AI turns Enron’s emails into an auditable, time-aware company knowledge base
The prototype makes thousands of cited claims navigable by what the company knew at a given date.
A prototype processed 45,047 public Enron emails into 214 entity pages and 49,401 facts, each linked to its source and expired when later messages showed it was no longer current. A held-out audit by a second AI found 93% of sampled facts fully supported and none of 500 samples wrong at their core, while a person selected which proposed entity types to retain. The project shows how specialist-information products can be assembled from unstructured archives with source-level provenance, shifting value from document search toward maintained synthesis.