C.H. Robinson's pursuit of $300 million in projected synergies from its RXO acquisition has put a public price tag on what AI-driven efficiency can deliver at the largest US 3PL scale — but the operator actually executing against that target still faces the same problem its smaller peers do.

That problem, according to Nexcade co-founder and CEO Dan Bailey, is not compute or model quality. It is context.

"Whether you're talking about international freight or on the brokerage side, there is so much embedded knowledge built up through years of experience and through real market nuance — and so turning that into insight that AI can use is an extremely, extremely tough challenge," Bailey said from London.

How deep is shadow AI use already in freight?

Bailey, whose company tracks between 40 and 50 leading freight forwarders and logistics service providers globally through every public AI announcement they make, called the volume of those disclosures over the last six to nine months "overwhelming." Adoption is already running ahead of policy.

"If you're not adopting it, I can guarantee you that your employees are in shadow capacity one way or another — there was a ton of shadow ChatGPT, shadow Claude being used across this industry and many others right now," Bailey said.

For C.H. Robinson, the risk attached to its $300 million RXO synergy figure is execution rather than strategy. The Robinson lean operating model has produced bottom-line results in its own business. Applying it to RXO, a separate operation with limited customer overlap and different embedded context, is the harder test.

What does Nexcade's quoting workflow actually do?

Roughly 90% of freight quote requests still arrive by email, Bailey said, and that is the surface Nexcade's agents target. The agents read customer inboxes, classify freight type, extract shipment details, and execute spot procurement lookups — checking contracts, spot market rates, and overseas agent portals via API or browser.

The intended end-state for an operator: arriving each morning to 10 to 15 quotes already staged for a pricing decision, rather than working through four or five emails and multiple lookups per quote. Teams using the platform have doubled files-per-head throughput, according to Bailey. One customer now sends automated responses to overseas agent requests that arrive at 3 a.m., before competitors are open for business.

Customer volume splits about 75% air and ocean, with the remaining 25% on road freight including US domestic and European lanes.

How should companies that haven't piloted AI start?

Bailey recommended a dual strategy. On one track, leadership identifies strategic workflows and builds the data foundations an enterprise rollout requires. On the other, willing managers run small pilots without waiting for that rollout.

Companies often learn as much from those experiments as from the flagship project, he said, getting "three or four bites at the apple" before the biggest initiatives even reach rollout. He drew on prior experience scaling the communications platform Sedna from 20 to more than 250 customers, where change management mattered as much then as it does now.

What metrics capture AI ROI beyond hours saved?

Two numbers tend to get overlooked, Bailey argued: risk reduction and revenue impact.

On the risk side, reconciliation tools and exception handling reduce demurrage, detention, and unexpected charges. The aggregate effect becomes material in a 12 to 18 month window. On the revenue side, speed-to-quote data — average win rates by response time and margin by lane — quantifies the commercial upside of AI, not just hours reclaimed.

For a market where 3 a.m. responses and lane-level win rates already differentiate winners, the path from shadow AI usage to documented ROI runs through context, measurement, and a willingness to run several small pilots before committing capital to the largest ones — a discipline C.H. Robinson will need to scale fast if its $300 million RXO synergy number is to hold.