Automate replenishment with MMF, Databricks Genie, and Amazon Quick
Foundation models made catalog-wide demand forecasting easy; the hard part is now acting on the forecast. This post builds a closed detect-decide-act loop on Databricks and Amazon Quick that reconciles demand surges against live supplier availability and places replenishment orders unattended, escalating to a human only when no supplier can cover a surge.
The article outlines a process for automating replenishment in retail using AWS Machine Learning tools. The key steps involve generating a demand forecast using AWS's Many Model Forecasting (MMF) system, detecting surges in demand, deciding which supplier can fulfill the demand, and finally, acting on the decision to place an order.
The forecast is created using Databricks' MMF, which predicts 7-day demand for each SKU. An agent monitoring the forecast identifies SKUs whose demand has surged, defined as a 7-day average that is at least 1.5 times the 14-day average and greater than 1. The decision process involves reconciling the surging SKUs against live supplier availability stored in Amazon S3 Tables. The cheapest supplier capable of fulfilling the demand is selected.
If a single supplier can meet the surge, an order is placed through a Supplier Order API. If no single supplier can cover the demand, a human review ticket is raised instead. The entire process is managed by Amazon Quick, which acts as the intermediary between Databricks and the external systems. Amazon Quick uses a Model Context Protocol (MCP) connector to retrieve the forecast and a Direct Query dataset from Amazon S3 Tables to access supplier information. Orders are placed through an OpenAPI connector.
The solution includes detailed setup instructions, including prerequisites and configuration steps, as well as a repository with the necessary scripts and JSON files to automate the entire process.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
