Kumar’s role at the global retail giant includes overseeing technologists at Walmart US, Sam’s Club, and Walmart International. Kumar is also an executive vice-president and the chief development officer at Walmart. 

In an interview with Mint, Kumar, who graduated from the Indian Institute of Technology-Madras, speaks about how Walmart combines technology advances such as artificial intelligence and generative AI with its ‘adaptive retail’ strengths. Edited excerpts:


Can you describe specific instances where Walmart has successfully implemented advanced solutions to boost operational efficiency and customer satisfaction?

Every aspect of the industry is undergoing a dramatic transformation. Earlier, people used to go to a (physical) store to find products. Then with e-commerce, you could type a few keywords and be presented with a list of products. Now with things like social commerce, like ‘Shop with Friends’, there is an aspect of product discovery that extends far beyond traditional ecommerce search. 

(Walmart’s ‘Shop with Friends’ feature enables virtual fitting rooms that can be shared across messaging and social platforms such as Pinterest and Instagram.)

In India, payments and identity (read: Aadhaar) have had a huge impact on commerce. Supply chain is dramatically evolving as well. It’s not just about drones and self-driving cars, but also about how the product itself—all the way from manufacturing—gets delivered into our customers’ homes, and sometimes even into their refrigerators. 

At Walmart, you use technologies ranging from augmented reality, robotics, digital twins, AI and now GenAI. How do you connect the dots for the customer?

The thread that brings it all together is adaptive retail—it is about how technology adapts to what the customer needs. Take social commerce or GenAI—it’s about reducing the friction for a customer as part of their shopping mission. Most of what we are trying to do with technology is eliminate the friction out of the customer experience in terms of finding, transacting and receiving products.

What’s the process you adopt to enhance customer experience with the help of analytics and AI?

Our process puts the customer right at the very centre. It’s how we start every meeting. We don’t start with, ‘Hey, here’s a nice, cool technology, let’s go figure out if it is going to help the customer or not’. Instead, we look at customer usage patterns and directly talk to customers and get information about their likes and dislikes. 

We have also got a huge wealth of data… about customers and their shopping habits and preferences online, as well as when they walk into the store or pick up their deliveries. We gather all of this data (from multiple channels) and make use of some AI models we have built in-house, and also leverage the best (large language models, or LLMs) that are available in the market. 

We train (these LLMs) on our data and we build our own custom models on top of it so we can gather insights out of that. From those insights, we start designing the experiences based on what we see as the most important problems that we need to solve for the customer. 

What we essentially want to do is to continuously keep closing the loop between what the customer wants, what’s the best technology to solve that problem, implement the technology, go back and see whether it has helped or not, and then keep refining it.

Can you give us an example of how this process is helping customers?

Consider the example of our home-delivery, which we started as a relatively small project or proof-of-concept, where a customer places an order, following which we send an associate to deliver the product not just to your home but inside your home and in your refrigerator. This calls for a lot of customer trust. 

We also have to understand substitution preferences. Let’s say you had placed an order for orange juice but we may have run out of stock due to a spike in demand, and maybe you want a substitution. So we listen, we go back and we fix the problem. It’s an ongoing process, which helps us enhance the customer experience.

How is GenAI helping Walmart increase productivity, efficiency, and efficacy of the conversations?

With GenAI, we focus on four broad areas. One is around customer experience. The second area is for a lot of our internal applications. We recently launched a product that helps our associates inside the store. It’s called Sidekick. It allows the store associate to focus on the next best thing that they need to do. 

The store is a very complex environment. People are running around and many tasks need to be done. GenAI, when combined with our in-house models, now not only identifies the perfect associate for a specific task but also assigns the task to the associate. The ability of AI and GenAI to understand context is helping our associates become a lot more effective.

The same thing is happening on the corporate side as well. If you look at our supply chain, we have built models that understand demand and are able to optimise the inventory flow. Our supply chain is very dynamic. When demand spikes in one area, we are able to now automatically figure out the most optimal way to divert inventory there. 

