Practical Applications of Generative AI in the Supply Chain

One of the core functions of generative AI is to create content. Through probability and extrapolation, it uses large language models (LLMs) to understand and generate text, images, sounds, and even video. These models work quickly and at scale. As a result, generative AI can help create content for product descriptions and listings, marketing and media, website design, or e-commerce stores.

According to an article in CIO Dive, Procter & Gamble developed an internal generative AI tool that aids their marketing team in the creation of photorealistic images and other assets. The enhancement to the content creation process frees up their employees to focus more on strategic planning and the customer experience.  These images can then be used as part of the company’s online product listings.

For e-commerce, Retail Dive reports that both Amazon and eBay use generative AI to help their sellers generate product descriptions. The AI take in a piece of information, such as a product’s image or key words provided by the seller, and use that data as a seed from which to create a longer, more detailed product description. From there, the seller can review, edit, and update the content, making the product listing process more efficient.

Content generated by AI needs to be reviewed for its authenticity. GS1 US Data Hub® can be used to search and verify product data. In theory, this capability could be expanded to aid human reviewers so that they can work to confirm the accuracy of the information AI generated into a product description or listing. In addition, the Global Product Classification (GPC) can be used to help determine a product’s category for compliance. This can help ensure products have the appropriate certifications and disclaimers incorporated into their listings. A ground source of truth can mitigate the risk of AI hallucinations and make it easier to identify and correct content errors when they occur.

Generative AI can use a communication interface to share valuable information across the whole organization, so it doesn’t require specialized knowledge or skills to understand. As a result, decision-makers at all levels can access key information about inventory, demand forecasting, and logistics. This should help them to respond more quickly and make well-informed supply chain decisions.

Another article from CIO Dive noted that The Home Depot has developed an AI application called Sidekick. It gives store associates access to AI insights to help them manage inventory. It uses computer vision systems to see real-time stock levels in the store and provide informed suggestions.

Walmart rolled out an AI application that can help manage inventory by analyzing structured and unstructured data from a variety of sources, according to a report from the Wall Street Journal. For example, it can examine weather patterns and social media trends to better inform inventory decisions. These insights about large-scale unstructured data enhance traditional predictive AI forecasting.

Insights are built on data. GS1 US offers a Data Hub View/Use subscription that allows users to query and view large amounts of product data that may be used for AI insights. GS1 Standards and data can help provide the building blocks for comprehensive AI operations spanning an industry’s supply chain.

Digital twins are virtual recreations of a real-world environment. They can be used for simulations, studies, and predictive analytics. Generative AI enhances digital twins by producing synthetic data to increase the scope beyond current limitations. Generative and predictive AI work together in a digital twin to simulate and respond to supply chain disruptions or sudden changes in demand.

A recent article from AutomationWorld states that Mars collaborated with Accenture to create digital twins of their products and factory spaces. They wanted to see how adjustments to either would unfold before committing to making changes in the physical space. The digital twin simulation helped them identify bottlenecks and quality issues. They also learned how they could respond proactively to unexpected disruptions or shifts in demand.

UPS is in the process of creating a digital twin of its logistics network, as highlighted in Supply Chain Dive. Their goal is to track packages more accurately and update routes based on real-time updates. They are looking to coordinate many signals into a single cohesive vision of the route. To do so, the digital twin would be built with a combination of Internet of Things (IoT) sensors and AI. This would also aid in optimization and route planning.

The creation of an expansive digital twin ecosystem is not a singular effort. Companies want to gain maximum benefit from large-scale AI simulation. To do so, they would need to interoperably share data with their trading partners. Electronic data interchange (EDI)¹ standards form the core of that interoperable exchange of data. In addition, the digitization of key events within a supply chain can be achieved through electronic product code information services (EPCIS) events to share visibility data and build a truly transparent digital twin.