From catalogs to context: Why Enterprise Architecture must pivot to knowledge bases and context graphs
For most of the past decade or so, the answer to the question “How can you find out more about the context of your data?” was to build or refer to the data catalog. That paradigm is now changing. In an era where we want to use generative AI to express possible target states in the architecture, the data, business process, and all other aspects of enterprise architecture must be more rigorously defined. This is where the context graph comes in. The context graph gives any logical component in the architecture the ability to express semantic context and relationships between concepts. This becomes important foundational knowledge for a generative AI solution to understand how the components are intended to fit together and therefore how to construct the output based on the available data.
Couple the context graph with a knowledge graph that stores the tacit knowledge of the organization arranged in a relational model conducive to focusing on the relationships between the data elements rather than just the data itself, and it creates a powerful combination of inputs for generative AI.
Compounding the errors
If we suppose that a dataset has 92% factual accuracy, then a ten-step AI process would compound the possible accuracy rate of the output down to 43%, which may be unacceptable to many organizations. With the introduction of the context graph (usually realized as a navigable ontology or graph) this error rate significantly decreases. Enterprise architecture platforms are in a unique position to take advantage of this technology because they already house IT services, applications, infrastructure, business processes, data domains and more. Combining these components and asking a generative AI to decide possible target states without context would certainly give you a quite different answer every time. With the context graphs, the repeated consistency of the output increases.
Ardoq’s move to secure this functionality
Ardoq recently announced that they have acquired Graphlake, a technology that can realize these concepts in the core Enterprise Architecture solution offering. We believe this new combination will propel Ardoq into an era where consistent generative architecture to augment the existing architecture team can be a viable future. In time, other enterprise architecture platforms will follow suit, and generative architecture will become table stakes.
