- Walmart is leading retailers in adopting artificial intelligence and machine learning, but the world's largest retailer still runs into cultural issues that undermine the push to implement the advanced technology.
- The company currently has roughly 1,500 data scientists, according to chief data officer Bill Groves, and is hiring more, including a role to develop voice-activated shopping applications.
- Walmart has three core questions that guide all of its AI and machine-learning projects. If the answer to any of them is "no" then the initiative is shelved immediately.
- The retailer just opened up a new warehouse completely ran by robots that can fulfill online grocery orders 10 times faster than store associates.
- Click here for more BI Prime content.
Walmart is a leader in the push to adopt artificial intelligence and machine learning. This week, it unveiled a new robotic system called Alphabot that can fulfill online grocery orders roughly 10 times faster than store associates.
The Salem, New Hampshire, facility that houses the system has roughly 30 robots that can independently pick and pack items from around 4,500 total products. The aim is to significantly increase Walmart's ability to fulfill online orders, a key growth area for the company. Humans, however, still manually pick fresh produce, while the robots gather everything else.
Still, the world's largest retailer runs into many of the problems other organizations experience when pursuing the advanced technology.
The company currently employs roughly 1,500 data scientists throughout the enterprise, according to chief data officer Bill Groves, who directly oversees a smaller staff of 100 tech workers. Groves also said the company employs around 50,000 software engineers, though a Walmart spokesperson later said the numbers is closer to 10,000.
Those employees help support the over 100,000 different machine learning or AI-based projects the organization currently has in production. Among the applications that Walmart is currently rolling-out are AI-powered cameras to monitor for theft.
"I do more work in the AI and [machine learning] space then I have ever done in my life," Groves said at the MLOps NYC conference two weeks ago.
"We're involved in robotics, we're involved in micro-personalization, we're involved in probably the biggest supply chain in the world," he added.
And it's continuing to build-out that staff. Among the positions Walmart is hiring for is a data scientist to help develop voice-activated shopping applications. The company already uses the technology in grocery pickup and deliveries. Overall, Walmart has 67 data science openings, 43 software development vacancies, and 90 available data analytics jobs, according to its careers website.
But the success rate for artificial intelligence or machine-learning projects is still just 75 percent, Groves said at the event. One way Walmart is aiming to address that is through its core three tenets that guide all the high-tech initiatives.
"If the answer is 'no' to any of these three, we'll typically put a stop to the project immediately so that way we aren't spending money that we shouldn't spend," he said.
1) Why are you doing it?
One of the first questions that Walmart employees have to answer when deciding whether to pursue an AI-based project is: will someone pay for it? That can include the company itself, or a vendor that may purchase the application from Walmart.
Groves forbids what he refers to as "cool projects," or those that might be fascinating to pursue but produce no tangible benefit for the company.
"If nobody will pay for, then why am I doing it?," he said. "The business has to see the value, the business has to want it."
While that's a relatively simple question to answer, Groves says it "doesn't happen often enough." One key reason is the lack of communication across teams.
2) Can you explain it?
A problem that companies routinely run into is how to break down the organizational barriers between sectors. That encourages more collaboration between technology teams and those on the operational side of the business, like managing the supply chain.
Key to that succeeding, however, is software engineers or data scientists being able to explain tentative or pending applications to other business units. "If I cannot explain to an executive what I'm doing, then why am I doing it," Groves said.
Often those in departments like human resources think of problems from an analytical mindset, like how to make it easier to onboard new employees, and may not be as knowledgeable on the underlying technology. That poses a challenge to data scientists and engineers who are accustomed to outlining projects from a technical standpoint.
At Walmart, Groves knows an AI-driven project is "going to fail" if his team of data scientists and engineers discuss it and "the business isn't even really part of the conversation," he says.
3) Can you implement it?
While the tech teams are in charge of creating and managing the AI-based applications — like using cameras and sensors to help determine when shelves need to be restocked — the business side must be able to implement it to ultimately drive down costs or improve profits.
"It's a massive challenge just due to the size and scale we have," Groves said. "Money is being thrown out the window."
That's a key reason why communication between the teams is essential. A major impediment, however, is also price, particularly given the immense amount of projects Walmart already has in production. So initiatives must have a plan in place to go from development to production in an affordable way, and it must have buy-in from all parties involved.
"The data scientists talk to the business, they came up with an idea, they didn't include the business or the technology team with the implementation," said Groves. "They come back, they have a model that stands no chance of ever making it into production with the systems you have. Definitely not cost affordable."
While relatively basic questions, Walmart's approach to AI exemplifies just how critical cross-collaboration is for advanced tech initiative to succeed — and how quickly they can fail it there isn't broad support internally.