How AI is taking a centerstage at Walmart's security infrastructure

In one of the building basements of the global retail giant Walmart’s headquartered in Bentonville, Arkansas, large screen monitors constantly track anomalies, right from the moment of trucks to weather patterns to any disturbances in the market. It is one of the several strategically placed operations centres around the world, that operate in unison as a fully integrated system in prediction and prevention of security and other instances. Being the world’s biggest retail company, Walmart sees nearly 250 million customers every week in over 10700 of its stores and runs 19 websites. With many entry points of vulnerability in the age of generative AI, securing its data touch points is a critical aspect to its business with both information security teams and technology teams coming together for this. Just to illustrate the scale, nearly 2.1 million associates of the company at a given time are using its software in the stores additionally with the customers walking into a store logged on to its captive Wi-Fi network. At its distribution level, gig work force of nearly a million drivers use the app to deliver packages from stores to customers. Operating at a scale as big as Walmart, information security is an intrinsic part of its design and systems, and here agentic AI is playing a bigger role that ever before.

Blending Security with AI 

Walmart recently launched WIBEY, an agentic AI developer tool and platform for its software teams to collaborate and build newer functions as a next in line adoption of agents in its business transformation process. With 60 different security teams focused on things like applications, codes, network security, cloud tech, cryptography and more importantly the customer-facing and seller application, the company’s information security team is not just tasked with keeping Walmart systems safe for its associates or sellers, but also secure customer experience and also actively predict and prevent threats. Here AI is today playing a more critical role. Hari Vasudev, EVP & CTO, Walmart U.S says one big area where AI has hugely impacted is in regulating the bot traffic. Citing the example of how during the holiday season which ushers in new launches and higher traffic, it also encounters a higher bot traffic. “Typically, resellers try to buy as many items as possible in order to resell in the secondary market and obviously we want our customers to be able to buy the products first rather than resellers” he says. Here the information security team steps in to enforce security at the perimeter. Jerry Geisler, EVP & Global CISO, Walmart Inc says in the past few years the company has been adopting machine learning models into its security operations center, with AI integration making its entry in the last couple of years. With threat intelligence at the core - detecting threat actors, new emerging malwares and a ton of intelligence data coming it to an organisation, he says AI has allowed them to consume that feed much more efficiently, helping in getting to the root of what matters. “We're always engaging in real-time security posture adjustments and AI allows us to do that work quickly” Geisler says. 

Security threat at scale today as it stands, Hari Vasudev sees it like finding a needle in a haystack. With the world moving towards the use of more agentic systems, AI is turning out to be effective in determining the one needle in the haystack of agents that poses the threat. “What AI allows you to do is sift through information very quickly, and ultimately almost always it comes down to intelligent pattern matching, which AI does extremely well. So that's how we are hoping to evolve our systems to be able to detect the threat more effectively and eventually can move further up to be able to predict, looking at patterns of behavior” he says.

While artificial intelligence has aided in faster detection and working with large data inputs in security, “We are not at a point at nor is the world at a point where we are ready to hand it over (to AI) we still need humans in the loop. So, it does allow us to do quicker analysis, arrive at decisions maybe more rapidly, but we still very much have humans involved in all our security processes”, Geisler adds.

State v/s organised criminal groups as threat actors

Today with high-speed internet connectively widely available across regions, the threat landscape as the company sees very similar regardless of the geography. While intent of the threat actors may vary depending on the region but in terms of activity it remains fairly consistent regardless of where the threat is being observed. With threats where state actors or organized criminal groups are involved the former, as Geisler sees, are using state-of-the-art levels of technology which often have not been seen before. However, the organised criminal groups tend to follow some similar patterns which allows the company to better understand who they are dealing with. “Where state actors are concerned, you're not typically going to see what we call TTP (Tactics techniques process) you're not typically going to see those appear outside of the state environment”, he says.