Skip to main content

Multi-Agent AI System

Victoria AI is not a single chatbot—it’s a team of specialized AI legal assistants working together through an intelligent orchestrator. Each agent has:
  • Domain-specific expertise trained on relevant legal knowledge
  • Access to CaseMind memory for contextual awareness
  • LexVault integration for document-grounded responses
  • Cost optimization using the right model for each task

The 4 Chat Agents

Victoria provides 4 interactive chat agents for different users and use cases:
Victoria also includes specialized onboarding agents (firm, case, and client setup) and backend AI functions (case insights, memory extraction, deadline tracking) that work automatically behind the scenes. See Backend AI for technical details.

How Agent Routing Works

When you ask Victoria a question, here’s what happens:
1

Request Analysis

The Orchestrator analyzes your question to determine which specialist agent should handle it
2

Context Assembly

The selected agent gathers relevant information from:
  • CaseMind memory (past conversations and extracted facts)
  • LexVault documents (uploaded case files)
  • Case metadata (parties, dates, jurisdiction)
3

Expert Processing

The specialist agent processes your request using:
  • Domain-specific prompts and knowledge
  • Appropriate Claude model (Haiku/Sonnet/Opus)
  • Relevant legal frameworks and templates
4

Response Delivery

You receive a detailed, contextual response with:
  • Direct answer to your question
  • Relevant citations or references
  • Follow-up suggestions
5

Memory Update

New information is extracted and stored in CaseMind for future reference

Model Selection

Victoria automatically chooses the right Claude model for each task:
Task TypeModelSpeedCostBest For
Routing & ClassificationClaude Haiku 4.5~200ms0.80/0.80/4Quick decisions
Legal Strategy & DraftingClaude Sonnet 4.5~2-5s3/3/15Most tasks
Complex Financial CalculationsClaude Opus 4.1~5-10s15/15/75Precision work
Victoria automatically uses the most cost-effective model for each task. Most requests use Sonnet, with Haiku for routing and Opus reserved for complex calculations.

Conversational Context

Victoria maintains conversation context within each chat session:
  • Short-term memory: Recent messages in the current conversation
  • Long-term memory: All extracted facts stored in CaseMind
  • Document grounding: Relevant content from uploaded documents
This means you can have natural, flowing conversations without repeating context.
Each AI agent conversation starts fresh but has access to all CaseMind memory. If you need to reference a previous conversation, mention it explicitly (e.g., “As we discussed earlier about asset division…”)

Cost Tracking

Every Victoria AI request is tracked with:
  • Token usage (input/output)
  • Model used (Haiku/Sonnet/Opus)
  • Cost calculation in real-time
  • Attribution to case, lawyer, and agent
View detailed usage in SettingsAI Usage Dashboard.

Best Practices

  • Co-Counsel: Legal strategy, motions, general questions
  • Financial Analyst: Calculations, asset division, support
  • Discovery Manager: Discovery planning and document requests
  • Client Liaison: Client-facing 24/7 support (in Client Portal)
The more context you provide, the better Victoria’s responses:
  • Reference specific documents: “Based on the financial affidavit…”
  • Mention jurisdiction: “Under California law…”
  • State your goal: “I need a motion to modify support because…”
Before asking Victoria about a case:
  1. Upload key documents to LexVault
  2. Wait for AI processing to complete
  3. Then ask questions that reference those documents
Always review and edit Victoria’s output before filing or sending:
  • Check facts against source documents
  • Verify calculations independently
  • Ensure jurisdiction-specific requirements
  • Add your professional judgment

Agent Capabilities Matrix

CapabilityCo-CounselFinancialDiscoveryClient Liaison
Motion Drafting✅ Expert
Legal Research✅ Expert⚠️ Basic⚠️ Basic
Financial Calculations⚠️ Basic✅ Expert
Asset Division⚠️ Basic✅ Expert
Discovery Strategy⚠️ Basic✅ Expert
Document Requests✅ Expert
Client Q&A⚠️ Basic✅ Expert
Document Search (Client)✅ Expert
General Legal Info⚠️ Basic✅ Expert
Case insights, task extraction, and deadline tracking run automatically in the background via Victoria’s Backend AI functions. See Backend AI for details.

Next Steps