Cencora 的 AI 需求感知方法：比对预测与订单模式，早期结果预测改善 10%

Cencora's Matt Sample explains how AI compares forecasts with ordering patterns to flag pharma demand shifts early, with 10% forecasting gains.

Artificial intelligence (AI) is giving Cencora an earlier read on shifts in pharmaceutical demand by comparing historical forecasts against incoming ordering patterns, according to Matt Sample, senior vice president of manufacturer, quality and replenishment operations at Cencora.

Sample spoke with Pharmaceutical Commerce at LogiPharma USA 2026, joined by Gregg Gorniak, MBA, vice president of manufacturer operations and data services and secure supply chain lead at Cencora. Sample says the company thinks about AI in three verticals. The first is legacy machine learning, which has been around for 20 years. The second is generative AI, which lets users interact with a data set. The third is agentic AI, which can take actions and recommend solutions.

Cencora does not receive demand data directly from its customers, Sample says. Traditional supply chains, particularly in the consumer space, get point-of-sale data that signals in real time whether sales are running above or below forecast. “We don’t really get something like that,” Sample says.

To close that gap, Cencora recently rolled out an AI tool to a handful of buyers. The tool compares the forecast against ordering patterns to flag demand variations that call for correcting the forecast, along with the reasoning behind the correction.

Buyers can also question the system directly, for example by asking what drove a demand spike. The answer might point to a specific customer in a specific area. It might also flag that the spike could be seasonal because the products involved are flu-related. Previously, buyers pieced that picture together themselves from spreadsheets and the company’s SAP ERP system. “Now, all the data is in one place,” Sample says.

So far the tool covers a limited number of SKUs and buyers, and early results show a 10% improvement in forecasting. According to Sample, that means either more product on the shelf or less excess inventory from overbuying.

Cencora is also working to use knowledge graphs, which Sample describes as not true AI but a different way to link data. The goal is to look at manufacturers alongside FDA actions and geopolitical and financial factors to identify risk patterns that may drive a market shortage. When a demand signal shows up alongside a risk warning, “that’s a conversation that needs to happen,” Sample says.

That kind of early warning matters as shortages climb again. ASHP and the University of Utah Drug Information Service counted 227 active national drug shortages in the second quarter of 2026. That was the third consecutive quarterly increase.¹

Reference

1. American Society of Health-System Pharmacists; University of Utah Drug Information Service. National Drug Shortages: January 2001–June 2026. 2026. https://www.ashp.org/-/media/assets/drug-shortages/docs/Drug-Shortages-Report.pdf

Related to this article