The third area is on the creative side, whether GenAI is creating personalised advertising and marketing campaigns for our customers so that we can understand what the customer needs are, or getting deeper insights into our product catalogue. 

Fourth, we are also using GenAI to build our own tools such as for better testing, better operations, and understanding how our systems can be scaled up and down.

That said, with GenAI tools you also have to take care of data privacy, security, and things like hallucinations.

Yes. One of the fundamental pillars is trust. The key is to make sure that any technology we deploy is going to be built on the trust that our customers and our associates have placed in us. 

We don’t pluck LLMs from the wild and let them loose on our customers and associates. We have built an in-house platform where we build data privacy and security safeguards to protect our customer data and other sensitive data since we also run one of the largest pharmacies in the world.

We also do not want to compromise our supply chain data. 

Second, we have built an internal layer which helps us select the LLM that is most appropriate for the task we are trying to perform. For instance, we’re in the process of rolling out a pilot that will help our legal department sift through the complex regulations we have to deal with. A general purpose LLM will not help us in this case. This also helps us address hallucinations.

What role do your India centres play in implementing emerging technologies such as AI and GenAI?

India is a great place for engineering talent of different flavours—from software engineers to data scientists to product and UX (user experience). We have grown our presence in Bengaluru, and opened another location in Chennai during the pandemic. Our India presence is one of the largest groups of technologists anywhere around the world. But we work as one global virtual team. 

Take GenAI as an example. There’s a very strong team here in our IDC (Indian Development Centre), which works very closely with our teams based in Hoboken (New Jersey) and Sunnyvale (California). Their sole focus is to generate the best search experience using LLMs.

Are there any synergies between your other businesses such as Flipkart, Myntra or PhonePe where you’re sharing technology expertise across units?

Our subsidiaries in India are largely self-sufficient. But since we are majority owners (in these units), we continuously keep looking at places where it makes sense to share information, best practices, etc. 

Teams in Flipkart, as an example, work very closely with our teams in our IDC and the US to share learnings, for example, in the area of GenAI.

What about the role of automation and robotics in enhancing productivity and customer experience?

We have announced a large set of investments to automate large parts of our supply chain. This is important as we plan to build a supply chain of the future, or, more broadly, our retail future. 

The focus is to use automation, whether it’s in the form of robotics or our automated storage and retrieval systems, to tackle difficult and repetitive tasks so that our associates can focus on what they do best—serve our customers with a human touch and deal with cases that need human intervention. 

We recently opened an automated storage and retrieval facility for our cold chain to make sure that the entire facility is chilled several degrees below freezing. This is not an ideal situation for most associates. Imagine having to lug big product cases while wearing cumbersome thermal clothing. 

This is where automation helps a lot, and also allows us to serve our customers a whole lot better, and reduce the time it takes to get the products to them. 

How do you address the sensitive issue of job losses due to automation?

We have discovered that our associates are thrilled that we have invested in automation because they see us doing two things. One is creating a much better working environment for them because it dispenses with the most dangerous tasks they disliked the most. And two, it has given them a path to be able to upskill.

So from a job, they are able to carve a career. I agree that overall the nature of the job is going to change (due to automation, AI and GenAI). But almost always, it seems to change for the better.

Given the incredible pace at which technology is moving, what is your advice to CXOs when implementing these technologies?

Technology evolves quite rapidly. And part of the challenge is to keep abreast of all these changes. So one has to keep three things in mind. 

First is to make sure that we have a team of skilled, talented people, and empower them. 

Second, it’s important not to get enamoured about new breakthroughs in technology for technology’s sake. It can be distracting, so ground yourself. The way we do this is by starting with the customer, seeing what the real problem is, and then exploring how technology can benefit them. 

The third thing is to start small. At Walmart, we have a number of proof-of-concepts where we test and learn. We didn’t switch over our entire search algorithm, for instance, just because we had access to an LLM. 

You need to understand the trade-offs and the risks, like regulatory issues and hallucinations. When the technology matures, you can do a wider rollout.


Source link