Can Amazon’s ML-Powered AWS Supply Chain Solution Achieve Widespread Adoption or Is It Too Narrow in Scope?

Amazon Web Services (AWS) announced AWS Supply Chain, a new application that helps businesses increase supply chain visibility to make faster, more informed decisions that mitigate risks, lower costs, and improve customer experiences. This cloud-based inventory monitoring and forecasting application came into general availability last month after the application made its debut at the cloud giant’s AWS re:Invent conference last December. AWS Supply Chain automatically combines and analyzes data across multiple supply chain systems so businesses can observe their operations in real time, find trends more quickly, and generate more accurate demand forecasts that ensure adequate inventory to meet customer expectations. End users can set up a unified supply chain data lake using AWS Supply Chain’s built-in connectors to understand, extract, and aggregate data from Enterprise Resource Planning (ERP) and Supply Chain Management (SCM) systems. Lifetime Brands, Traeger Grills, and Whole Foods Market are some of the large end users that have already adopted the AWS Supply Chain solution.

The AWS Supply Chain platform promises unified data, actionable insights powered by Machine Learning (ML), and contextual collaboration features to reduce costs and mitigate risk, making it apt for retail supply chains. With omnichannel fulfillment still growing rampantly, there is a growing need among retailers to avoid situations where a store has insufficient merchandise in stock to meet customer demand. At the same time, they must also ensure that store managers don’t purchase more of a given item than necessary. Excess inventory has to be discarded in some cases, which can erode a retailer’s profit margins. Besides helping retailers address inventory challenges, AWS Supply Chain can also predict future product availability. It includes a forecasting tool that anticipates customer demand based on historical sales logs and other data. Using the application’s forecasts, a retailer can evaluate if it has enough merchandise in stock to meet future demand and order more when necessary. AWS now also has solutions to ensure the best possible route is taken. This model continually learns as delivery drivers follow their routes and can reoptimize in the middle of a route using real-time data. This will especially be useful for last-mile delivery, which is the most challenging part of getting a package from its origin to its destination. This step is the most expensive aspect of SCM, so getting it right is critical.

Similar solutions in the market have comparable capabilities around supply chain and stock management. Prominent ones include Slimstock, Microsoft’s Dynamics 365 Supply Chain, Oracle’s Supply Chain and Inventory Management, and Manhattan Associates’ Supply Chain Solutions and Inventory Optimization platform. Google also launched its Supply Chain Twin in September 2021, which can be viewed as a comparable solution as well. The platform lets companies build a digital twin—a representation of their physical supply chain—by organizing data to get a more complete view of suppliers, inventories, and events like the weather or dynamic traffic. All these emerging ML-backed solutions relate to supply chain digital twins, as they address a lot of pain points for complex supply chains. Digital twins provide connectivity, metadata management, data management, advanced analytics, and integration with business applications and process systems. This enables participants to develop a feedback loop to optimize their supply chain operations. Cognitive analytics can discover patterns and detect variations. This can optimize inventory, reduce or increase capacity plans, monitor supply chain risks, and test alternates to supply, transportation modes, and locations.

Although there are similar platforms in the market with identical capabilities, AWS stands out with its pre-trained ML models that are based on Amazon.com’s nearly 30 years of supply chain experience. ML requires a large amount of information to function correctly. By feeding it just a few data points, it may make incorrect decisions because it doesn't have the experience to deal with a new situation. When an ML model is fed the multitude of data generated by a store the size of Amazon, it becomes easier to trust the suggestions the AI model makes.

ML allows the computer to learn new information and change its model without being programmed by a human. Having this aspect is vital, as a human cannot process all of the information that millions of Amazon transactions generate. AWS can use information like time of year, location, and last year's demand to suggest lawn care supplies during the fall, while also ensuring its warehouses are stocked with those products before the rush hits. This level of supply chain optimization cuts down on expensive transportation and inventory management costs.

While the solution seems very appropriate for retail and last-mile use cases, it is limited to a narrow range of functionality and is positioned as a demand planning and visibility complement to existing supply chain systems. Currently, even though the solution has demand planning, insights, and a data-lake foundation, it lacks functionality for supply, manufacturing, distribution planning, and execution capabilities. However, it is possible that AWS Supply Chain is only a start, and in due time, it will definitely need to leverage more of its ML capabilities to deliver scenario planning and proactive decision-making capabilities that can enable widespread adoption as a standalone system.