Agentic AI in Logistics: Why 55% Accuracy Fails

Agentic AI in logistics is struggling with spatial reasoning, according to Bart Coppelmans of HERE Technologies. General-purpose AI models correctly answer only about 55% of basic direction questions. This accuracy gap is preventing the widespread use of agentic AI in logistics. Here's the thing: while AI can understand language, it fails on tasks like truck routing, navigating low-clearance bridges, finding parking, and managing congestion. Location intelligence needs to be built into logistics AI from the start, not added later. HERE Technologies, with decades of experience, emphasizes this foundational approach. What's interesting is that current AI models lack the spatial reasoning needed for real-world freight operations. For example, rerouting trucks around bridge restrictions requires specific location data that large language models don't have internally. HERE's map data includes over 800,000 truck-specific data points. Without connecting the AI model to a strong geospatial model, the responses can be dramatically incorrect. The bottom line is that integrating location data can create a feedback loop, connecting planning with real-time driver input. This moves logistics from static planning to dynamic orchestration, optimizing for cost, safety, and sustainability.