How AI is spreading throughout the supply chain
AI is making companies swifter, cleverer and leaner
DELIVERING 25 PACKAGES by lorry or van might seem straightforward enough, but it is devilishly complex. The number of possible routes adds up to around 15 septillion (trillion trillion), according to Goldman Sachs, an investment bank. Integrating AI into the complex web of production and distribution—the supply chain—will have a bigger economic impact than any other application of the technology and affect a larger number of businesses, says Sudhir Jha of Infosys, a large IT company. McKinsey estimates that firms will derive between $1.3trn and $2trn a year in economic value from using AI in supply chains and manufacturing (see chart). Many firms are already using robots powered by machine learning to improve the running of their factories and warehouses. But AI will transform several other aspects of supply chains as well.
One is the unglamorous work of managing finances and paying suppliers. Just as Microsoft’s Excel spreadsheets changed finance departments, AI will make routine back-office work more efficient, says Morag Watson, BP’s chief digital-innovation officer. Some early adopters are starting to use AI to scan invoices and predict payments. Workday, a software firm, offers a financial-planning tool using AI to forecast which clients are going to pay late.
Another opportunity is to improve manufacturing through computer-vision systems that can inspect products on assembly lines and spot flaws. These systems are more accurate than humans, says Andrew Ng of Landing.AI, a startup that works with Foxconn, a big Taiwanese contract manufacturer, and others. Nvidia, a chipmaker, already uses computer vision to ensure that its chips are properly assembled.
Companies will also use AI to predict when their equipment might fail. This will benefit firms that operate large assets, such as airlines, oil firms, energy companies and industrial giants, where unexpected breakdowns come at a big cost. Companies can combine data on past performance with those generated by smart sensors on machinery (part of the much-hyped “internet of things”) to predict when a jet engine or a wind turbine is likely to fail, so they can do maintenance before that happens. America’s air force and defence department are working with C3 IoT, a startup, to scan maintenance logs and past technical problems for signs that aircraft are wearing out. Companies are also building “digital twins”—virtual representations of assets—to run simulations of how weather and other factors affect machinery.
Next year’s hits
Better predictions will improve inventory management and demand forecasting, too, freeing up cash and storage space. This is especially important for retailers, which often have very thin margins, says Chen Zhang, chief technology officer of JD.com, a Chinese e-commerce firm. In 2015 the cost to companies of overstocking was around $470bn and of understocking $630bn worldwide, according to IHL Group, a research firm. Amazon now has algorithms to predict demand for hundreds of millions of products it sells, often as much as 18 months ahead. Among the most difficult items are clothes, where the company must decide which sizes and colours to stock at which warehouses, depending on nearby buyers’ shapes and tastes, says Ralf Herbrich, Amazon’s director of machine learning.
Lineage, a firm that keeps food cold for clients such as grocers and restaurants, uses AI to forecast in what order items will arrive at and leave a warehouse, so that it can put the pallets in the right position. “I put the toothbrush by my sink because I use it three times a day, and my Christmas tree in the attic for a reason,” says Greg Lehmkuhl, Lineage’s boss, adding that using AI for smart placement has boosted efficiency by 20%.
AI is also helping firms track the movement of their goods. Most of the businesses in global shipping, from ports and lorries to container ships, have been technological laggards, so their customers never knew when their goods might show up. This is starting to change. Systems are getting better at routing items efficiently and predicting their arrival, and companies are investing more in them. To forecast arrivals, they can put sensors on shipments or design whole systems to use data like the GPS signals put out by lorries. Packages are also being routed more efficiently, with big potential gains. Jack Levis, director of process management for United Parcel Service (UPS), a package-delivery firm, says that for every mile that its drivers in America are able to reduce their daily route, the firm saves around $50m a year.
Goldman Sachs expects AI to bring logistics costs down by at least 5%, which could generate additional profits of $25bn over the next ten years. That would make a big difference in this cut-throat and low-margin business. It may also introduce new competitors who completely rethink old processes. “When you build a new jet, you don’t just put a jet engine on the Wright Brothers’ plane,” says Ryan Petersen of Flexport, a logistics startup. Many firms, including JD.com, are investing in AI-powered dronedelivery technology.
Now mighty Amazon is moving into the logistics business, piloting a service in Los Angeles for picking up packages from businesses and delivering them to customers, which puts it in direct competition with FedEx and UPS. The e-commerce giant has become “everyone’s competitor”, says Ibrahim Gokcen, chief digital officer of Maersk, a global shipping firm. “Everybody in the supply chain has a heightened awareness they have to up their game, in part because of the capabilities of Amazon,” says Rich Carlson of Savi, a smart-logistics startup. Amazon’s rivals may fret, but consumers will be pleased.
The Economist, 31 March 2018