Why AI Initiatives Fail in Supply Chain

Introduction

As we approach the end of 2024 and a new year begins, it’s essential to examine the landscape of AI initiatives in the supply chain. In the recent discussion with Joannis, it became apparent that many AI endeavors in this space are unlikely to succeed. Joannis laid out a candid perspective that challenges common practices still prevalent in many businesses today. Here’s an in-depth look at the key factors driving failure in AI initiatives for supply chains.

The Current Landscape of Supply Chain Practices

For over four decades, the mainstream approach to supply chain management has stagnated. Companies continue to rely on outdated practices that have not significantly changed since the late 1970s. This includes processes like Request for Proposals (RFPs), time series forecasting, safety stock formulas, and service level metrics. Despite decades of technological advancement, organizations often cling to these established methods, assuming that merely overlaying AI will magically fix the underlying issues.

RFPs: A Dysfunctional Framework

RFPs have become a catch-all for sourcing vendors but are inherently flawed. They require exhaustive specifications and assume that organizations fully understand their own needs, a dangerous presumption. This complexity often leads to missed opportunities and unsuitable vendor selections, effectively creating a bureaucratic nightmare rather than a streamlined procurement process.

Time Series Forecasting: Simplistic Assumptions

Time series forecasting is hailed as the cornerstone of demand prediction, yet it too has significant shortcomings. While aggregated data appears reliable, it fails to capture critical nuances like customer behavior variability or external market influences. Relying solely on this model can lead to misguided decisions, particularly in volatile environments.

Safety Stock: A Non-Economic Perspective

Safety stock calculations are intended to mitigate risk in inventory management. However, they often employ simplistic and deterministic models that ignore the actual economics of inventory decisions. Rather than optimizing profits, safety stock practices can exacerbate costs and inefficiencies. This disconnectedness becomes even more apparent when considering multiple SKUs in a store’s inventory, making it illogical to treat them in isolation.

Service Levels: A Flawed Measure of Quality

Service levels offer a poor proxy for evaluating customer satisfaction. High service levels may suggest reliability; however, they fail to account for product phases and customer needs. Consequently, decisions based solely on service levels can lead to poor inventory choices and ultimately a diminished customer experience.

AI as a Silver Bullet: A Misguided Belief

The prevailing belief that applying AI to flawed existing practices will yield positive outcomes is fundamentally misguided. AI tools cannot overcome the limitations inherent in outdated paradigms. Even the most advanced algorithms will fail to provide meaningful insights if predicated on flawed theories. Consequently, organizations tend to revert to manual overrides to make sense of their supply chains, thus avoiding the consequences of their prior decisions.

Joannis emphasizes a need for organizations to rethink their foundational assumptions about supply chain management. There’s a call to acknowledge that the existing paradigms are broken and require re-evaluation before new technologies are introduced.

The Way Forward

While companies may embrace AI technologies as an additive solution, there is merit in also considering the removal of outdated practices. This shift opens the door for more rational decision-making processes that align with the realities of modern supply chains.

In sum, the failure of AI initiatives in the supply chain boils down to the persistence of outdated frameworks and the belief that technology alone can solve systemic issues without fundamentally altering how decisions are made.

Keyword

• AI initiatives
• Supply chain
• RFPs
• Time series forecasting
• Safety stock
• Service levels
• Decision-making
• Economic perspective

FAQ

Q: What are some common practices that hinder AI initiatives in supply chains?
A: Outdated practices such as RFPs, time series forecasting, safety stock formulas, and service level metrics are often hindrances.

Q: Why are RFPs considered dysfunctional?
A: RFPs assume organizations fully understand their needs, leading to complex and often ineffective vendor selections.

Q: What is the flaw in time series forecasting?
A: Time series forecasting can oversimplify customer behavior and external influences, resulting in misguided decisions.

Q: How do safety stock calculations fail?
A: Safety stock calculations often do not optimize profits and can exacerbate issues when treating SKUs in isolation.

Q: Why are service levels not a good measure of quality?
A: Service levels do not capture actual customer satisfaction and can lead to poor inventory choices if treated as definitive metrics.

Q: What is the proposed way forward for organizations engaging in AI initiatives?
A: Organizations should both reconsider their foundational assumptions in supply chain management and embrace the removal of outdated practices alongside introducing new technologies.