Microsoft and the US Department of Energy grabbed headlines by announcing a partnership to develop AI-powered applications to help first responders reacting to natural disasters. One of the first AI prototypes in development will employ computer vision to detect and predict the frontiers of active wildfires and floods. The second application uses an AI tool that should be in every risk manager’s toolbox: simulation. This simulator will aid teams in running mock scenarios to better plan and prepare for the next big natural disaster.
Many companies were caught flat-footed by COVID-19, perhaps the largest disruption to global trade in a hundred years. While they scramble to realign their supply chains to meet the reality of 2020 and beyond, now is precisely the time for these organizations to consider AI-based risk management tools, from cutting-edge predictive analytics techniques to the tried-and-true methods of simulation and optimization.
Mathematical simulation and optimization form a powerful combo that helps create lean, cost-efficient supply chains that are also resilient to disruption. The first step in integrating these technologies into your organization’s operations involves digitizing the supply chain, often referred to as a digital twin. This digital twin of your physical supply chain should detail contingency options of all kinds, such as alternate suppliers, lead times, inventory levels, site-specific redundancies, bills of materials, and even business continuity plan documents (BCPs). This step can be taken for granted and should be collected before, not during a crisis.
Next, simulate full or partial outages: customer regions being flooded, fires disabling production lines, etc. You can ascribe probabilities to these scenarios by obtaining relevant external data such as hundred-year flood maps and county-level natural disaster records.
After using these scenarios to illuminate any vulnerabilities in the supply chain, the next step is to ensure reasonable protections against the worst weaknesses. In one study, Ford found that it was not necessarily the most costly vehicle components that presented the highest risk, but rather a small, overlooked group of critical components like o-rings and valves. Strengthening these areas of weakness might involve onboarding a new supplier or carrying more safety stock inventory.
As COVID-19 demonstrates, we cannot assume that disruptions are short-lived or that any resulting changes will be temporary, even for supply chains that have been carefully designed and optimized to balance cost and risk. When something inevitably goes awry, having a ready-made contingency playbook can be a game-changer. And if the disruption takes a more permanent form, it’s a good idea to re-optimize the supply chain to ensure that your operations are cost-effective, whether that disruption has resulted in undersupply or oversupply.
There are many risks in this world that are truly unpredictable, especially in complex global supply chains. Nonetheless, risk and supply chain managers would do well to intelligently incorporate techniques from the ever-advancing field of predictive analytics into their toolboxes. For example, part failure can be predicted on a manufacturing line, in a truck engine, or in equipment at a customer site. Ryder Logistics and other fleet operators have deployed predictive maintenance software to reduce asset downtime. What other risks can be predicted? The spread of natural disasters such as hurricanes and flooding, poor crop yields, late delivery, capacity utilization, or even financial performance of a supplier or customer. Furthermore, better predictive models make simulation and optimization models even more useful. If a fleet operator improves a probabilistic predictive parts failure model, he or she can optimize operations with more certainty or run simulations with more realistic scenarios.
As we adjust to the new normal brought on by COVID-19, it’s worth taking a fresh look at how AI can boost supply chain resiliency, whether that’s creating a digital twin, building predictive models, or adding simulation tools to prepare for whatever may come your way.
About the Authors
Nate DeJong is a Senior Consultant with LLamasoft. He has spent seven years applying algorithms to solve a variety of supply chain problems from forecasting to supplier risk. He spends significant time across business leadership, data science and engineering teams with the goal of delivering analytically sound solutions across organizations.
Colleen Eland is an Engagement Manager with LLamasoft. She has 18 years of diverse global experience in fast moving consumer goods companies. Her roles have involved helping to increase scale in supply chain, procurement, customer marketing and sourcing strategies for many fortune 500 companies.
Sign up for the free insideBIGDATA newsletter.
Speak Your Mind