With supply-chain disruptions over the last two years showing no signs of abating anytime soon, businesses are turning to a new wave of AI-powered simulations known as digital twins to help them deliver goods and services on time. These technologies not only forecast future disturbances, but also advise on how to avoid them. Desperate businesses grappling with the demise of just-in-time shipping are turning to them to strike a critical balance between efficiency and resilience.
The following items have been difficult to obtain at some point in the last few months: new cars, new phones, contact lenses, cleaning products, fresh produce, garden furniture, books, and the color blue.
“It’s not like when everyone ran out of toilet paper in March 2020. This time the missing items feel personalized.”
– Chris Nicholson, creator of Pathmind
COVID-19 has shed light on a wide range of global networks, from the internet to international air travel. But the supply chains that crisscross the globe – the ships, trucks, and trains that connect factories to ports and warehouses, transporting practically everything we buy thousands of miles from where it’s made to where it’s consumed – are under more scrutiny than ever.
“It’s fair to say that whatever you’re selling, you’ve got a problem right now. We’re having talks with clients every day where they’re just crying. For months, they haven’t been fully in stock for one 30-day period in a row.”
– Jason Boyce, Founder and CEO of Avenue7Media
Digital twins aim to tackle supply chain breakdowns by anticipating them and then utilizing artificial intelligence to find a remedy. The name encapsulates the central concept of mimicking a complex system in a computer, resulting in a kind of twin that replicated real-world objects – from ports to products – and the processes they are a part of. For many years, simulations have been used in industry to assist workers explore new product ideas or streamline the layout of a warehouse. However, the availability of vast amounts of real-time data and processing power means that more complex processes, such as the chaos of global supply chains that frequently rely on several vendors and transportation networks, can now be replicated for the first time.
For years, this type of technology has given Amazon, which already has the advantage of owning its own trucks and warehouses, a competitive advantage. Others are now embracing it as well. Google is developing supply-chain digital twins, which Renault said in September that is has begun utilizing. International shipping conglomerates such as FedEx and DHL are developing their own simulation software. And AI companies like Pathmind are developing bespoke tools for anyone willing to pay for them. Nonetheless, not everyone will benefit. Indeed, the powerful new technology has the potential to deepen the global economy’s widening digital divide.
Withstand the storm
It’s simple to blame the pandemic for the current supply-chain issues. Factory closures and manpower constraints disrupted production and delivery hubs at the same time an increase in online shopping and comfort buying boosted demand for home deliveries.
But, in reality, the pandemic exacerbated an already dire condition.
“There are global forces driving this, all combined into a perfect storm.”
– D’Maris Coffman, Economist at University College London
To weather this storm, trillions of dollars will need to be invested in global infrastructure, increasing ports and delivery fleets, and investing in better management, better working conditions, and better trade accords.
“Technology is not going to solve these problems. It’s not going to allow ships to carry more containers.”
– David Simhci-Levi, Data Science Lab Lead at the Massachusetts Institute of Technology
However, AI can assist businesses in surviving the worst of it.
“Digital twins allow us to identify issues before they happen.”
– David Simhci-Levi
According to Hans Thalbauer, Google’s managing director of supply chains and logistics, the biggest issue firms confront is an inability to foresee events farther up the supply chain.
“It doesn’t matter which company you talk to, everyone in the supply-chain world will tell you they don’t have the visibility they need to make decisions.”
– Hans Thalbauer
It is supply-chain visibility that enables Amazon, for example, to estimate when an item will arrive at your door. Every item that Amazon delivers itself – including the millions of things it distributes on behalf of third-party sellers like Boyce and his clients – comes with an exact delivery estimate. According to Boyce, it may not seem like much, but if Amazon gets these predictions wrong, it will start losing customers.
According to Deliverr, a US company that oversees delivery operations for a variety of e-commerce companies such as Amazon, Walmart, eBay, and Shopify, an anticipated delivery time of two days vs seven to ten days improves sales by 40%; an estimated delivery time of one day increases sales by 70%.
