AUG 10, 2022
How Artificial Intelligence could still save the Christmas business
NOV 03, 2021
It does not look like the gaps in the supply chains that have emerged over the past two years will be closed anytime soon. That is why companies are turning to a new generation of AI-supported simulations. The so-called digital twins are supposed to help them deliver goods and services to customers on time. These tools not only predict interruptions, but also suggest what to do about them. Desperate companies grappling with the breakdown of just-in-time shipping use them to strike the critical balance between efficiency and resilience.
The list of things that have been hard to come by in the past few months is as varied as it is long: new cars, new phones, contact lenses, cleaning supplies, fresh produce, garden furniture, books, the color blue. “It’s not like when everyone ran out of toilet paper in March 2020,” says Chris Nicholson, founder of Pathmind, a company that uses AI to solve logistics problems. “This time the missing articles are far more personal.”
“It’s fair and fair to say that no matter what you’re selling, you have a problem right now,” said Jason Boyce, founder and CEO of Avenue7Media, a consultancy that advises on top Amazon sellers. Boyce says he has customers who would turn over millions of dollars annually if they had the goods they wanted in stock. “We have conversations with customers every day who just cry,” he says. “Your warehouse was not full for months – 30 days in a row.”
Digital twins are trying to fix disruptions in the supply chain by predicting them before they happen and then using AI to find another path. The term “digital twin” stands for the basic idea of ??simulating a complex system in a computer and creating a kind of twin that reflects real objects – from ports to products – and the processes associated with them. Simulations have been instrumental in decision-making in industry for a number of years. They help, for example, to check different product designs or to make a warehouse more optimal. Since large amounts of real-time data and computing power are now available, more complex processes can now be simulated for the first time, including the mess in global supply chains with their numerous providers and transport networks.
This type of technology has given Amazon, which has been working with its own trucks and warehouses for years, an extra edge. Now other companies are also taking advantage of this technology. Google is developing digital twins for the supply chain that automaker Renault has been using since September. International shipping giants like FedEx and DHL create their own simulation software. And AI companies like Pathmind create bespoke tools for anyone who can pay for them. But not everyone will benefit from it. Rather, the powerful new technology threatens to widen the digital divide in the global economy even further. It is easy to blame the pandemic for the current problems in the supply chain. But the truth is, the pandemic only made a bad situation worse. To get this situation under control, trillions of dollars must be poured into global infrastructure, ports and delivery fleets expanded, and investments in better management, better working conditions and better trade deals.
“Technology won’t solve these problems. It won’t allow ships to move more containers,” said David Simchi-Levi, who heads the data science lab at the Massachusetts Institute of Technology and who helped develop digital twins for several large companies Has. But AI can help companies weather the worst. “Thanks to digital twins, we can spot problems before they occur,” he says. A lot of effort – a lot of profit Thanks to the transparency of the supply chain, Amazon can, for example, predict when an item will arrive at a customer. For every item that Amazon ships itself – and that includes the millions of items it ships out on behalf of third parties like Boyce and its customers – there is an accurate estimate of when it will arrive. That might not seem like a lot, Boyce says, but if Amazon got those predictions wrong, it would lose customers – especially around the holiday season when people buy gifts at the last minute and trust Amazon to deliver them. “Just this little thing, just specifying a delivery date, requires an enormous amount of computing power,” he says. “But people freak out if they don’t get their things on time.”
According to Deliverr, a US company that manages delivery logistics for several e-commerce companies like Amazon, Walmart, eBay, and Shopify, an estimated delivery time of two days versus seven to ten days increases sales by 40 percent; an estimated delivery time of one day increases it by as much as 70 percent. So it’s not surprising that other companies want such a crystal ball too. Just-in-time supply chains are as good as dead. The problems of the past two years have ruined many companies that have gone too far with efficiency. Storage space is expensive, and in times of abundance it may seem extravagant to pay for it when you may not need it for a week. But if in the following week no new goods have arrived, you can no longer sell anything. The right balance “Before the pandemic, most companies were focused on cutting costs,” says Simchi-Levi. Now they are ready to pay for resilience (the ability to adapt to the circumstances, for example by always having goods in stock). But just concentrating on that is also wrong: you have to find the right balance between the two. That is exactly the strength of simulations.
“We’re seeing more and more companies stress testing their supply chains using digital twins,” he says. By exploring different possible scenarios, organizations can find the best balance between efficiency and resilience for them. With the help of deep reinforcement learning, in which an AI learns through trial and error what measures to take in different situations, digital twins become machines that explore what-if questions. What if a drought breaks out in Taiwan and the scarcity of water brings microchip production to a standstill? A digital twin could predict the risk of such an event, understand the impact on a particular supply chain and – using reinforcement learning – suggest what measures should be taken to minimize the damage.
Natural Language Processing