ORION doesn't chat — it just quietly saves 100 million miles a year
UPS · 2013–ongoing
What they did
UPS's ORION (On-Road Integrated Optimization and Navigation) is operations-research-plus-machine-learning applied to a brutally concrete problem: each delivery route has roughly 200,000 possible sequences, and picking a near-optimal one across 250 million address data points saves real fuel and time at UPS's volume.
What happened
UPS reports ORION saves around 100 million miles and 10 million gallons of fuel a year, cutting CO2 emissions by roughly 100,000 metric tons, for an estimated $300–400M in annual savings. More than 70% of UPS's ~55,000 U.S. routes now run on it, at an average reduction of 6–8 driving miles per route per day — rolled out gradually, with driver training built around the fact that an optimal-looking route often looks wrong to an experienced human until they trust the system.
The so-what
The single longest-running, most durable AI success story in this list predates the LLM era entirely — a reminder that "AI" delivering real value has never required a chat interface, only a well-defined problem and a metric worth optimizing.