The yards are hiding in the Inland26 agenda

The obvious conversation is carrier sourcing. The less obvious one is facility interoperability.

A dedicated fleet learns the quirks of a network over hundreds of visits. A small carrier arriving for the first time cannot. If shippers need to pull capacity from a broader, more fragmented base, facilities have to get easier for a stranger to navigate on the first try.

Every enterprise knows roughly what its TMS owns and what its WMS owns. Things get fuzzy in between.

Who owns arrival. Who owns gate state. Who owns the appointment after reality stops matching it. Who owns trailer state. Who decides which door a truck actually gets. Who records that a yard move physically happened.

The asphalt between the TMS and the WMS increasingly needs its own live state and its own orchestration.

Cargo theft is usually discussed as a transportation-security problem. It is becoming a facility-access problem, and the numbers say so before most people notice.

CargoNet’s own numbers make the case. In its most recent quarterly report, theft incidents fell 26 percent year over year, but losses more than doubled to $304.6 million. That combination only happens one way: theft is getting more targeted, not more common. Keith Lewis, the company’s own VP of Operations, put it plainly: lower incident volume should not be mistaken for lower risk, and the vulnerability is not the distribution center, it is the way the loads are moving. Fewer, smarter hits mean the next fix has to happen before the pickup, not after it.

That has an interesting implication for facilities. When trucks and drivers are abundant, wasting 30 minutes is an efficiency problem. When usable capacity tightens, those same 30 minutes become capacity.

A facility that cuts a carrier’s turn time is returning productive truck and driver hours to the network. That argues for a different yard ROI number.

J.B. Hunt’s own CEO is already saying the quiet part. “AI, automation and other tools have to be applied with purpose, tested in real-world conditions and proven through the work that gets done,” Shelley Simpson told Fortune this August.

Her own company already prices what happens when that is not true. J.B. Hunt’s “660 Minutes” research puts bad shipper dwell at roughly $15,000 a year in lost driver earning potential, the sum of slow appointments, slow loading and no drop-and-hook access. Capacity is tightening at the same time. Put those two facts together and a bad facility stops being a line on a scorecard nobody reads and starts being a number a carrier can put in a rate.

Less interesting than everyone’s directional freight forecast is what they think changes operationally: what gets more expensive, what gets harder to source, what workflows become unacceptable, what data suddenly matters.

Market turns expose operational debt. That is what is worth listening for.

A credential is not a person, and a person is not the truck they claim to be driving. The CDL crackdown is the first domino, not the last one.

FMCSA’s own Deputy Administrator, Jesse Elison, just proved a paper credential can fail at national scale. He has called the non-domiciled CDL system demonstrably failed, and the fix is pushing an estimated 194,000 drivers, about 5% of the CDL workforce, out of the industry within two years. Matthew Leffler is already asking the next question, about training mills that certify competency in 24 hours or two days. Once regulators start pulling that thread, it does not stop at the credential. It reaches the gate, where nobody is currently checking whether the driver holding that license is the one who actually showed up.

Rail does not remove execution complexity, it changes it. Containers, chassis and trailers still have to arrive somewhere, be identified, staged, moved, loaded, unloaded and handed off.

More modal handoffs mean more physical state transitions inside the fence line.

That is effectively a 45-minute enterprise discovery session if you listen for the right thing.

Everybody wants better data. What data do they actually wish they had. Everybody wants automation. What process stops them from automating today. Where does the network still depend on somebody knowing which trailer is behind Row C.

This might be the sleeper session of the conference for anyone who owns a distribution network, precisely because it is arguing past the detention framing most of this space defaults to.

Does facility performance affect transportation procurement, not theoretically, actually. Does a carrier price a consistently fast facility differently. Does it allocate scarce capacity differently. If that connection can be measured, yard modernization moves out of the narrow warehouse-efficiency bucket and into transportation strategy.

Werner’s own CIO is walking into Inland26 with a line that is going to look obvious in two years and sounds contrarian today: build versus buy versus invest, and his own view that software itself will get less important over time.

That is a strange thing for a CIO to say, unless the layer underneath the software, the verified data it actually acts on, is what he thinks is doing the real work. Clean up the physical layer first. Automate second. Every enterprise buyer who has been burned by an AI pilot that failed on bad data is about to start agreeing with him, whether or not they use his words for it.

Every company vetting carrier identity right now is checking the road. That is where the money and the vendors are in 2026. It will not stay that way.

Ryan Joyce, GenLogs’ CEO and a former CIA case officer, gets to almost the same place from the highway side. “The future of freight will not be won by the operators who trust the digital record. It will be won by the operators who verify it against the ground truth,” he has said. His company runs a roadside camera network logging nearly 20 million truck images a day just to catch a plate or DOT number swap. SearchCarriers stitches together roughly 30 FMCSA datasets for the same reason, because carrier self-reporting breaks. Once the road is this well covered, the next uncovered ground is the one place the cameras cannot reach: the dock, where the load actually changes hands.

Human out of the Loop is a false binary, and within a year nobody serious will still be arguing it that way. The dividing line was never human versus agent. It is verified versus unverified, and every company on this stage is about to get sorted onto one side of it.

The freight-agent vendors here already split two ways. Some are built to pause and ask a human when they cannot verify something. Others treat a clean TMS feed as enough to act on, full stop. The second group is building on borrowed time: the first physical-world failure that reaches a customer is what turns human in the loop from a marketing slide into a regulatory requirement.

Vooma, Envoy AI and HappyRobot are all brokerage-side automation: sourcing, negotiating and booking freight faster. None of the three touches the point where a load actually changes hands, and that gap is where the next wave of agentic freight startups gets built.

Agentic brokerage is a leading indicator. Transportation software is moving from telling humans what happened toward executing work directly, and physical logistics will follow the same arc, just later. Our view: establish ground truth, standardize the operation, orchestrate it, then automate it. An agent that books a load faster still hands it to a dock that has no idea it is coming, and that is the version of this panel someone will have to run again in two years.