The problem
Walmart’s supply chain has to put the right products in thousands of stores, fulfillment centers and online orders every day, at a scale where small inefficiencies add up. The middle mile, the trucks that move goods from distribution centers to stores, was the first target: multi-stop journeys have to fit store delivery windows, trailers have to be packed to use space and keep temperature-controlled items fresh, and trucks returning empty waste capacity (Walmart).
Inventory is the second problem. In markets where shelf space is scarce, overstock in one store and shortages in another turn into waste and missed sales. Walmart describes its goal as fixing imbalances before they show up in stores (Walmart). Weather adds a third pressure, because demand for essentials shifts days before a storm arrives.
What they did
Walmart built its routing and load-planning technology in-house. According to INFORMS, the company used advanced optimization models to build a roadmap for long-term supply chain capital investments, plus an application that supports daily decisions on truck routing and loading (INFORMS). INFORMS describes the work as operations research; Walmart markets the same tool as AI-powered.
Walmart says the tool plans multi-stop journeys around time, location and delivery windows, packs trailers, reacts to weather and traffic, and plans backhauls so trailers return loaded. In March 2024 it began selling the tool, called Route Optimization, to other businesses as software as a service through Walmart Commerce Technologies. “We have invested significant time, resources and operational knowhow into building solutions like Route Optimization,” said Anshu Bhardwaj, chief operating officer of Walmart Global Tech (Walmart).
The company then extended the same shared stack beyond the United States. In July 2025 it described predictive warehouse and transportation systems in Costa Rica and Canada, Self-Healing Inventory in Mexico, which detects stock imbalances and redirects product automatically, a unified view of inventory across stores and online, and agentic AI tools that let associates ask which items were shorted in a store (Walmart). Its reuse strategy is explicit: “At this scale, the only way to move faster is to move smarter,” said Vinod Bidarkoppa, chief technology officer of Walmart International.
Ahead of Winter Storm Fern in January 2026, Walmart used forecasting models and a simulation of its middle-mile network to stage supplies such as ice melt and water, and rerouted hundreds of thousands of perishable and dry goods to secondary distribution centers, Retail Brew reported (Fortune).
Results
According to INFORMS, the routing application avoided 72 million pounds of CO2 and saved $75 million during fiscal 2023, work that won the 2023 Edelman Award (INFORMS). These are the figures of the award submission as published by the awarding body; INFORMS does not publish its own audit of them.
In its March 2024 release, Walmart said the same technology had eliminated 30 million unnecessary miles, avoided 94 million pounds of CO2 and bypassed 110,000 inefficient paths (Walmart). These figures are self-reported, and the release does not state the period they cover.
For inventory, Walmart said in July 2025 that Self-Healing Inventory alone “has already saved Walmart more than $55 million”, a self-reported figure with no period or method stated (Walmart). It also said that, in early deployments of the shared platform, work that once took quarters now happens in weeks.
During Winter Storm Fern, Walmart placed storm supplies several days earlier than it could without AI, according to Retail Brew’s report. “By utilizing advanced forecasting models and our simulation platform, we can anticipate shifts in demand and dynamically reroute essential supplies,” Indira Uppuluri, Walmart’s senior vice president of supply chain technology, told Retail Brew (Fortune). No sales or availability figures were given.
What didn’t go to plan
Public sources report no setbacks or reversals in Walmart’s supply chain AI, but the evidence has gaps. Only the fiscal 2023 routing figures carry an outside body’s name, and even those come from Walmart’s award submission. The later figures are self-reported, cover unstated periods and do not reconcile neatly: 72 million pounds of CO2 for one fiscal year, then 94 million pounds with no period given.
The label also shifts with the audience. The work that won the Edelman Award is described by INFORMS as operations research and optimization, while Walmart’s commercial materials call it AI-powered. Leaders comparing vendors should know which technique delivered the result. Walmart has not disclosed how much it invested, how the $55 million inventory figure was calculated, or how much outside revenue Route Optimization earns. The storm response was reported without measured outcomes.
Lessons
- Start where the math is solvable. Walmart’s first measured win came from routing and trailer loading, a well-defined optimization problem with clear cost and mileage measures.
- Build once, reuse across markets. Walmart credits a shared stack for moving deployments from quarters to weeks across Costa Rica, Canada and Mexico.
- Act on the forecast automatically. Self-Healing Inventory reroutes stock itself rather than only flagging imbalances, which is where Walmart attributes its inventory savings.
- Get outside scrutiny of the headline number. The Edelman submission gave Walmart’s routing savings an external record that its later self-reported figures lack.
Adapt it in your organization
Is this for you?
- You run your own transport or distribution network with enough routes or stores for small savings to add up.
- You have clean data on delivery windows, trailer capacity, store demand and inventory positions.
- Your planners can act on automated recommendations, or let the system act within limits.
- You can measure cost per mile, empty miles or inventory transfers before and after.
- A supply chain executive owns the budget and the outcome.
A 90-day version
Pilot: run route and load optimization on one region’s store deliveries for a quarter, comparing planned miles and trailer fill with a matched region that keeps current planning. Owner: vice president of transportation. Success measure: miles per delivered case down at least 5% with no fall in on-time delivery.
Metric to track
Miles driven per unit delivered, with empty backhaul miles alongside it. Walmart’s own baseline is 30 million unnecessary miles eliminated across its network; for a smaller fleet, track the same ratio rather than the total. For an operations case with measured results in another industry, see GE Aerospace.
Failure modes to pre-empt
- Reporting savings with no period or method, as with the $55 million inventory figure, so the board cannot compare them.
- Calling optimization AI, or the reverse, and buying the wrong capability.
- Automating transfers between stores without limits, so the system moves stock faster than stores can receive it.
- Scaling to new markets without a shared platform, so each country rebuilds the same tools.
Questions to take to your leadership team
- Which of our supply chain decisions are well-defined enough to optimize today?
- Who outside the team would verify our savings before we report them?
- Where should the system act on its own, and where must a planner approve?
Executive FAQ
How much has AI saved Walmart’s supply chain?
INFORMS reports $75 million in fiscal 2023 from truck routing and load planning, and Walmart says Self-Healing Inventory has saved more than $55 million.
Can other companies use Walmart’s routing technology?
Yes. Since March 2024 Walmart has sold Route Optimization to other businesses as software as a service.
About the curator & AI disclosure
Curated by Luciano Ayres
Luciano Ayres is Director of Artificial Intelligence at ArcTouch, with more than 20 years of experience in software and technology leadership and specialized studies in deep learning. Connect with him on LinkedIn.
AI-generated briefing. This case was produced through an automated AI research and writing workflow curated by Luciano Ayres. Sources are linked for independent verification; summaries and analysis may contain errors and do not replace professional due diligence.