Getting Started
Set Up Your Case Context
Provide the AI with essential context at the start of a conversation. The AI maintains this throughout the chat. Suggested information to include:
Case type and your role
Primary objectives
Key parties and timeline
Main legal issues
Current stage and immediate priorities
Example: "Personal injury case representing plaintiff injured in June 2023 slip-and-fall at grocery store. In the discovery phase, the trial is set for March 2024. Need help analyzing medical records and liability evidence."
Select Relevant Documents
Use the Source selection button or drag-and-drop files into the chat to specify which documents the AI should reference.
Benefits:
Reduces credit consumption
Improves response relevance
Delivers faster results
Best practice: Start with key documents (depositions, critical records, expert reports) rather than all case files.
Document Intelligence Processing
When uploading files via chat, you'll see a prompt to process with Document Intelligence. This enables semantic search and precise, clickable citations.
When to process:
Critical case documents
Files requiring detailed analysis
When you need citations linking to specific passages
When to skip:
Supplementary materials
Reference documents
When managing the Document Intelligence page allocations strategically
Note: Document Intelligence has a separate usage tier (page-based), not AI credits. Processing every document can quickly consume your monthly allocation—prioritize files critical to your case strategy.
Learn more: Document Intelligence Overview
Switch Models Anytime
Change AI models during conversations based on your needs. The current model is displayed in the query box.
Balanced (default) - Most questions
Fast and Low-Cost - Simple lookups, yes/no questions
Premium Intelligence - Complex multi-document analysis, nuanced reasoning
Crafting Effective Prompts
Be Specific and Focused
Clear, detailed prompts get better results.
❌ Avoid | ✅ Better |
"Summarize the case" | "Summarize plaintiff's negligence claims in the March 2024 complaint." |
"What does the doctor say?" | "What did Dr. Martinez testify about the standard of care on pages 45-60?" |
"Find important emails" | "Find emails between Johnson and Smith discussing contract breach, June 1-15, 2023" |
"Tell me about damages" | "List all economic damages in plaintiff's itemized statement." |
"Tell me about the medical records" | "What medical treatments did Dr. Smith perform on the plaintiff between March 15-20, 2023?" |
"What happened in the deposition?" | "Summarize Jones' deposition testimony about vehicle maintenance records." |
"Find information about the case" | "Find merger agreement mentions in documents tagged 'board meeting minutes'" |
Domain-Specific Examples
Personal Injury: "What injuries did the plaintiff report immediately after the accident, per ambulance report?"
Employment Law: "What performance reviews did the plaintiff receive in 12 months before termination?"
Contract Disputes: "What are payment terms and deadlines in the Master Services Agreement?"
Product Liability: "Find consumer complaints about similar defects between 2020-2023"
Use Legal Keywords
Frame questions with terms and ideas likely referenced in your documents.
✅ "What is the date of injury stated in the incident report?"
❌ "When did someone get hurt?"
Narrow with Tags and Dates
"In documents tagged 'medical records,' find mentions of plaintiff's knee injury."
"Review correspondence between January-March 2024 for settlement discussions."
"Search Smith deposition for testimony about accident reconstruction."
Break Down Complex Questions
Instead of: "Review all documents and tell me about injury history, defendant actions, damages, liability issues, and expert opinions."
Try this sequence:
"What injuries did the plaintiff sustain per ER records?"
"What medical treatments has the plaintiff received since the incident?"
"What did Dr. Thompson's expert report conclude about permanent impairment?"
"What economic damages are itemized in the plaintiff's statement?"
"What defenses did defendant raise in answer to the complaint?"
Understanding AI Assistant Tools
AI Assistant uses specialized tools automatically. You'll see status messages showing which tools are active (e.g., "Searching all sources for keywords..."). While the AI selects tools automatically, you can request specific tools using trigger phrases.
Search Tools
Case Search: Searches case records: facts, issues, tasks, contacts. Does NOT search document content.
Best for: Finding timeline facts, locating issues, reviewing tasks/deadlines, searching entities
Triggers: "Search for facts about...", "Find issues related to...", "Show me tasks...", "List all tasks"
Keyword Search: Exact keyword matching in all documents (works without Document Intelligence). Supports wildcards, phrase matching, and boolean operators.
Best for: Exact terms, unindexed documents, wildcard searches (e.g., "medic*")
Triggers: "Search documents for...", "Which documents mention...", "Find the word/phrase..."
Semantic Document Search: AI-powered conceptual search. Requires Document Intelligence processing. Finds meaning, not just exact matches.
Best for: Natural language queries, conceptual searches, understanding context
Triggers: "Find information about...", "What do documents say about...", "Find discussions of..."
Note: For comprehensive coverage, AI may use both semantic (indexed docs) and keyword search (unindexed docs) together.
Document Tools
Read Document Content: Reads full text or specific page ranges. Check page count with Source Metadata first. Read large docs in 10-20 page chunks.
Best for: Complete transcripts, full contracts, detailed sections
Triggers: "Read source [name]", "Show me content of...", "Read pages X to Y..."
View Images: View photos, diagrams, charts, or document pages. Works with image files (PNG, JPG) and PDF pages.
Best for: Accident photos, charts/diagrams, forms, signatures, handwritten docs
Triggers: "Look at the image...", "View page X...", "Display the photo..."
Source Metadata: Gets document info: page count, file type, processing status. Use BEFORE reading large documents.
