How Walmart is increasing delivery speed using AI models

Walmart wants to get faster at shipping, sorting, and delivery.

The retail giant has been building its e-commerce business for years, powered largely by its ability to deliver groceries almost anywhere in the US within hours. Walmart’s global e-commerce business grew 26% in Q1. In the US, online sales were also up 26% thanks in part to Walmart shipping out orders at high volume and growing its marketplace.

Walmart is using what it calls “intelligence layers,” which sit at the heart of its AI models, to better predict what customers are going to order and when, and put the right product in the right part of its supply chain so it gets to people faster.

Experts say advances in large language models (LLMs) have made this possible, particularly Walmart’s improved ability to run predictive analysis that helps with things like supply chain planning.

During Walmart Associates Week, Suresh Kumar, Walmart’s global CTO and CDO, said the company is pouring investments into AI models that handle everything from inventory forecasting and order replenishment to scheduling, transportation, and route optimization.

Right place, right time: “AI is now this conductor of this orchestra or product movement,” Kumar told a group of reporters during a session on how Walmart is using AI and digital innovation to reshape commerce. “By having AI be able to understand all aspects of inventory movement…we can now be much better at not just predicting what customer needs are, but when do they need something? What is the best way to deliver for them?”

Kumar said Walmart has invested in AI models for its “selection algorithm,” “routing algorithm,” and “scheduling algorithm” to make complex real-time optimization decisions and determine the best way to deliver products as quickly as possible.

“Intelligence is where we are investing in all of these places,” Kumar said, “from forecasting to replenishment forecasting, ordering replenishment, scheduling, transportation, and route optimization.”

A lot of work has also gone into how Walmart manages the flow of inventory from manufacturers all the way into people’s homes through the retailer’s AI models, Kumar said.

“We spent a bunch of technology resources on building custom models, and in also incorporating sort of the best of breed that’s out there to be able to predict as accurately as possible the trends and what customers are really shopping for, and how we believe that the sales of these products, these individual products by region, by area, are going to vary,” Kumar said.

Delivery models: With LLMs having a breakthrough moment last year, thanks to agentic commerce, retailers like Walmart essentially have “predictive analytics in a can,” Scot Wingo, CEO of Refibuy, an AI startup focused on solving problems for retailers using advanced AI frameworks, told Retail Brew. “It’s trained on things that machine language [models] couldn’t train on. That ends up being helpful.”

LLMs can now handle much more intelligent and complex tasks, Andy Szanger, director of strategic industries at CDW, a company that manages tech stacks for retailers, told Retail Brew.

“What AI is enabling them to do is have a better understanding of who their customers are, where they are, and anticipating what they’re going to buy,” Szanger said. “What AI will enable them to do is have a better understanding of what are the stores that are within that proximity, and which ones can handle the bandwidth that’s needed to have the right amount of employees to do the picking and efficiently get that product out the door quicker, and then also allow them to map in external factors such as weather and traffic to understand how long it’ll truly take them to get product to their customer.”

Ultimately, AI is giving retailers the ability to analyze their data at a scale and speed that simply wasn’t possible before.

“Similar to what Walmart’s doing with delivery, it’s allowing retailers to leverage their data and do analysis at scale and speed that they were never able to do before,” Szanger said.