It’s no surprise that others want their own crystal ball. Justin-in-time supply chains are on the verge of extinction. The disturbances of the last two years have sunk numerous organizations that have pursued hyper-efficiency to its logical conclusion. Warehouse space is costly and paying to hold product that you may not need for a week may appear lavish in times of plenty. But, if next week’s shipment does not arrive, you will have nothing to sell.
“Before the pandemic, most companies were focusing on cutting costs.”
– David Simchi-Levi
They are now willing to pay for resiliency but focusing solely on resiliency is also a mistake: you must strike the correct balance between the two. This is where simulations truly shine.
“We’re seeing a growing number of companies starting to stress-test their supply chains using digital twins.”
– David Simchi-Levi
‘What If’ Scenarios
Companies can find the greatest mix of efficiency and resiliency for them by experimenting with various scenarios. Add deep reinforcement learning, which allows an AI to learn what behaviors to perform in diverse contexts though trial and error, and digital twins become machines for pondering what-if scenarios. What if Taiwan suffers from a drought, causing a water deficit that prevents microchip production? A digital twin might forecast the likelihood of this happening, track the impact on your supply chain, and recommend wats to reduce the harm using reinforcement learning.
If you’re a car manufacturer in the Midwest of the United States, a digital twin might advise you to acquire extra components from a distributor on the West Coast that still haws surplus. Yet, when many instances are combined, the situation quickly becomes extremely complex. For example, according to Simchi-Levi, Ford has more than 50 plants worldwide, using 35 billion parts to build 6 million vehicles each year. Ford interacts directly with around 1,400 suppliers distributed across 4,400 manufacturing facilities, and there is a stack of suppliers and suppliers’ suppliers up to 10 layers deep between Ford and the raw materials that go into its vehicles. Any of those links could fail, and a thorough stress test would need to examine each one.
To perform simulations and train AIs, digital twins use as much data as feasible. There includes logistical information on the company and its suppliers, including inventory and shipment data. Then there’s customer behavior data based on market research and financial estimates. And global data, such as geopolitical and socioeconomic trends. Simchi-Levi has even used social media data to forecast people’s behavior, particularly during the pandemic.
Google’s digital twin, which takes into consideration global weather trends, may be loaded into Google Earth. Thalbauer explains that if you are a vegetable farmer in California, you may run simulations to identify which of your fields are at risk from La Nina. When Google creates a digital twin for a customer, such as Renault, they can select which of the several data sources to incorporate.
Pathmind takes a more light-weight approach. Its digital twin simply wraps around a company’s existing supply-chain management systems, utilizing the data that they already generate. It then augments this data by executing what-if simulations and incorporating the resulting synthetic data into the pot in which it trains its AI.
A digital twin can learn to adapt to previously unknown situations, including global pandemics, with the correct synthetic data.
“This is where we get into the whole secret of ‘Why is AI smart?’, it lives more than we do, in these many different worlds, some of which have never existed before.”
– Chris Nicholson, Founder of Pathmind
This technology, in theory, can benefit anybody. There will be winners and losers.
“Digital-twin technology presents a powerful opportunity for companies of any size.”
– Rick Lazio, Lawyer, Former US Congressman & Senior Vice President at Alliantgroup
However, he observes that larger organizations, which are already the most protected from losses, are the first to use this technology.
Lazio believed that many smaller enterprises would require assistance, possibly in the form of government investment, to avoid falling behind.
“Companies that adopt technology early see benefits greater than the sum of its parts.”
– Rick Lazio
Simchi-Levi is more upbeat. Many firms used to believe that establishing a digital twin would require a large investment and years to pay for itself, but that is no longer the case: a million dollars and 18 months can provide many of the benefits.
Simchi-Levi is confident that the hype surrounding digital twins will continue even after the worst of the current problems have passed. He predicts that if it isn’t the pandemic, it will be something else. The last few years have taught firms how to properly plan for and compete.
“When we go back to normal, it won’t be the same as before. The pandemic proved that the future is here.”
– David Simchi-Levi
Information From MIT Technology Review