Supermarket Chains Are Using AI To Keep Groceries and Supplies Stocked

The supply chain was already complex but has become more so as borders close, migrant workers are prevented from harvesting, stores and manufacturers close as workers get sick and regional changes in virus impact shift needs and priorities. It is the perfect job for AI. JL

Jared Council and David Uberti report in the Wall Street Journal:

In late January, AI began detecting coronavirus-related shocks to food supply chains across the Asia-Pacific region. Algorithms respond to daily fluctuations in customer behavior. by the end of January. They started to move supplies as close to production sites as possible. AI added new information to shore up algorithms that predict food supply and demand. Manufacturers have raced to feed AI tools with new data to help teams shift materials as needed and keep factories running. Distributors have turned to model where to direct fleets of trucks and trains that keep store shelves stocked.

In late January, the artificial intelligence startup Noodle.ai began detecting coronavirus-related shocks to food supply chains rippling across the Asia-Pacific region.

It was the early stages of a pandemic that has put such systems—increasingly seen as key to keeping the world fed—to the test.

Software firms like Noodle.ai have urgently added new information to shore up algorithms that predict food supply and demand. Manufacturers such as global food giant Danone S.A. have raced to feed AI tools with new data to help teams shift materials as needed and keep factories running. Distributors have turned to the machines to help model where to direct fleets of trucks and trains that keep store shelves stocked.

Despite shortages of toilet paper, hand sanitizer and highly sought-after products, these efforts have helped keep many food supply chains functioning, helping divert food meant for restaurants, airlines and other industries toward grocery stores.

“We’re seeing it as a rallying cry to help the world’s supply chains run,” said Noodle.ai Chief Executive Stephen Pratt, whose firm shortened the forecasting windows for its algorithms from months to days. “There’s nothing more important right now.”

Michael McGowan, vice president of merchandising insights and activation for Kroger -owned data firm 84.51 LLC, said the crisis has required his team’s human analysts and algorithms to respond to near-daily fluctuations in customer behavior, like increased weekday shopping or preferences for bulk packages of a particular product.

“To be honest with you…it’s been a frantic time in everybody’s career,” said Mr. McGowan, whose firm helps Kroger’s teams stock more than 2,700 supermarkets across its brands. “There’s no previous model that can just be dusted off and say ‘let’s review that’ at this time.”

Blue Yonder Inc., whose software helps project demand for products like dairy and produce, started integrating Covid-19 death statistics from the Centers for Disease Control and Prevention into its systems in late February to help them measure the scope of the outbreak and make better predictions as the crisis evolves. The Arizona-based firm, which counts clients like Albertsons Companies Inc. and Loblaw Companies Ltd., also established virtual “war rooms” with about 50 of its customers to monitor fast-changing facts on the ground.

“People might get sick and you might have to shut down a warehouse,” said Uwe Weiss, executive vice president of machine learning at Blue Yonder.

One such closure by a supplier or manufacturer could reverberate up and down a particular supply chain. Some retailers and food manufacturers across the U.S., including a number of meat plants, have already closed shop or reduced hours to prevent employees from getting sick.

Katharina Stenholm, senior vice president and chief cycles and procurement officer at Danone, said her company used AI and machine-learning tools to identify such pressure points and shift resources from other areas.

Ms. Stenholm, who oversees a team of 1,000 people from her office in Amsterdam, said her group had a global response team in place by the end of January. They started to move supplies as close to production sites as possible. “I was very afraid logistics could get in the way,” she said.

“Not a single factory idled for lack of materials,” Ms. Stenholm said. “The bottlenecks were complex, but manageable.”

The company already was investing in data and AI-driven tools to manage its supply chain. “We have been working on advanced analytics, machine learning, all of those projects are now well-appreciated,” she said.

But with little data shared between commercial and consumer-facing supply chains, farmers and manufacturers could face more long-term pressures as canceled orders from one sector aren’t immediately replaced by new orders from another, said Richard Tiffin, chief scientific officer at the U.K.-based data analytics firm Agrimetrics.

Food production in key countries like Italy and Spain could plummet in the coming months, Mr. Tiffin said, and travel restrictions might squeeze the migrant labor that harvests many of the world’s crops. Each could add additional wrinkles for the AI-powered tools that underpin food supply chains.

Mr. Tiffin, an agricultural economist at the University of Reading, compared the complexity of that system today to that of the global financial system leading up to 2008.

“They work, but we’ve got no idea how those systems work,” he said. “And as we know, it’s a characteristic of those sorts of systems that small perturbations can cause really big, damaging consequences.”