How Sam’s Club taps AI to help run its stores more efficiently

Every morning around 8 or 10 at the Sam’s Club in Secaucus, New Jersey, a robotic floor-scrubbing vehicle named Roberta roams up and down the warehouse’s aisles.

Roberta is one of a fleet deployed at the company’s locations across the United States. But it does more than clean the floors: Roberta is tasked with using its computer vision algorithms to check the stock levels of items, ensure every display area is presentable, and flag anything that seems out of place.

The autonomous daily rounds are just one element of how Sam’s Club seeks to use robotics and algorithms to overhaul how its stores are run. Peter Rowe, Sam’s Club VP of technology, merchandising, and AI labs, eventually envisions Roberta’s data feeding a comprehensive system that links all of the different ways the company is currently using AI for forecasting and stocking decisions.

“The next thing we’re creating is basically what I would call our membrane,” Rowe told Tech Brew. “The data from Roberta—let’s say on Dunkin Donuts [coffee stock]—I want to use it again for replenishment at our distribution centers. I also want to use that same forecast, potentially, for merchandising decisions for seasonal items. That membrane is going to allow us to connect all of those different siloed decisions.”

Providing efficiency: The use of automated systems like these has become more popular in brick-and-mortar stores as they compete with algorithm-driven online competitors.

While parent chain Walmart scrapped its first fleet of inventory-checking robots in 2020 for efficiency reasons, Sam’s Club’s VP of retail product management, Todd Garner, said Sam’s Club was able to save resources by retrofitting sensor towers on the already-existing cleaning robots, though he said the store doesn’t track how many hours of human labor it saves this way.

“Our pallet-driven model makes this much more conducive to the inventory reading that we’re doing,” Garner said. “For us, it makes a ton of sense, and it’s providing a ton of value.”

He also confirmed to reporters who happened to be standing in Roberta’s path that the machine is trained to adjust course to avoid hitting shoppers—even tech journalists.

Forecasting football fans: Also included in Sam’s Club’s web of technology are algorithms that adjust demand for the more than 100 items produced on-site—like rotisserie chickens or baked goods—around localized events. Those events can include big sporting events or concerts—the Taylor Swift appearances at nearby MetLife Stadium last weekend may have been one, for instance—and they rely on store associates inputting data via one of the in-house apps used by staff.

“Our algorithms have to adapt so that we have enough rotisserie chickens or back ribs on [college] football Saturdays in the fall,” Rowe said. “We need to ramp up that production. But then we don’t want to carry it over. Because through the week, you’re not consuming that.”

Rowe said demand-forecasting algorithms first went into effect in the stores in 2018, but they’ve since evolved to the point where they are providing data on an almost hourly basis.

Apps for associates: In-store workers are on the front lines of many of the ways that Sam’s Club deploys AI. The company also gives each associate a natural-language tool that allows them to ask questions about location and stocking of items into their phones.

Another tool uses algorithms to plot the quickest route through the store for a worker fulfilling a curbside pickup order. The same demand-forecasting systems also help predict periods of increased orders.

Now that each of these systems are starting to take hold, Garner said the company’s next challenge is bringing them together into one comprehensive system.

“While we may start off with a very simple problem that is specific to that silo,” Garner said, “over time, as we continue to build out the technology, as we continue to have more AI brains as we continue to have more applications, then the opportunity exists to bring them together.”