air Assist

aiR Assist is a generative AI-powered conversational search feature embedded within RelativityOne, Relativity's cloud-native eDiscovery and legal data intelligence platform. The tool enables legal professionals to interact with large document sets using natural language queries and receive synthesized, evidence-grounded responses generated by a large language model.

Key features and functions include:

Natural Language Querying
Users submit questions in plain English against an indexed document set within a RelativityOne workspace. The system identifies the most relevant content and generates a structured answer with citations to source documents, supporting defensible review workflows.

Retrieval-Augmented Generation (RAG) Architecture
aiR Assist operates on a RAG framework that combines document retrieval with LLM-based generation. When a query is submitted, the system performs a similarity search across indexed document passages, selects the most contextually relevant results, and passes them to the LLM along with the original question to produce a coherent, citation-supported response of up to 25 citations.

Document Indexing
Users can create up to five indexes per RelativityOne workspace, each accommodating up to 300,000 documents. Only documents with extracted text stored in Relativity's Data Grid are eligible for indexing. Individual documents must not exceed 5 MB of extracted text.

Conversation Management
aiR Assist supports multiple named conversations within a workspace. Users can create, rename, and delete conversations, and activity across all conversations — including questions asked, answers generated, and index changes — is captured in the Relativity Audit application for governance and review.

Common Use Cases
The tool is designed to support three principal litigation workflows: early case insight (surfacing potentially relevant documents and identifying key actors and themes), case strategy development (building timelines and mapping communications), and deposition and trial preparation (identifying potential exhibits and confirming whether specific events or communications are documented).

Language Support
aiR Assist currently supports English-language content only. The system has been designed and tested on English-language datasets; use with non-English content is not formally supported and may produce variable results.