Using AI for managing the supply chain network

- Supply chains of tomorrow must be designed to respond to much greater uncertainty than was true even 10 years back. The numerous factors buffeting supply chains include, pandemics, demand volatility, raw material disruptions, labor shortages, geopolitical tensions, cyber attacks, terrorism and more frequent extreme weather patterns. As a result in designing and managing supply chains, the ability to recover quickly from unexpected developments has become as important as maximizing efficiency. Every supply chain is a network of organizations. Think about raw material mines, factories, trucks, ships, trains, ports, warehouses, all parts of a complex network. How does one ensure that such a network can be not only efficient, but also resilient and responsive? The answer lies in ensuring that the network exhibits the following properties. One, end-to-end visibility. Like control towers, network managers must maximize visibility into every node and every link, as well as the external factors that may cause turbulence in these nodes and links. Visibility helps ensure that unexpected changes anywhere in the system, get anticipated, recognized, and if needed, acted upon with minimum delay. Two, agility. When unexpected changes do occur, network managers must be able to figure out the best response and act on it rapidly. To create such a network, leading edge companies are starting to build digital twins of their supply networks, deploying AI to manage the digital twins and using the output to augment their decisions and actions. A digital twin refers to a simulation of the physical network in a computer, creating a replica that mirrors real world objects and the links between them. The digital twin integrates the company's internal private data, including IoT sensors with external public data, such as satellite images, weather predictions and political developments. Let's say you have a supplier in Wenzhou, China, the digital twin can automatically incorporate the risk of an incoming hurricane, something that will be hard for people to do when dealing with dozens or even hundreds of nodes and links. In this manner, a digital twin can anticipate problems before that happen. It can also use AI to recommend or even execute the optimal solution. While simulating reality is not a new idea, the value of AI models lies in their ability to decipher non-linear interactions among very large and constantly changing sets of data, both structured and unstructured. Similar to interactive games such as chess and Go, reinforcement learning algorithms, train specific AI models by playing millions of what if games on the multidimensional board of a digital twin. Once the model has mastered a particular supply chain network, it springs into action when needed. Let's say there is an accident at a parts supplier in Malaysia. The AI would not only alert you, but also suggest actions such as placing an immediate order with a distributor in California who happens to have inventory in stock. AI models can deal with such complexity far better than even the smartest supply chain managers. They also keep retraining constantly. Companies such as Amazon have been using the digital twin concept to manage their supply chains for some time. Now, tech providers, such as Google and Nvidia, have started offering these capabilities to other companies. Building on these ideas, you may want to analyze your company's supply chain as well. Who should be the members of a team to assess how to build a digital twin for your supply chain? How many tiers upstream could you go in building an excellent mirror of the physical reality?