Enterprise systems generate a vast amount of valuable data that spans across different departments, as well as data from ecosystem partners, customer interactions, and external sources, which can provide valuable insights. Yet, bringing all this data together can be complex. Once accomplished, however, it can lead to significant benefits such as a single source of truth, advanced analytics, differentiated AI use cases, and even data monetization. Creating a data pipeline and harmonizing disparate sources might not be a core competency that organizations want to invest in. Luckily, there are tools and techniques available to make this journey easier, including breaking it down into multiple iterations based on various parameters to deliver faster business benefits. There are also advanced analytics and low code AI tools to harness the data to help business to detect anomalies, identify patterns and derive predictive insights.
Many times, complexities arise when translating insights into action. For example, if an organization obtains an insight on port congestion and its impact on the lead time of certain supply lines, it could potentially trigger multiple actions such as adjusting the production plan due to the anticipated delay or changing control parameters, such as lead time or ordering calendar configuration, in the supply planning system. However, the challenge involved in this could be due to the lack of maturity or complex nature of the enterprise applications to consume this change in an automatic manner. Another possibility is that the organization may lack domain / system expertise, to understand the combinatorial effects of changing control parameters in value chain processes.
One possible solution to this challenge is the use of adaptive domain-specific LLMs. By uptraining a base model on the nuances of enterprise systems involved in value chain orchestration using the available documentation (e.g. legacy mainframes, bespoke applications, off the shelf ERPs & planning systems such as SAP or Blue Yonder, etc.) and effective prompting, they could become super domain experts aware of everything in an organization’s context and can help ‘LLM in Loop’ decisions as part of the workflow. However, this possibility raises the question of whether this is a scary prospect or a path to 10X benefits.
