CaseBot

Overview

CaseBot is an interactive AI-powered chatbot that turns your document collection into a continuous, conversational source of case intelligence. CaseBot can be run as a standalone job on a document population, or as an add-on to an existing Case Insight project. Instead of relying on static outputs or one-time searches, CaseBot allows attorneys and investigators to ask unlimited natural-language questions about their case data and receive immediate, document-backed answers.

CaseBot uses an agentic, multistep process to generate responses. For each question, CaseBot creates a plan, runs multiple searches, reviews the original source documents, verifies the information it finds, and then prepares its response. CaseBot supports its factual claims with citations and verbatim excerpts from source documents. Selecting a citation opens the full document and shows the cited passage in context.

Prerequisites

System Requirements

1. Extracted Text - All text-based documents (emails, Word docs, PDFs, etc.) must have extracted text populated in the Extracted Text field.
2. File Size Limitations - All text-based documents should have an Extracted Text Size of less than 2 MB.

Running CaseBot

CaseBot can be run as a standalone job on a document population, or as an add-on to a Case Insight project.

Data Scope

• CaseBot operates on the complete data set submitted for the CaseBot job, whether run standalone or as an add-on to Case Insight
• There is currently no option to further narrow the scope of data available to CaseBot though metadata limits
• All documents included in the CaseBot job will be indexed and searchable through CaseBot

Core Capabilities

Conversational Q&A

• Ask unlimited natural-language questions about your case data
• CaseBot plans, searches, reviews source documents and verifies its findings before responding
• Follow-up questions do not maintain context from previous conversation
• Questions can focus on events, people, dates, concepts or specific documents

Document Citation

• Every answer is tied directly to source documents
• Citations include verbatim excerpts from the cited passage on a best-effort basis
• Relevant portions of cited documents are automatically highlighted
• Direct links allow instant navigation to full document content in Relativity

Specialist Skills

CaseBot includes eight specialist skills organized into two categories. Each skill applies a structured, repeatable process to a common case task and can be started directly, without having to describe the task from scratch.

Fact Development

• Fact Narrative: Organizes what the documents show about a topic by issue. Identifies what the record establishes and where information is unavailable.
• Chronology: Creates a dated, attributed and cited sequence of events for a matter, topic or person. This allows users to prepare a more narrow or specific timeline than the general chronology that is a part of Case Insight
• Entity Map: Identifies the people and organizations connected to a topic, their roles and their relative importance.
• Hot Document Finder: Ranks documents as Hot, Warm or Notable based on factors such as admissions, anomalies, awareness of wrongdoing and changes in tone.

Case Preparation

• Deposition Kit: Creates a witness preparation package covering the witness's background, prior statements, relevant documents, potential contradictions and possible lines of examination.
• Claim Analysis: Evaluates a single allegation or claim from both perspectives and provides a balanced assessment without reaching a conclusion.
• Request Analysis: Breaks a request for production or subpoena into numbered requests and sub-elements. Identifies potentially responsive documents for each element and provides citations and a coverage summary.
• Record Check: Breaks witness testimony into individual assertions and classifies each as Supported, Contradicted, Mixed or Unsupported. Provides the evidence supporting each classification.

Skills are started from the cards displayed in a new conversation or from the Skills menu next to the message box. Users may also describe the requested task in plain language instead of selecting a skill.

File Attachments

Files can be added to a conversation by dragging them into the window or selecting the paperclip icon. CaseBot can analyze attached files together with the matter's document population.

Supported file types include PDF, Microsoft Word, Microsoft Excel, Microsoft PowerPoint, Outlook .msg, text, Markdown, CSV, JPEG, PNG, WEBP and GIF. Microsoft Office files, PDFs, Outlook messages, text, Markdown and CSV files are processed as text. Supported image files are processed directly.

Attachment Limits

• Up to five files per conversation
• Maximum file size of 10 MB per file
• Very large attachments may not be processed in full
• Attachments apply only to the request submitted with them and are cleared afterward

Using CaseBot

Step 1: Access CaseBot

1. Navigate to the Hub and locate your CaseBot job, whether it was run as a standalone job or as an add-on to a Case Insight project
2. Locate and click on the CaseBot tab or interface
3. CaseBot is enabled by default and is available immediately after the job completes - no request or activation is required
4. The interface opens with a conversational chat window

Step 2: Start a Chat Session

1. CaseBot chat sessions are active for 3 hours from initiation. Additional 3-hour increments can be added by clicking "Extend Purge Time"
2. Start a specialist skill from the cards in a new conversation or the Skills menu, or type a question in natural language in the chat input field
3. Click Ask to submit your question
4. CaseBot plans its approach, runs searches, reviews source documents and verifies its findings before generating a response with supporting citations
5. Responses stream as they are generated. The interface displays the searches CaseBot performs and the documents it opens during processing

Step 3: Review Answers and Citations

1. CaseBot provides an answer to your question in conversational format
2. Supporting documents are cited with each answer, along with a verbatim excerpt of the cited passage where available
3. Click on any cited document to view:

• The full document in Relativity
• The full document text in the Hub

4. Review the highlighted sections to verify and understand CaseBot's reasoning

Step 4: Ask Follow-Up Questions

1. Continue the conversation by asking follow-up questions
2. Refine your queries based on initial answers
3. Explore different angles or dive deeper into specific topics
4. Build your case understanding progressively through interactive dialogue

