35% Fewer Stockouts: What AI Demand Forecasting Changes in Supply Chain Operations

Arizona State University's Brett Duarte on AI supply chain strategy, part 2: demand forecasting, inventory optimization, and safety stock

Get your forecast wrong in one direction and you've got empty shelves. Get it wrong the other way and you've got a warehouse full of product nobody wants. Professor Brett Duarte pointed to Amazon reporting a 35% reduction in stockouts from its AI inventory tools. So which parts of the AI promise hold up in supply chain operations?

Season 2, Episode 14 of Data Blades continues our 3-part mini-series with Professor Brett Duarte of Arizona State University's W. P. Carey School of Business. The series follows ASU's AI-Enabled Supply Chain Strategy executive program running October 5 to 7 on the Tempe campus. Part 1 covered procurement. Part 2 moves downstream into operations, where a forecast turns into inventory and inventory turns into either a sale or a markdown.

Brett walks through the AI demand forecasting toolkit layer by layer: extrapolation from historical data, regression and econometric models, machine learning for nonlinear patterns, and large language models acting as a reasoning layer on top. That top layer handles what he calls narrative synthesis and multi-source fusion, pulling qualitative signals like a Wall Street Journal article on consumer sentiment, the consumer price index, or inflation data into the forecast.

Casey admits capacity planning in ERP often felt like "pretend functionality." Ricardo describes the retail version of real time: however fast someone can update the spreadsheet. Then Brett turns to inventory optimization, where AI moves teams off static reorder points and monthly forecasts. He describes an agent that senses a demand spike, raises a purchase order, checks the supplier contract for budget limits and terms, sends routine orders automatically, and routes important ones to a human. He closes on multi-echelon inventory optimization and why the location of your safety stock matters as much as the amount.

What you'll learn in this episode.

Why AI demand forecasting and inventory optimization are two halves of one decision [00:03:43]How AI demand forecasting stacks extrapolation, regression, machine learning, and LLMs [00:06:14]What narrative synthesis and multi-source fusion mean for a forecast that can read the news [00:07:07]Why forecasting is moving from Python code into the hands of managers and executives [00:08:31]How AI agents replace the days of manual spreadsheet update