The following is an excerpt from the upcoming Designed Analytics report, “Generative AI in Supply Chain: Key Considerations“. The report will go-live on http://www.Designed-Analytics.com on 02/25/2024.

If I ask you to think beyond text and images when thinking about Generative AI, it may seem contrary to the very notion of Generative AI. But that is when you can expand how you can think about the possibilities of leveraging this amazing technology in your organization, and in this specific context, in transforming supply chain management.

To envision Generative AI this way, you must keep one central aspect in mind- when you interact with an LLM leveraging text, it needs to translate that text in the backend to understand what you are trying to communicate, translate that into what it needs to do, go ahead and execute that, and then translate its own gibberish into a language you understand. The model has been trained to mimic that it can understand English (hence the word “Artificial” in AI) or any other language. 

Therefore, you do not always need to interact with a LLM using text. In fact, you do not always need to initiate a dialogue with the model. The model may trigger a dialogue based on certain triggers. And understanding this aspect is key to understanding the plethora of ways you can leverage Generative AI across your supply chain and think about co-pilots in a whole different way.

As shown in Figure 3, when you interact with the model, the model constructs a query based on your question. One key highlight of the figure is that the database that is being tapped is not always a vector database. This aspect of the architecture opens the possibility of marrying your structured, graph DB, and vector DB to build unique capabilities. This is no futuristic architecture and is already being leveraged for many applications.

Figure 3: From query translation to generation

It is therefore critical that when you think about Generative AI, shake-off the notion of text, images and videos being generated by tools that are available for public to play with. While you do not need to be a techie, understanding the capability behind the poems and images will help you explore the possibilities in a more practical way. Let us explore an example from the procurement area.