What a Context Graph Actually Is, and How to Build One

Context graphs represent a novel approach to understanding and modeling how work is conducted within an organization, differentiating from traditional knowledge graphs by incorporating temporal sequences and behaviors rather than just static relationships. While knowledge graphs focus on the state of entities and their relationships, context graphs delve into the flow and process, answering questions about how tasks and resolutions occur over time. They operate on three layers: the foundational knowledge graph, a personal graph tracking individual actions, and the context graph itself, which aggregates and anonymizes patterns from personal graphs. Unlike process mining, which deals with structured workflows, context graphs handle fragmented data across various tools without a single event log, aiming to create an adaptive model for agents to understand organizational behavior. This model helps agents decide the best course of action in real-time scenarios by offering a probabilistic map of typical workflows, thus filling the gap between static documentation and real-time decision-making. The construction of context graphs relies on deep connectors, a semantic layer for consistent definitions, trace stitching, and hybrid storage solutions that balance structural and semantic access, all while maintaining privacy and adaptability as the organization evolves.