Free two-day shipping rewired what customers expect from delivery, and it blew a hole in last-mile budgets that retailers are still scrambling to patch.<\/p>\n
FarEye, an AI-Powered Delivery Management System (DMS), recently launched an agentic AI dispatcher called PILOT to plan, execute, and monitor final-mile deliveries with minimal human oversight. Gaurav Srivastava, the company\u2019s co-founder and chief product and technology officer, told FreightWaves two forces made the launch possible: relentless consumer demand for free, fast delivery, and the leap in AI capability that followed Large Language Models (LLMs) debut.<\/p>\n
\u201cThe two big drivers of AI: first is what we call the Amazon Prime effect,\u201d Srivastava said. \u201cEvery consumer wants things to be delivered free and fast, and that poses a big challenge for retailers and carriers across the world.\u201d<\/p>\n
That demand lands hardest on the last mile, the final leg between a warehouse or store and a customer\u2019s door. It is also the most expensive.<\/p>\n
The Amazon Prime Effect Meets a Cost Problem<\/p>\n
Last-mile delivery consumes 40% of total supply chain cost, Srivastava said, more than any other segment of the network. It is also the most fragmented, with decision-making absorbed daily by a single, overworked role.<\/p>\n
\u201cThe dispatcher is the person sitting there locally in the city, managing your drivers, managing that chaos of a driver not showing up, a truck running into an accident, a shipment running late, a customer\u2019s urgent call,\u201d Srivastava said. \u201cAll of that is absorbed by this one persona, that one human. His entire day goes by firefighting.\u201d<\/p>\n
FarEye has worked with dispatchers for over 10 years, Srivastava said. Agentic AI opened a new possibility: an agent that works alongside the dispatcher instead of just handing them another screen to click through.<\/p>\n
Inside FarEye\u2019s Agentic AI Dispatcher<\/p>\n
PILOT covers what a dispatcher normally handles across a 10-hour shift: scrubbing order data, planning routes, sourcing carriers and drivers, handing off shipments, and monitoring the day as problems surface. It can run autonomously or with a human reviewing its calls.<\/p>\n
\u201cThis agent can proactively plan, execute, and also monitor your last-mile operations,\u201d Srivastava said. The system is built to be modular, and sits on top of the enterprise\u2019s (FarEye or otherwise) existing route optimization, compliance, and track-and-trace tools, all integrated into one decision layer. Working in a Human-In-The-Loop (HITL) model, it ensures the required human oversight while reserving human interception for critical exceptions.\u00a0<\/p>\n
For carriers and shippers, that means fewer decisions routed through one overworked person\u2019s judgment, and more routed through a system that never stops watching the board. On the contrary, enabling the dispatcher to drive sharper decisions with up to five times higher productivity.\u00a0<\/p>\n
Right Truck, Right Price, Right Time<\/p>\n
Srivastava pointed to Tractor Supply, which operates more than 2,400 stores nationwide and ships everything from a hammer to livestock feed, as an example of the orchestration problem AI is built to solve. The retailer runs its own trucks but also leans on FedEx, UPS, and regional providers.<\/p>\n
\u201cThere are days their trucks would sit idle at the store with just under 50% capacity while the store ships out products with FedEx and UPS,\u201d Srivastava said. \u201cAnd then there are days where their trucks aren\u2019t able to fulfill all orders.\u201d<\/p>\n
The agent decides, order by order, where capacity should go.<\/p>\n
\u201cEven if it\u2019s very close to your store, you might still want to give it away to a FedEx or UPS because it\u2019s more cost-effective,\u201d Srivastava said. \u201cYour cost of driving that big truck just to deliver one small hammer might not be a great use of your assets.\u201d<\/p>\n
That decision runs on live, dynamic cost data, not a fixed rulebook.<\/p>\n
\u201cShould it be loaded on your truck? Should it be loaded on a small van? Should it be given to a FedEx or a Roadie or an Uber?\u201d Srivastava said. \u201cOr should it just be deferred to tomorrow because there\u2019s no SLA breach?\u201d<\/p>\n
Srivastava said the problem is harder in last-mile than in middle-mile trucking, where networks run between fixed nodes. \u201cIn last-mile you don\u2019t have nodes. You don\u2019t know where you\u2019ll be delivering,\u201d he said. \u201cYou can get a delivery from a very remote suburban ZIP code you have never serviced before, and now you\u2019ve got to create a plan for it. That problem vastly more complex.\u201d<\/p>\n
Omnichannel Growth Outpaces the Store<\/p>\n
The shift shows up in retailers\u2019 own strategy. Enterprises like Tractor Supply spent years trying to pull customers into its stores, Srivastava said. That has changed.<\/p>\n
\u201cUntil about two years ago, Enterprises had this strategy that they wanted to drive customers to their offline store,\u201d he said. \u201cThat strategy has evolved now. They want to be present wherever the customer would find it most convenient, whether in store or to order online and get it delivered.\u00a0<\/p>\n
One of the largest global furniture retailers shows the same pattern at a larger scale. Srivastava noted the large retailer saw substantial growth on their e-commerce side compared to their retail sector.<\/p>\n
Across FarEye\u2019s retail customers, he said, online ordering is growing three to five times faster than in-store sales, pushing last-mile volume up faster than most networks were built to handle.<\/p>\n
Looking ahead, Srivastava said the same orchestration logic will extend to whatever delivery modes come next, including drones and autonomous vehicles, as those become viable alongside human drivers. For now, the agent\u2019s job stops short of the loading dock.<\/p>\n
\u201cIt can\u2019t replace a driver and a floor supervisor. People actually delivering the product can\u2019t be replaced,\u201d Srivastava said. \u201cBut the whole planning and orchestration of the last mile can be done by this agent.\u201d<\/p>\n