AI retrieval assistant cuts energy asset management task time 42% in pilot study

Energy utilities rely on decades-old enterprise asset management (EAM) platforms for engineering work management, procurement, and inventory. Replacing these systems is cost-prohibitive and operationally disruptive. This paper from IEEE's SEGE 2026 introduces a retrieval assistant that adds an AI-powered knowledge access layer on top of legacy EAM. It operates in three modes: answering questions from vendor documentation, querying the operational data store (ODS) schema, and providing UI how-to guidance. The runtime pipeline combines intent understanding with query rewriting, hybrid semantic and vector retrieval, context engineering under token limits, grounded answer generation, and deterministic hyperlink conversion for panel identifiers and citations.

The data preparation pipeline treats semantic enrichment as the primary quality lever: adding table and field descriptions, normalizing acronyms across sources, and indexing representative row-level context. A pilot with a small sample showed significant gains across all metrics. Precision at five improved from 0.56 to 0.72, mean reciprocal rank from 0.43 to 0.58, and nDCG at five from 0.51 to 0.66. Median task completion time dropped from 14.2 to 8.3 minutes, a 42% reduction. User-rated usefulness and confidence both hit 4.0 on a five-point scale. Beyond immediate gains, the approach establishes reusable foundations for future analytics and automation tools, showing that even modest AI interventions can unlock significant operational value in legacy-heavy industries.