eDiscovery and DeepDive- Introduction

What are eDiscovery and DeepDive?

eDiscovery and DeepDive are AI-powered features within Cognitive CDE that help users find and understand information from project documents.

eDiscovery searches across all project documents to answer user queries. It scans multiple files and returns relevant information along with source references and relevancy indicators, making it easier to locate information across large volumes of data.
DeepDive allows users to select a specific file and ask detailed questions about its content. It provides responses based only on that file, helping users explore and understand document-level information more clearly.

NOTE: eDiscovery and Deepdive are part of the Cognitive CDE add-on feature and incur commercial costs. Please contact your Asite representative for more details.

Prerequisites

System Requirements

System Requirement	Description
Relevant Project Vectorised	The Project you want to access eDiscovery for must be vectorised by the admin for AI processing.

Privileges

Role Privilege	Description
Can Access Cognitive CDE	Allows users to access and use Cognitive CDE features.

Steps to Access eDiscovery

Go to the Files tab.
Click Cognitive CDE.
In the Top Bar, click the Dropdown Menu icon.
Select the relevant Project.

The Cognitive CDE screen appears, where you interact with the eDiscovery agent for the selected Project.

Understanding eDiscovery

Once a Project has been vectorised, eDiscovery enables you to ask questions and generate summaries across all files within a selected project.
Users can interact with it conversationally and retrieve insights from across multiple documents.
It works with the Cognitive CDE's Translation agent, allowing users to ask questions in their preferred language and receive translated answers from the source files.
It also allows users to track the conversation history and manage how the results are displayed.
The screen is divided into three sections:
- Left section: Shows the project you are working in, your previous conversations with the bot, and an option to start a new conversation.
- Middle section: Displays the selected conversation. You can view and continue the chat here.
- Right section: Shows the list of files the bot used to generate the answer, along with their AI relevancy percentages.
Refer to the image and table below to understand each part of the eDiscovery screen:

Number	Title	Description
1	Project Name	Shows the Project selected for AI analysis.
2	RAG (Retrieval Augmented Generation) Bar	A fully customisable list of attributes describing the current files used to generate AI's answer. The key attribute to note is AI Relevancy (%), which appears next to each file and indicates how much of the AI's answer came from that file. You can adjust order and appearance using the Customise button (bottom-right).
3	Chat History	Lists all current and previous conversations with the bot.
4	New Conversation	Start a new conversation with the bot.
5	View buttons: (how many files are displayed at once, Customise button, and maximise screen button)	A set of buttons to manage how results are displayed: Rows per page: choose how many files are displayed at once. Customise: adjust the order and appearance of metadata fields shown on the right. Maximise screen: expand the screen to full view. Click again to minimise.

How Do I Access DeepDive?

Once you've implemented eDiscovery (refer to Steps to Access eDiscovery), you can access DeepDive directly from the file list on the right-hand side:

Find the File you'd like to explore.
Click Explore with AI.

The file will open in full view, with the chat panel on the right-hand side. From here, you can type any question or request, and Cognitive CDE will generate detailed responses based solely on the contents of that file.

Understanding DeepDive

DeepDive is designed for situations where you need to focus on a single File rather than an entire Project. Instead of searching across hundreds of documents, DeepDive allows you to stay within one file and explore it in depth.

With DeepDive, you can:

Ask highly specific questions about the file's content.
Request clear explanations or summaries of complex sections.
Engage in a conversational back-and-forth, refining your questions as you go.

The system is responsive and context-aware; therefore, it doesn't just provide surface answers; it works through the file with you, understanding your prompts and returning insights that match the detail you're looking for.

eDiscovery and DeepDive

Function	eDiscovery	DeepDive
Scope of query	Processes queries across the entire vectorised project.	Restricts queries to a single file selected by the user.
Summarisation	Generates summaries spanning multiple documents.	Produces focused summaries on an individual file.
Interpretation style	Operates in literal retrieval mode. E.g. a query for "empty pages" matches only that text string.	Operates in a cognitive interpretation mode. E.g. detects whether a document structurally contains an empty page.
Metadata configuration	Customisable attributes in the RAG bar via the Customise button.	Not applicable.
System prompts	Currently ignores system prompts.	Supports advanced system prompts with deeper control over scope, offering more accurate and adaptable output.
Scalability	Suitable for very large projects (hundreds of thousands of files).	Suitable for deep analysis of individual files, including lengthy reports and technical drawings.
Intended use case	Breadth: discovery, analysis at the project level.	Depth: exploration, interrogation of a single file.