Interview: Walmart’s David Glick discusses technology scale

Build or buy

Under the leadership of chief technology officer (CTO) Suresh Kumar, Glick says Walmart has transitioned to building software in-house rather than taking the route of sending out a request for proposals (RFP) and bringing in software providers.

He says many off-the-shelf products are unable to meet the scale the business requires: “We need to be able to customise exactly what we need, and it needs to be resilient and stay up 100% of the time.”

“Generative AI comes up every single day – you read about it on Twitter, you hear people say it’s going to change the world and reduce headcount by 30%. But at Walmart, we’re actually trying to use it”

David Glick, Walmart

Glick says Walmart has invested a lot of dollars and engineer time in moving off the mainframe into the public cloud and its own premises cloud. “We have a platform called the Walmart Cloud Native platform, which we use to abstract away which cloud we’re on, or if we’re on a mainframe, or anything else, so that we can deploy software quickly,” he says. “That’s allowed us to move much more quickly with advancing our technology.”

One of those technologies is artificial intelligence (AI).

Looking at the noise across the industry surrounding AI, Glick says: “Obviously, generative AI comes up every single day and you can read about it on Twitter. You can hear people say it’s going to change the world and reduce headcount by 30%. But at Walmart, we’re trying to do is actually use it.”

Glick says that within Walmart’s enterprise business services team, there is a two-pronged approach to generative AI.

“One of the things we want to do is go from the top down and pick an application to see how it goes,” he says. As such, the company has developed an application to help augment the employee benefits helpdesk. It works by listening to employee questions and then helps the helpdesk agent provide answers. “Who’s better at memorising a 300-page benefits guide? Is it a computer or is it a person?”

Glick says the business is also taking a bottom-up approach to generative AI by “crowdsourcing ideas for generative AI”. Walmart has introduced an AI assistant as part of an app, providing what Glick sees as a “safe and IP-protected way” to try things out. “We can monitor those queries and monitor what they use the AI for, which tells us what’s important to our users,” he says.

This informs the enterprise business services team at Walmart on what purpose-built functionality to focus on. It has found that security and protecting the company’s intellectual property ( IP) are “super important”.

Walmart uses both public cloud providers and its own in-house generative AI technology stack. “It’s paramount for us that our user data or customer data and our IP stays within our firewall and isn’t used to train other datasets, so we spend a lot of time figuring out how to do that,” he says.

This effectively means Walmart needs to work very closely with its public cloud providers to ensure commercially sensitive data remains within the company’s network. But the company also runs its own graphical processing units (GPUs) for AI within Walmart datacentres.

Over time, Glick believes public cloud providers will have to come up with a way to protect IP out-of-the-box. “We’re right at the cutting edge of a generative AI, so we’re working very closely with both of our cloud providers to make sure that between what we have in guardrails and what they’re doing, none of our IP leaks into the public domain,” he says.

Perhaps the most interesting takeaway from the conversation with Walmart’s Glick is his acknowledgement that the industry does not have all the answers. Off-the-shelf software is unable to scale to the level required by Walmart, and the limitations of public cloud providers’ generative AI offerings means Walmart has had to build and manage its own enterprise systems and AI infrastructure.

Listen to the podcast >>