Artificial Intelligence (AI) has been a buzzword in supply chains for over a decade. The promises of leveraging machine learning to deliver predictive and prescriptive analytics have been touted by supply chain software companies for at least that long. The reality is that you should be leveraging certain forms of AI in your supply chain – McKinsey & Company estimates that companies leveraging AI in their supply chain are gaining $1.3 to $2 trillion in economic value. At a higher level PWC estimates that AI could contribute $15.7 trillion to the global economy by 2030. These are significant numbers and do a good job to frame the impact and importance of AI within the supply chain.
It is no accident that experts feel AI within the supply chain will have far-ranging impacts. The sheer size and scope of global, and even regional, supply chains are enormous. Reducing inventory, even a few percentage points, across a supply network can free up significant amounts of cash and increase operating leverage. Adding to the picture the complexities of a global supply chain, and the number of different levers that can be pulled, allows AI with machine learning (for the purposes of this post we will assume that machine learning is a subset of AI) to do what it does best: run a huge volume of calculations and scenarios to drive insights and learnings.
There are a few reasons why AI has failed to deliver on the promise that it is the panacea to all supply chain challenges. While pulling live data from twitter about protests and typhoons in Asia and automatically rerouting supplies and pulling raw materials from other suppliers with no human intervention sounds great in practice, it is certainly challenging to execute in the real-world. First, the entire supply network needs to have live accurate data; for multinational companies that often means trying to pull live data from dozens of ERP/MRP systems that are decades old. As software is updated and fields get changed, these connections to legacy systems come crashing down. Second, the entire network needs to be implemented to truly see any gains. How can a supply manager adjust where we pull raw materials from if the entire network cannot communicate? Third, pulling together data from dozens of ERP/MRP systems across the globe takes years to complete and millions of dollars to implement, if it is even feasible at all. Typically, these types of projects never get fully implemented and the only real winners are the system integrators.
Lastly, the major challenge around implementing these global supply chain AI solutions is making sure that the data and insights that the solution provides are actionable and used. Periodic insights or analytics that are viewed at the end of each quarter are cumbersome to understand and execute on, have limited benefit, and often are rejected as impractical solutions.
When AI is effectively, efficiently, and successfully deployed most often there are two important characteristics: it seamlessly fits into a planners’/supply chain leaders’ workflow and it integrates easily back into existing systems of record such as ERP or MRP. If the solution is too complex or time-consuming supply chain professionals will not use it; eventually people will revert to the easier path which is often Excel based. If the solution lives in a bubble and provides insights only in that bubble (i.e. inventory of SKU-XYZ/location-ABC needs to be increased 17%) but has no way of tying that action back into a procurement system, as an example, it will not be a part of the day-to-day workflow.
AI is at its best when it can take a huge quantity of data, identify trends, and make recommendations and predictions. When that information is used to empower a smart inventory analyst to make the right decisions at the right time, we can see significant improvements in supply chain metrics, i.e. better fill rates with reduced inventories across the entire supply chain network. It is still important to have a human being with their fingertips on the controls. The recent COVID-19 pandemic is a good example of where human intervention is important; AI is not likely to predict a plant shutdown due to a black swan event, but it can augment decision making around this type of event. AI can help you quickly run scenarios and predict the impact of these types of events as they are happening. Along with that, solutions that leverage AI can help executives run numerous what-if scenarios around supply chains so that companies can have a better understanding of both risks and opportunities within the supply chain.
Intuitive workflows are key when leveraging AI in practice. AI should augment human decision making through identifying trends, surfacing problems early, and managing by exception. The end goal at a tactical level is to stay ahead of the curve and stop fighting fires. At a strategic level we can drive huge and significant improvements throughout the supply chain. We can dynamically predict what inventory levels should be today and in the future by location and by SKU. We can begin to prescribe changes to inventory levels ahead of indicators a human might be able to detect.
Supply chains will continue to grow in complication and sophistication. The Amazon Effect has already begun to bleed into non-consumer goods – companies will expect goods to be delivered within a day meaning that there could be a rise in local distribution centers as well as Vendor Managed Inventory (VMI). The good news is that AI is well equipped to handle this new world. Dynamic and predictive inventory stocking by SKU/location combination will be seen as a “must have” and no longer a “nice to have” as companies realize real time changes in demand and volatility throughout the supply network. Corporations that do not adopt these practices will quickly find stocking locations completely stocked out or with huge amounts of excess inventory.
As we continue to see the growth of AI software solutions integrated into existing workflows that empower smart humans we feel that there will be significant improvements to supply chain metrics (increased inventory turns, lower inventory levels, increased fill rates, etc.) while being able to deliver both raw materials and finished good in a more timely manner.