LUX-ZEPLIN (LZ) is the world’s largest dark matter direct detection experiment using a dual-phase xenon time projection chamber with a 7-ton active volume, situated a mile underground at the Sanford Underground Research Facility, South Dakota, USA.

Large experimental physics collaborations, such as LZ, produce and maintain a broad range of technical knowledge, including detector documentation, operation records, analysis notes, software repositories, and data-processing workflows. While this information is essential for scientific productivity, it is often distributed across many platforms and evolves continuously over the lifetime of an experiment. This creates a practical challenge for collaboration members. Finding context-aware answers often requires knowing where to search, which terminology to use, and which information is current.

We present LZ Oracle, an AI-assisted platform being developed for the LZ dark matter experiment to improve access to collaboration knowledge and support scientific and operational workflows. The system combines retrieval-augmented generation with an agentic exploration that can search indexed documentation, inspect relevant context, and synthesize responses grounded in collaboration materials. Current development focuses on improving reliability, expanding knowledge coverage, and designing evaluation workflows that allow domain experts to benchmark answer quality and provide structured feedback.

This talk will describe the design goals, architecture, deployment considerations, and lessons learned from developing an AI-assisted knowledge retrieval platform for the LZ collaboration.