Step 5: Manage Chat Sessions

1. Session Duration - Each chat session remains active for 3 hours
2. Extending Sessions - You can add additional 3-hour increments to continue an active session
3. Session Privacy - All chat sessions are specific to individual users; you will not see questions and answers from other users
4. Session Retention - Chat sessions are not retained indefinitely by eDiscovery AI. Plan accordingly by documenting important findings and downloading chats or documents prepared by CaseBot
5. New Conversation - Start a new session at any time to begin fresh conversations
6. Persistence - Conversations, citations and generated files remain available if the page is refreshed or you move to another browser tab within the 3 hour window

Step 6: Document Your Findings

Since chat sessions are not retained indefinitely, it is important to document your findings:

1. Copy relevant answers and citations into case notes or strategy documents
2. Save important document citations for future reference
3. Download skill results as formatted reports, or export the full conversation transcript
4. Flag documents identified through CaseBot for further review in Relativity
5. Share findings with team members before the session expires

Best Practices

Effective Question Formulation

1. Be Specific - Ask focused questions about particular people, dates, or topics for more precise answers
2. Use Natural Language - Write questions as you would ask a colleague; no need for Boolean operators or complex syntax
3. Start Broad, Then Narrow - Begin with general questions, then refine based on initial answers
4. Include Context - Reference specific timeframes, departments, or projects when relevant
5. Ask One Question at a Time - Break complex inquiries into individual questions for clearer answers
6. Use Specialist Skills for Structured Tasks - When your task matches a specialist skill, start from the skill rather than a freeform question for a more consistent, complete result

Verification and Quality Control

1. Always Review Citations - Check the source documents that support each answer
2. Verify Critical Information - Confirm important facts through manual review of cited documents
3. Cross-Reference Findings - Compare CaseBot answers with other case intelligence sources
4. Document Key Exchanges - Save important questions and answers for team reference and work product
5. Download Chat - The full chat along with document citations can be downloaded at any time

Understanding CaseBot's Index

Index Creation

• When you run CaseBot, whether as a standalone job or as an add-on to a Case Insight project, CaseBot automatically creates a searchable index of all submitted documents
• This index is retained for the life of the CaseBot job
• The index enables rapid question-answering without re-processing documents for each query

Index Scope

• The index includes all documents submitted to the CaseBot job
• You cannot further narrow or expand the scope after index creation
• To add new documents to CaseBot's knowledge base, you must include them in a new or updated CaseBot job. You can work with the eDiscovery AI support team to ensure appropriate billing for updated CaseBot populations. 

Results and Output

Answer Format

• Answers are provided in clear, conversational language
• Responses include a direct answer followed by supporting citations
• Complex answers may include multiple document references

Downloads and Exports

• Specialist skill results can be downloaded as formatted reports in PDF format
• Full conversation transcripts can also be exported
• Transcript export is currently limited for conversations exceeding 200,000 characters

Response Time

• Most answers to basic questions are generated within seconds
• Complex questions involving multiple documents, or that invoke a specialist skill, may take several minutes

Important Considerations

Session Limitations

Chat sessions are only retained for 3 hours. This is a critical limitation to understand:

• Plan your investigation sessions accordingly
• Document important findings before sessions expire
• Important questions and answers should be saved to external notes or case files
• You can extend active sessions by adding 3-hour increments, but sessions do not persist indefinitely

Accuracy and Verification

• CaseBot provides highly accurate answers based on your document set, but all findings should be verified through attorney review
• The AI serves as an investigative tool to accelerate understanding, not a replacement for professional legal judgment
• Always review cited documents to confirm accuracy and context
• Critical case decisions should be supported by attorney verification of CaseBot findings

Data Completeness

• CaseBot's knowledge is limited to documents included in the CaseBot job
• Ensure all relevant data is included in the CaseBot job, whether run standalone or as an add-on to Case Insight, before relying on CaseBot for comprehensive answers
• Documents added after the CaseBot job completes will not be available to CaseBot without reprocessing
• Consider document completeness when evaluating answer comprehensiveness

Understanding CaseBot's Limitations

Proving Negatives

CaseBot cannot prove that something did not occur or does not exist. If CaseBot cannot find documents on a particular subject, this only means that responsive documents were not identified in the indexed data set. The absence of documents does not prove that an event did not happen, a communication did not occur, or a fact does not exist. Consider the following scenarios:

• Documents may exist but were not included in the CaseBot job's data set
• Relevant communications may have occurred through channels not captured in discovery (phone calls, in-person meetings, non-preserved documents)
• Search terms or question phrasing may not match how the topic was discussed in documents
• Documents may be present but not sufficiently relevant to be cited in response

When CaseBot indicates it cannot find information on a topic, treat this as an investigative finding that requires further analysis, not as proof of absence.

High-Volume Quantitative Questions

CaseBot is optimized for qualitative analysis and targeted document retrieval, not comprehensive quantitative counting. Questions that require analyzing more than approximately 50 documents to provide a complete answer may not retrieve effective results. Examples of questions that may exceed CaseBot's optimal scope include:

• "How many times did John Smith and Jane Doe email each other over the past year?"
• "Count all documents mentioning Product X across all departments"
• "How many employees attended training sessions in 2024?"
• "List every instance where the pricing policy was discussed"

For these types of high-volume quantitative queries, consider using our Relevance solution instead which will return all documents relevant to any particular issue. CaseBot works best for questions like:

• "What were the key topics discussed between John Smith and Jane Doe regarding the merger?"
• "What are the most important documents about Product X's safety concerns?"
• "Summarize the company's approach to employee training based on relevant communications"
• "What pricing changes were discussed in executive meetings?"