We are building the first fully robotic cross-dock facility for freight.

In June 2025, we announced we were building the first fully robotic cross-dock facility for freight — automating inbound receiving, dimensioning, smart sortation, and outbound dispatch.

TechCrunch covered the announcement. Our investors called it "full-stack automation, not just digital wrappers on legacy processes."

This is the progress report: what we have built, what we have learned, and what the data says about automating a cross-dock facility that processes thousands of pallets per month.

Why Automate the Cross-Dock

A cross-dock is the simplest building in logistics. Freight arrives on one side, gets sorted, and departs from the other side. There is no long-term storage, no complex pick-and-pack, no inventory management. It is a flow-through operation: in, sort, out.

This simplicity is exactly what makes it automatable. The operation has a small number of discrete steps, each of which can be instrumented and eventually replaced with robotics:

The first four steps are in active development. The last three are planned for subsequent phases.

What We Built First: The Digital Twin

Before deploying robots, we built the data layer. Starting in early 2025, we installed cameras across the LA test facility and used computer vision to create a complete digital model of facility operations.

The digital twin captures:

• Pallet flow patterns — where freight enters, how it moves through the facility, where bottlenecks form
• Dwell distribution by sort zone — which areas of the facility accumulate freight and why
• Dock door utilization — which inbound and outbound doors are active, when, and for how long
• Sort accuracy — how often pallets are routed to the correct outbound lane on the first pass
• Labor density mapping — where human workers spend their time and which tasks consume the most labor hours

This digital twin allows us to simulate changes before deploying them physically. We can model what happens when a robotic sorter replaces a forklift in Zone 3 before buying the robot.

FIGURE 1: Cross-Dock Dwell Decomposition — Where Time Goes

CHART 1: Automation Phases and Dwell Reduction Projections

What the LA Test Facility Taught Us

Lesson 1: The bottleneck is scheduling, not sorting

We expected the sort-to-outbound step to be the constraint. It is not. The data shows 70% of dwell time is freight waiting for an outbound truck. The sort itself takes ~2 hours. The wait for the next departure takes 22+ hours.

This means the first automation priority is not a robot — it is an AI scheduling system that reduces the gap between inbound arrival and outbound departure. The robot makes the physical handling faster. The scheduling makes the wait shorter. The scheduling has 4x the dwell reduction impact.

Lesson 2: Vision-based identification eliminates the barcode bottleneck

Traditional cross-dock operations require every pallet to be scanned with a handheld barcode scanner. This creates a sequential bottleneck: one worker, one scanner, one pallet at a time.

Computer vision can identify pallets by shape, label, and marking without a barcode scan. Early testing shows identification accuracy above 95% for palletized freight with readable labels — and the system improves with training data. At 95%+ accuracy, the remaining 5% get flagged for human review rather than requiring 100% manual scanning.

Lesson 3: Robotic sortation works for pallets, not yet for mixed freight

Palletized freight on standard 48×40 pallets can be moved by autonomous pallet movers (APMs) with high reliability. Non-standard freight — oversized, oddly shaped, floor-loaded — still requires human handling.

In our cross-dock network, approximately 78% of freight is palletized on standard pallets. This means robotic sortation can address the majority of volume on day one, with human workers handling the long tail of non-standard freight.

Lesson 4: The economic case is about throughput, not labor replacement

The primary benefit of automation is not reducing labor cost — it is increasing facility throughput. A robotic sorter operates 24/7. A human dock worker works an 8-hour shift. The same facility footprint can process 2–3x more freight per day with automation than with manual operations.

At $45M revenue across 50+ cross-docks, throughput is not a constraint. At $200M+ revenue, it will be. The automation investment is about building capacity for the next 5x of growth without building new facilities.

The Roadmap

Target: sub-16-hour average dwell (0.65 days) at the flagship facility, with subsequent rollout to Chicago, New Jersey, Dallas, and Miami.

For context: the ORD (Chicago) facility already achieves 0.67-day average dwell with human operations and 85.5% same-day throughput. The robotic facility aims to match that performance with less labor and more consistent 24/7 operation.

What This Means for the Industry

No traditional LTL carrier is building robotic terminals. ODFL invests $75M/year in IT. XPO has the most advanced AI stack in legacy LTL. Saia is building physical terminals. None of them are automating the dock.

The reason is structural: their terminals are built for storage and consolidation, not flow-through. Automating a 200,000 sq ft terminal with 100+ dock doors and multi-day freight staging is a different problem than automating a 30,000 sq ft cross-dock with sub-24-hour flow-through.

The cross-dock architecture — simpler operations, shorter dwell, fewer freight touches — is inherently more automatable than the terminal architecture. This is the second-order advantage of the cross-dock model: not only does it reduce dwell and damage today, it creates the physical substrate for robotic automation tomorrow.