Best for: Checking length, verifying processing status, planning document workflow
Triggers: "How many pages is...", "Is [document] processed...", "Check document [name]"
Case Management Tools
Propose Issues: Creates legal issues and claims to organize your case. Use Case Search first to avoid duplicates.
Best for: Building a legal framework, organizing by theory, and structuring discovery
Triggers: "Create issues for...", "Set up issues for...", "Generate issues for..."
Create Tasks: Adds reminders and to-dos with due dates. Use Case Search first to avoid duplicates.
Best for: Deadline reminders, discovery schedules, action items
Triggers: "Create a task to...", "Remind me to...", "Add deadline for..."
Update Case Information: Updates case summary with goals, strategy, context—the "executive summary" of your case.
Best for: Setting strategy, documenting theories, recording status
Triggers: "Update case information...", "Set case objectives...", "Document case strategy..."
Request Case Statistics: AI provides statistics ranging from instant (existing data) to detailed (multiple tool calls).
Quick stats (instant): "How many documents?", "Show case statistics", "How many facts?"
Document metrics (requires analysis): "Total page count?", "Which documents longest?", "Processing status?"
Timeline/deadline metrics (requires search): "Timeline date range?", "Overdue tasks?", "Deadlines this week?"
Issue/entity metrics (requires search): "Facts per issue?", "Which issue has the most facts?", "Most mentioned entities?"
Note: The AI Assistant will notify you if analysis requires multiple tool calls and may ask for confirmation.
Quick Reference: When to Use Which Tool
What You Need | Tool | Example |
Find timeline facts | Case Search | "Which facts reference the accident?" |
Find exact keywords | Keyword Document Search | "Which sources mention 'settlement agreement'?" |
Find concepts / discussions | Semantic Document Search | "Find discussions about liability" |
Read the full document | Read Document Content | "Read the Smith deposition." |
View photos/charts | View Images | "Look at page 5 of the medical report." |
Check document size | Source Metadata | "How long is the contract?" |
Build a legal framework | Propose Issues | "Set up issues for negligence" |
Add reminders | Create Tasks | "Remind me to file a response in 10 days." |
Update case overview | Update Case Information | "Update our case strategy" |
Optimizing Results
Validate AI Responses
Click on citations to open the Document Viewer with highlighted text and fact creation prompt. We recommend reviewing for:
Accuracy - Does the cited passage support the AI's statement?
Context - Does the surrounding text change the meaning?
Reliability - Is the source credible and properly cited?
Important: Always verify source material supports conclusions. Examine surrounding text to ensure citations aren't taken out of context. This verification is essential for responsible AI use in legal work.
Use Follow-Up Questions
"Provide more details about the maintenance records you mentioned."
"What did other witnesses say about that incident?"
"Are there contradicting documents?"
Request Specific Formats
"Create a bulleted list of all dates in the complaint."
"Make a table listing all witnesses and which documents mention each."
"Summarize each witness's version in separate paragraphs."
"Compare the witness statement with the police report for conflicts."
"List damages by category (medical, lost wages, pain/suffering)"
Manage Conversation Context
Monitor the "Context used" percentage in the chat box. When approaching 100%, you'll see: "The conversation has been compacted to free context. For more accurate results, consider starting a new conversation."
Compaction allows conversation to continue, but may reduce accuracy for earlier topics.
Start new chat when:
Context is high, and AI loses track of earlier details
Shifting to a different topic/document set
Conversation achieved its objective
Fresh conversations maintain optimal quality and start credits at zero.
When AI Can't Find Information
Troubleshoot:
Check document selection
Verify that the document is uploaded
Check Document Intelligence status (unprocessed = keyword search only)
Rephrase with different keywords
Break into smaller queries
Add more context
Example:
Original: "What did the witness say about the car?"
Better: "What did Jane Doe testify about vehicle speed in her deposition?"
When Responses Seem Incomplete
Try:
Validate citations
Provide more context
Switch to Premium Intelligence
Rephrase with different terminology
Ask clarifying follow-ups
Start a new chat if the context is high
Pro Tips
Leverage Case Context
After providing context initially, reference it naturally:
"What does defendant's investigation reveal?" (AI knows which party)
"Find evidence supporting our negligence claim" (AI knows your theory)
"What damages can we prove?" (AI focuses on plaintiff damages)
Use for Pattern Recognition
"Compare plaintiff's deposition testimony with complaint statements—any inconsistencies?"
"What common themes across all customer complaints?"
"Do witnesses corroborate defendant's timeline?"
Combine Tools Strategically
Document analysis workflow: Source Metadata (check size) → Search (find sections) → Read/View (specific pages)
Avoid duplicates: Case Search (check existing) → Review results → Create only if needed
Comprehensive search: Semantic Search (indexed docs) + Keyword Search (unindexed) + Case Search (records)
Balance AI with Your Expertise
AI excels at:
Locating information across documents
Summarizing testimony/reports
Identifying fact patterns
Creating timelines
Generating statistics
You excel at:
Legal strategy
Credibility evaluation
Relevance judgments
Legal analysis
Understanding client objectives
Provide Feedback
Use thumbs up/down buttons to help improve AI Assistant. Feedback stays internal to Casefleet—not used for external AI training.
Remember: AI Assistant provides evidence-based answers from your case documents, not general legal advice. Quality depends on document completeness, clear prompting, and citation validation.