VI. AI Automation in the Supply Chain

Watch the videos below, and then think about your responses to the discussion questions below. Make notes for our discussion in class.

Introduction

The COVID-19 pandemic exposed a fundamental vulnerability in the global economy: the modern supply chain is a complex, far-flung, and fragile web of global production. For decades, companies shifted manufacturing to Asia to lower costs, creating intricate systems where a single disruption can cause massive backlogs and shortages. In response, some companies are considering expensive, long-term strategic shifts like nearshoring (moving production to a closer country) or reshoring (bringing it back to the home country). However, overhauling these systems can take years and cost trillions of dollars globally, making such moves impossible for many businesses.

This is where Artificial Intelligence introduces a new set of solutions—not by changing the geography of the supply chain, but by radically improving its intelligence and efficiency. The transformation is happening at every level, from the warehouse floor to the corporate boardroom.

At the micro-level, AI is optimizing the physical operations inside distribution centers. Using computer vision, AI can identify arriving vehicles, count and scan boxes, and detect damaged packages before they reach a customer. Intelligent automation can adjust conveyor belt speeds to maximize employee safety and productivity, preventing jams that can cost thousands of dollars per minute in downtime. Using simulation platforms like NVIDIA Omniverse, companies can design and test entire virtual warehouses, training fleets of autonomous mobile robots (AMRs) to sort, store, and retrieve products with maximum efficiency before deploying them in the real world.

At the macro-level, generative AI is empowering strategic decision-making. NVIDIA, which operates one of the world's most complex supply chains, developed an "AI Planner"—an AI agent that allows its operations team to "talk" to their supply chain data. Built on NVIDIA Inference Microservices (NIM) and using Retrieval-Augmented Generation (RAG) to access proprietary data, this agent can run thousands of possible scenarios in seconds. Using a simple natural language prompt, a manager can ask how to reallocate products globally based on hypothetical disruptions—a task that would normally take an entire night to analyze. This leap from physical automation to strategic, conversational AI gives businesses a powerful new toolkit to navigate the uncertainty of the global supply chain with unprecedented speed and agility.

Discussion Questions

Chapter 2 introduces the Value Discipline Model. Which of the three disciplines (Operational Excellence, Product Leadership, or Customer Intimacy) is the primary driver for the AI solutions presented in the videos? Use a specific example from the NVIDIA videos to justify your choice.
In Chapter 3, we distinguished between structured and unstructured tasks. Identify one structured task that AI is automating on the warehouse floor (from the "Efficient Supply Chain Operation" video) and one highly unstructured task that the "AI Planner" is tackling (from the "Talk to Your Supply Chain Data" video).
The WSJ video focuses on mitigating strategic risks like shipping delays and factory closures. The NVIDIA videos present AI as a solution. However, as discussed in Chapter 5, new tools introduce new risks. What is the catastrophic "tiger" a company must consider if the "AI Planner" agent hallucinates or bases its global reallocation plan on flawed or biased data?
NVIDIA's "AI Planner" is a powerful tool built on a platform of proprietary technologies like NIM and cuOpt. Using the "Build, Buy, or Borrow" framework from Chapter 7, how would you classify their approach? How does this AI Planner highlight the critical importance of the Data and Model components of the DIMASI Cube?
Chapter 12 defines an AI agent as a system that is given a goal, not just a command. How does the "AI Planner" exemplify this definition? What is its high-level goal, and what are some of the tasks it autonomously performs to achieve it?