used to be years of hand-built mappings,” Allen said. “With AI, I can do it in the span of a heartbeat.

It takes an intelligence layer that’s trained on nuance, knowing that this code on one carrier’s bill matches that code on a different one.” The result, she said, is a relentless, always-on system that is running every single decision Shipium makes against the gold standard and explaining what’s wrong in the moment.

The product is not designed to be fully hands-off. Murray said humans still handle the toughest exceptions, but every case feeds back into the model. “We do have humans to backstop some of this, right? You can think of them as being hyper-efficient, in terms of they’re only going to get the nastiest of exceptions,” he said.

“Once the human does it one time or a couple times to make it all tie out, then the next time through, the system is going to know what to do with it.” The “digital twin” is Shipium’s internal model of what a shipment’s cost and physical specifications should be, built at the moment it rates and routes a package.

Murray traced the idea back to earlier parts of his career. “When we’re rating and costing a shipment, we have effectively a digital twin of what we believe the carrier contract is, what we believe the physical dimensions of this item are, and all of the constraints around this,” Murray said.

Shipper-side errors, according to Allen, are what quietly drive many billing disputes. “This is the half that nobody wants to talk about,” she said. “Many of the problems are somewhere inside the shipper’s infrastructure: declared weight, a dimension problem, something that doesn’t match what was actually shipped.”

Because Always On Audit retains the full shipment record, not just the final invoice, the Shipium team can trace an error to its exact origin. “We can forensically recreate what actually happened,” Allen said.

“Not just that it happened on Wednesday at 2:00, but we can say it happened on this lane, on this dimensioner, at this station, with this operator, or it came in on this data stream from your tenant.” Allen described the combination of execution and audit data living in one system as the hardest part for competitors to replicate.

“It comes down to this: one system, one record, one evidence chain, no integration seam, no emailing spreadsheets, no waiting for a report to run, and no trying to use archaeology to figure out what happened,” she said. Because findings can surface while a shipment is still on site, Shipium’s agents can act before a bad charge is ever generated.

“We can move an agent to act on that shipment before it leaves the facility to choose a different carrier or service method, or in extreme cases, stop the shipment and hold it until a human in the