Generative AI Use Cases in Supply Chain

Use Cases & Projects, Dataiku Product, Featured Christine Andrews

From demand forecasting to inventory management, delivery dock optimization to warehouse management, enterprises rely on classic machine learning more and more to transform supply chains. In the last five years, organizations have leveraged the technology to make more proactive decisions. This includes streamlining processes as well as unlocking new opportunities for growth. So what does Generative AI bring to the table on top of these existing capabilities?

→ Read Now: Why You Need an AI Platform to Scale Generative AI

The Two Tiers of Applications

Before we get into Generative AI in supply chain specifically, let's take a step back. Imagine the first generations of artificial intelligence (AI) were like the steam power of the first industrial revolution. That makes these new models the electricity of the second industrial revolution. Generative AI models will be transformative in ways that we do not yet anticipate.

Think about potential applications for Generative AI technologies in supply chain in terms of how they will be transformative:

  • Moonshots, or use cases that introduce fundamentally new capabilities.
  • Mundane, or use cases that augment hundreds or thousands of processes and decisions throughout the business.

You Need Mundane and Moonshot Use Cases to Succeed

The most commonly imagined moonshot application of Generative AI in the enterprise is the all-knowing, oracular chatbot. This always-on, always-accurate assistant can provide immediate answers or predictions about the current and future state of the business.

Could this work? Maybe, but there are many caveats and many reasons to doubt that such a system would ever be fully trusted. Receiving less attention is the potential for Generative AI to transform the business through the augmentation of countless mundane tasks. In this scenario, the business chooses different models for different applications, balancing considerations like performance, cost, and privacy.

Teams have the power to apply approved technologies to the challenges that they face. Democratized Generative AI leads to a massive increase in productivity.

Here's the thing — you need to be able to execute on both moonshot and mundane Generative AI use cases. Yes, you should start building the rocket for that moonshot. But even more immediately, you must start augmenting processes throughout your organization, finding the right balance between autonomy and control. 

LLM-Enhanced Demand Forecasting

Let’s walk through one example of a Generative AI use case in supply chain management that would have immediate value in augmenting your workforce. That is Generative AI-enhanced — or more precisely, large language model (LLM)-enhanced — demand forecasting

 

Imagine you already have machine-learning powered demand forecasting (perhaps even using Dataiku’s demand forecasting solution). A supply chain analyst interacts with data through a dedicated interface, allowing them to select specific SKUs, markets, stores, and timelines. 

What Generative AI can bring to the table is the ability to turn analytics into action. For example, say there’s a planned surge in demand for a SKU that’s in limited stock. This could not only be identified, but acted upon. Generative AI could provide the analyst with an accelerated capacity to follow up on this insight with an email to engage for a procurement activity. 

For this use case, Generative AI can bring:

  • Improved response time in purchase and markdown decisions with insights from real-time analytics.
  • Increased revenue thanks to continuous response to product demand.
  • Empowerment via improved communication and collaboration between teams, who are now able to turn analytics into action.

Supplier Risk Management

Supply chains have never been more global or diversified. That provides opportunities but introduces great risk. Everything from geopolitical changes to environmental disruptions can upend tier one, two, or three suppliers and put entire supply chains at risk.  

Shortages of components as significant as semiconductors or seemingly inconsequential as fasteners have caused production stoppages, delays, and even canceled orders. The law of unintended consequences means that well intentioned just-in-time inventory management strategies can save short-term costs that can be readily offset when raw materials become scarce. Navigating this insecurity and surfacing risk is a critical need of supply chain leaders. 

Generative AI affords the ability to leverage massive amounts of unstructured text data across myriad sources, from publicly available information to internal documents and contracts.

Organizations seeking to minimize and manage supply chain risks rely on individuals to maintain supply scorecards and supplier risk analysis reports. The challenge here is two-fold: 

  1. Overreliance on manual reporting with opaque oversight from other stakeholders in the organization and 
  2. Limited data inputs that provide an incomplete picture of supply chain risks through the different layers and tiers.   

Using LLMs and Retrieval Augmented Generation (RAG), enterprises can automate risk identification in the context of supplier order patterns, BOM (bill of materials) breakdowns, and demand plans. From there, they can automate the computation of, for example, reliance on sole sourced components with corresponding supplier risk and identification of alternate suppliers. 

Generative AI can go even further by helping rate and score those suppliers that meet compliance considerations and regulatory requirements. Traditionally, some of this data is surfaced in BI reports that are manually created and stitched together based on extracts from ERP systems and, too infrequently, enhanced with institutional knowledge from supply chain professionals.

Materials Management

It’s one thing to identify your risks and exposure, but what can organizations do about it? Sometimes alternative suppliers aren’t a viable option. In many cases, that means alternate materials or processes are the only avenue for addressing risk and avoiding complete production stoppages. Often, this design exploration can be quite time consuming and, thus, is undertaken only after a problem emerges and time is of the essence. This is one way Generative AI can help — creating rapid ways of identifying alternative materials.  

Many organizations are leveraging Generative AI to help in the R&D space for new material identification and property prediction.  In the same vein, the technology can aid supply chain managers identify alternative components and materials with similar characteristics. Design engineers, chemists, or formulation specialists can ask questions like, “How does component or material A differ from B?”

It’s not replacing the need for a domain specialist but, using Generative AI, your scientists and specialists can more quickly identify alternate material compatibility and define the right experiments or simulations to run.  From surfacing risk to mitigating it, Generative AI has the potential to make highly diversified supply chains more resilient and less exposed.  

wide angle shot of a warehouse showing potential use cases for generative ai in supply chain

Achieving Generative AI at Scale

Maybe you'll succeed in building your moonshot. A single chatbot that is accurate, safe, and cost-effective, serving the needs of the entire organization. However, it probably should not be your organization’s sole Generative AI strategy. It is too risky, and it misses the myriad improvements that are much closer at hand. 

How, then, can an organization succeed in scaling the use of Generative AI throughout their organization? There are three essential pillars to achieving this scale:

  1. Democratization. There is a massive gap between the potential applications and the number of experts who could build those applications. The solution is enabling more people to use and help build Generative AI applications.
  2. Acceleration. Competitive pressure means that time is of the essence. Applications or assets that are developed need to then be reused and repurposed. In other words, you cannot start from scratch for every new application of the technology.
  3. Trust. People must trust Generative AI, and it has to be low-risk.

You May Also Like

From Vision to Value: Visual GenAI in Dataiku

Read More

Data Preparation Dataiku Hidden Gems: Part 2

Read More

Maximizing Enterprise Data Products Distribution

Read More