XPO

The LTL freight market entered a prolonged recession, with weak demand, suppressed rates, and rising operating costs forcing multiple carriers out of business in 2025. For XPO, operating hundreds of service centers across North America, the core challenge was granular: labor represents a large share of operating costs, yet staffing decisions lacked the precision to match hour-by-hour freight volume swings. Over-staffing burned margin; under-staffing degraded service quality. With the American Transportation Research Institute (ATRI) flagging labor as a top cost concern industry-wide, XPO needed a data-driven approach that could make defensible workforce deployment decisions at the service center and shift level — not just at the network level.

XPO developed XPO Smart, a proprietary in-house labor planning and productivity platform built on multiple machine learning and predictive analytics models. The system forecasts freight demand across both short-term (daily, hourly) and long-term planning horizons, translating those forecasts into staffing recommendations at the individual service center and shift level, down to the hour. Because the platform was built internally by XPO's team of more than two dozen AI and data science professionals, it integrates natively with existing operational workflows rather than sitting as an external layer. Running on a fully cloud-based technology stack — unlike legacy mainframe systems common elsewhere in LTL — XPO Smart enables rapid model iteration and continuous improvement. The result is a system that converts network-wide freight signals into precise, hour-level labor deployment decisions across the entire carrier network.

XPO Smart delivered a 2.5-point year-over-year productivity improvement — a result disclosed on the company's earnings call and significant because it was achieved while overall shipment volumes declined. Gains during a volume downturn indicate the improvement came from better labor alignment, not simply from carrying more freight. Qualitatively, the platform allowed XPO to maintain high service standards throughout the market downturn while positioning the network for rapid capacity expansion when demand recovers.

• 2.5 points productivity improvement year-over-year
• Achieved during a period of declining shipment volumes
• Service quality maintained across the service center network
• Network positioned for upcycle readiness without overstaffing during the downturn

• AI-driven labor matching can improve productivity even during volume downturns by eliminating both over- and under-staffing waste simultaneously
• Hourly granularity in demand forecasting unlocks optimization that daily or weekly planning cycles cannot capture at network scale
• Building AI platforms in-house enables faster iteration and tighter operational integration than relying on third-party solutions
• Productivity gains achieved during a recession create structural cost advantages that compound when freight volume recovers
• A cloud-native technology stack is a practical prerequisite for continuously retraining and improving AI models across a distributed service center network