Organizations gather evidence through conversations, documents, decisions and relationships, yet actions rely on interpretations—judgments assembled from multiple observations, grounded on assumptions and revised as contexts evolve. Existing retrieval systems, knowledge graphs and agent memories retain raw records but lose the interpretive context. We introduce a four‑layer architecture: Signals capture raw data; Domains cluster signals semantically; Claims encode interpretations built from several signals together with their justifications; Dimensions place claims on named, versioned axes. Interpretations persist alongside their evidence, and a truth‑maintenance network automatically updates claim credibility as new evidence arrives. All dimensions become searchable within a shared organizational cortex, providing a unified cognitive layer across the enterprise. In a registered retrieval experiment, adding the Dimension term improves nDCG@10 by 0.028 (p=0.002). The full configuration outperforms a Signals‑only baseline by 0.081 and a graph‑augmented Signals‑only setup by 0.053, both statistically significant.\n\nReview: The proposed architecture leverages hierarchical semantic abstraction and truth maintenance to retain interpretive knowledge, offering measurable benefits for cross‑functional collaboration and long‑term knowledge evolution.