Knowledge Graph

Beyond a List of Memories

A flat list of memory entries — even a well-classified and well-scored one — cannot represent the relational structure of institutional knowledge. The fact that a vendor decision was influenced by a prior incident, which was in turn shaped by a process failure, which led to a workaround that is still active today, requires a graph to express. Individual entries in isolation tell part of the story. The graph tells the whole.

RolegacyAI's Knowledge Graph models the connections between role memories, the entities they reference, and the other memories they relate to — providing a structured, navigable representation of the role's accumulated institutional knowledge.

Node Types

The graph contains multiple node types:

• Memory entries: Individual captured decisions, lessons, workarounds, and processes.
• Roles: The roles within the organisation — the primary unit of institutional knowledge.
• People: Role holders, key contacts, stakeholders (referenced in a professional capacity, not personally).
• Vendors and partners: External organisations with which the role has a documented relationship.
• Systems and tools: Technology platforms, integrations, and configurations referenced in role memories.
• Processes: Named operational processes that appear across multiple memory entries.
• Projects and initiatives: Significant programmes of work that generated role knowledge.

Edge Types and Relationships

Nodes are connected by typed edges that express the nature of the relationship:

• influenced — this decision was influenced by that earlier decision or event
• supersedes — this newer approach replaces an earlier documented approach
• relates-to — these two entries share relevant context
• caused — this event or decision led to a subsequent outcome
• applies-to — this workaround applies to this system or process
• involves — this entry involves this vendor, system, or stakeholder

Graph Traversal for Briefing and Retrieval

When generating a successor brief or answering a role-specific query, the Knowledge Graph enables traversal beyond the directly retrieved entries. If a retrieved decision is connected to a related workaround that is still active, the graph ensures the workaround is included in the output even if it was not directly retrieved by the semantic query. This traversal gives the RAG Engine access to contextually essential information that flat vector search alone might miss.

Cross-Role Knowledge

The Knowledge Graph also enables authorised cross-role analysis — identifying patterns of decisions, common workarounds, or shared vendor relationships that span multiple roles in the organisation. This analysis supports organisational learning: understanding which operational patterns appear repeatedly across roles can surface systemic issues that no individual role's memory reveals in isolation.