Fileread is an AI-based fact-finding platform for litigation, eDiscovery, corporate legal, and internal investigations matters. Users query full document productions in natural language, and each answer includes a citation to its exact location in the source material.
Key features and functions include:
Source-Linked Answers
Each answer includes a citation to its exact location in the source document, allowing facts to be verified directly rather than located through manual search.
Full-Production Search
Users can search the complete document production rather than a subset of the record, across file formats including text, images, and handwriting.
Work Product Generation
Generates fact memos, case chronologies, contradiction reports, deposition outlines, character lists, and organizational charts from the underlying record.
Workbench
Plans and executes multi-step research tasks, such as building a case chronology, across the full record, and determines subsequent search steps based on prior findings; recurring tasks, including contradiction reports, can be run as an automated workflow.
Axis
Builds a data extraction schema from an uploaded list of target fields or a written description, and extracts data such as parties, dates, and dollar figures into a structured spreadsheet; each exported data point includes a citation to its source.
Multimodal Analysis
Extracts information from visual content, including charts, images, and diagrams, in addition to text-based documents.
Corporate Investigations Use
The platform is also positioned for in-house legal and investigations teams, supporting review of internal communications and records to reconstruct events and brief leadership with source-cited findings.
Relativity Integration
Fileread integrates natively into Relativity workspaces and is listed on Microsoft Azure Marketplace; it is also available as a standalone platform outside of Relativity.
Processing Infrastructure
Document processing occurs in Microsoft Azure using private large language model inference; the vendor states customer data is not used to train a shared model.
Loading...