Automation: Amazon’s Supply Chain Optimization Technologies (SCOT)
summary
A large-scale automated system called SCOT runs buy/place/replenish/fulfill decisions for tens of millions of SKUs worldwide, using probabilistic forecasts (from Sparse Quantile Random Forest to convolutional and attention neural nets), capacity-aware placement that must decide in under a second, and a decentralized optimizer called MEPv2 multi-echelon planner that computes opportunity costs for per-ASIN policies. Experiments include four live RCT “labs” before rollout, assortment optimization using NLP plus choice models, and seller tools like the Inventory Performance Index.
AI-written summary · about
chapters (11)
- 0:00Introduction and journey
- 2:16Scott and supply chain
- 6:37Decisions and flywheel
- 10:50Phased automated journey
- 16:10Technical building blocks
- 25:07Scaling automation
- 30:00Phase three challenges
- 35:35New technologies
- 43:55FBA seller tools
- 47:55Future initiatives
- 50:22Principles for innovation
related talks
27mOptimizing Omnichannel Order Fulfillment with AI and Advanced Analytics
26mRiddles in Supply Chain Management - How AI solves them?
26mInnovate with AWS: Enabling industrial ops at Amazon fulfillment centers
16mAutomate supply chains with agentic AI systems
30mCustomer centricity in manufacturing with operational improvements
16mUsing AI to solve complex business problems: optimizing...
26mInnovate with AWS: Enabling industrial ops at Amazon fulfillment centers
55mAutomate Inventory Rebalancing with Agentic AI