Traditional Enterprise Knowledge Graphs rely on upfront Entity Resolution (ERRP) that strips away nuance. These heavy build-time commitments create brittle pipelines and discard the contextual signals necessary for handling unstructured data like SEC filings or risk disclosures.

Methodology JIT-Resolution

In this talk, Dr. Weidong Yang and I propose a Just-In-Time (JIT) Resolution, treating the graph as a retrieval engine rather than a static ""truth"" store. By deferring resolution to query time, the graph serves as a high-fidelity index over original sources, preserving provenance and temporal signals. This transforms ERRP from a rigid preprocessing step into an iterative, auditable process where confidence scores and alternative hypotheses are first-class outputs.

Technical Implementation

We demonstrate a SQL-native graph stack where ""SQL is ETL."" Extraction and transformation execute directly in modern data warehouses (BigQuery/Databricks), while a lightweight layer materializes a relationship index for traversal. This architecture allows for continuous ERRP and schema flexibility without pipeline rebuilds.

Case Study

Using SEC 10-K filings, we constructed a temporal competitor graph. The system resolves shifting references (e.g., ""Amazon"" vs. ""AWS"") and tracks evolving competition over time. This grounded AI workflow reduces hallucinations and enables non-graph teams to operationalize relationship intelligence.