At Transform 2019, Walmart’s VP of customer experience engineering Aanan Contractor and Walmart ecommerce’s VP, Lucinda Newcomb sat down with FEM CEO Rachel Payne to discuss Walmart’s latest forays into blending the online and offline worlds using AI and automation.

The fastest-growing illustration of this is grocery pick-up of orders made online; by the end of the year, the service will be available in 3,000 out of Walmart’s 5,000 stores.

“Think about a family,” says Newcomb. “You have two little kids strapped in the back seat, and instead of having to get them in a cart and push them around, you can just roll up and have somebody bring out everything you wanted.”

The goal is to enable Walmart customers to do this as efficiently and easily as possible. Their data shows that 65% of a customer’s order will be a repeat basket, re-stocking things they buy regularly. “It shouldn’t take, every week, 30 to 40 minutes to build that basket,” says Aanan. “How do we leverage voice commerce? Walmart has a very good partnership with Google in terms of launching our voice ordering system. It’s as easy as saying, hey, Google, talk to Walmart.”

If 65% of a basket is stocking regular items, Walmart wants to ensure it’s using the massive amounts of data it collects to help a customer find what they want, before they even know they want it. “What do we know about her?” explains Newcomb. “How do we make sure that we’re helping her realize this is something she seasonally does? Have we figured out that there’s an allergy in her family? Or maybe we’ve figured out that she’s very organic-sensitive and organic-focused. How do we help recommend other types of products specific to her needs? We’re awash in data. We’re constantly looking at how we can make sure we get to the most delightful experience possible.”

In building the models to make these recommendations and guide customers, accuracy has become a north star for the company. “Ninety percent precision confidence is very important in this mix for us to be really effective in terms of predictive basket,” says Aanan. “We don’t deploy the algorithm until it’s at least 90% confidence.”

That critical data is also used in inventory management to avoid ‘basket breakers’ where an item indispensable to a customer is unavailable. “We have all this other data to know which are the fast-moving inventory items,” says Newcomb. “These are the most likely to go out of stock. These are the ones that are seasonal. How do you use that to figure out what are the types of items that are at risk, as well as figuring out, which of those do we consider to be basket breakers?”

Yet, substitutions are a reality. “We have a great inventory management system, but at the same time, situations happen,” she adds. “We spend a disproportionate amount of effort on making sure we nail substitutions. They’re so infrequent relative to the broader basket, and yet when we have to substitute, it has such an impact on a customer’s NPS [net promoter score] and on her happiness with us.”

This is where the merger of AI and humans becomes essential. “In this smart substitution that we’re talking about, algorithmically, there’s a kind of place for these associates, these personal shoppers, who have known that customer,” says Aanan. “They know what’s really going to be accepted as a substitute for their customers. So how do we really add an aspect of a signal from our associates, these personal shoppers, into the algorithm, which is also taking input from the customer preferences and personal aspects? That’s a human aspect. Not just leveraging AI, but it’s really a combination of human intelligence and AI and merging them together.”

In the near future, those worlds are going to collide even more. Starting this fall, Walmart will be introducing in-home grocery delivery — not just to a customer’s front door, but right into their home, and into their refrigerator. There’s too much risk in leaving perishable items outside while a customer’s at work all day, so beginning in three markets — Pennsylvania, Florida, and Missouri — Walmart associates will be equipped with cameras (for security) as they deliver groceries right into customers’ kitchens.

“We’ve been in test for this, and anything with AI and personalization and in-home, people start to say, what about trust? What about the creep factor? Number one, we’re extremely careful about this. It’s always internal Walmart associates who’ve been there at least a year. They’re really well-trained. We have the cameras, so you can see what’s happening. At the same time, what we’re finding is that before Uber, would anyone say they’d get into a random car that drove up by the side of the road? Nobody would have trusted that. But now we find that customers just say, ‘Oh my God, somebody delivered my groceries and Marie Kondo’ed my refrigerator while they were at it. That’s pretty magical. I can’t go back to how I did things before.'”

It all comes down to how Walmart frames the mission of the customer journey. Their customers don’t want to go to the grocery store, they just want to feed their families, with a lot less stress and headache.

Watch the full video of the session and find all the other panels at this year’s Transform, right here.