Skip to main content

AI Assistant

Intelligent document analysis, search, and case management powered by secure AI.

What is the AI Assistant?​

The AI Assistant uses advanced artificial intelligence to help you work more efficiently with your documents, cases, and communications:

  • Understand documents instantly with AI-powered analysis
  • Find information faster with intelligent search
  • Manage cases efficiently with automated insights
  • Triage emails automatically based on priority and sentiment
  • Generate professional reports in seconds

Built for Security and Privacy​

Your data never trains AI models: ✅ Zero-knowledge processing: AI analyzes content without storing it ✅ No training data: Your documents never improve third-party models ✅ Encrypted analysis: All AI processing happens on encrypted data ✅ Audit trail: Every AI operation is logged for compliance ✅ Rate limiting: Prevents excessive AI queries for security

AI-Powered Features​

Document Intelligence​

Automatic Classification (U2)

  • Categorizes documents (contracts, invoices, legal, personal, etc.)
  • Extracts key information (names, dates, amounts)
  • Summarizes content in plain language
  • Identifies entities and relationships

Deep Analysis (U1)

  • Analyzes document context and purpose
  • Extracts structured data (tables, forms, metadata)
  • Identifies patterns and anomalies
  • Generates searchable summaries

Learn more →

AI-Enhanced Search (U4)

  • Interprets natural language queries
  • Suggests query refinements
  • Understands context and intent
  • Filters by document type, date, relevance

Example:

Query: "contracts signed in 2024"
AI Interpretation: Looking for legal contracts signed during 2024
Suggested Refinements:
  - Contracts signed in Q1 2024
  - Contracts with specific parties in 2024
  - Unsigned contracts from 2024

Learn more →

Case Management​

Case Insights (U5)

  • Identifies key issues and themes
  • Detects patterns across documents
  • Provides actionable recommendations
  • Flags potential risks

Case Triage (WO1)

  • Automatically prioritizes new cases
  • Assigns urgency levels (1-5)
  • Categorizes by type
  • Routes to appropriate team members

Timeline Generation (U3)

  • Creates chronological case timelines
  • Identifies key events and milestones
  • Links evidence to timeline entries
  • Visualizes case progression

Learn more →

Email Management​

Smart Email Triage (U6)

  • Categorizes emails (urgent, inquiry, spam, feedback, etc.)
  • Detects sentiment (positive, negative, neutral, angry, concerned)
  • Identifies action items and deadlines
  • Prioritizes responses (1=Critical, 5=Low)
  • Suggests case linking

Learn more →

Report Generation​

Automated Reports (U7)

  • Executive summaries with key findings
  • Case summaries with citations
  • Timeline reports with chronology
  • Custom reports from prompts

Report Types:

  • Executive Summary: High-level overview for stakeholders
  • Case Summary: Detailed case analysis with evidence
  • Timeline Report: Chronological event sequence
  • Custom Report: Generate reports from natural language prompts

Learn more →

How AI Processing Works​

Privacy-First Architecture​

Your Data Flow:

1. You upload document to Torvus
   ↓ [Encrypted at rest]
2. You request AI analysis
   ↓ [Encrypted in transit]
3. AI processes document (Google Gemini or Groq)
   ↓ [No data retention by AI provider]
4. Results returned to you
   ↓ [Encrypted, logged for audit]
5. AI provider forgets your data immediately

Security Guarantees​

What we DO:

  • ✅ Encrypt all data before AI processing
  • ✅ Use zero-data-retention AI providers
  • ✅ Log all AI operations for security audit
  • ✅ Rate limit AI usage to prevent abuse
  • ✅ Allow you to disable AI features entirely

What we DON'T do:

  • ❌ Store your data on AI provider servers
  • ❌ Use your documents to train AI models
  • ❌ Share your data with third parties
  • ❌ Process data without your explicit action
  • ❌ Enable AI by default (opt-in only)

Using the AI Assistant​

Enabling AI Features​

Requirements:

  • Pro plan or higher
  • Two-factor authentication enabled
  • Terms of Service acceptance for AI features

Setup:

  1. Go to Settings → Advanced Features
  2. Enable AI Assistant
  3. Review and accept AI Terms
  4. Configure AI Preferences (optional)
  5. Start using AI features

AI Preferences​

Customize AI behavior:

  • Provider: Google Gemini (quality) or Groq (speed)
  • Confidence Threshold: Minimum confidence for AI suggestions (0.7-0.95)
  • Auto-Classification: Automatically classify uploads (on/off)
  • Search Refinements: Show AI-generated query suggestions (on/off)
  • Email Triage: Automatically triage incoming case emails (on/off)

AI Feature Availability​

By Plan Tier​

FeatureFreeProBusinessEnterprise
Document Classification❌✅ 100/mo✅ 1,000/mo✅ Unlimited
Document Analysis❌✅ 50/mo✅ 500/mo✅ Unlimited
Intelligent Search❌✅ Unlimited✅ Unlimited✅ Unlimited
Case Insights❌✅ 50/mo✅ 500/mo✅ Unlimited
Email Triage❌✅ 200/mo✅ 2,000/mo✅ Unlimited
Report Generation❌✅ 20/mo✅ 200/mo✅ Unlimited
Case Triage❌❌✅ 500/mo✅ Unlimited
Timeline Generation❌❌✅ 100/mo✅ Unlimited
Policy Generation❌✅ 10/mo✅ 100/mo✅ Unlimited

Usage Tracking​

Monitor your AI usage:

  • View AI operations in Settings → Usage
  • See monthly limits and current usage
  • Upgrade plan for higher limits
  • Receive notifications at 80% and 100% usage

Pricing​

Pro Plan: $29/month

  • AI features for personal use
  • 100 document classifications/month
  • 50 case insights/month
  • 200 email triages/month
  • 20 reports/month

Business Plan: $99/month (per user)

  • All Pro features + more quota
  • Case triage and timeline generation
  • Priority AI processing
  • Team sharing of AI insights

Enterprise: Custom pricing

  • Unlimited AI operations
  • Custom AI models (fine-tuned for your org)
  • Dedicated AI infrastructure
  • SLA guarantees for AI availability

Best Practices​

When to Use AI​

✅ Good Use Cases:

  • Analyzing large volumes of documents quickly
  • Finding information in unstructured documents
  • Prioritizing emails and cases by urgency
  • Generating first drafts of reports
  • Identifying patterns across cases

❌ Not Recommended:

  • Making final legal decisions (always review AI suggestions)
  • Processing highly sensitive data without review
  • Relying on AI for 100% accuracy (always verify)
  • Sharing AI-generated content without editing

AI Accuracy​

Confidence Scores:

  • 0.90-1.00: High confidence, usually accurate
  • 0.80-0.89: Good confidence, verify important details
  • 0.70-0.79: Medium confidence, review carefully
  • Below 0.70: Low confidence, treat as suggestions only

Always verify AI results for:

  • Legal documents and contracts
  • Financial information
  • Personal identifiable information (PII)
  • Security-sensitive decisions

Rate Limiting​

Why we rate limit:

  • Prevents accidental excessive usage
  • Protects against account compromise
  • Ensures fair usage across all users
  • Reduces costs for everyone

Rate Limits:

  • Per-operation limits: Max 10 AI operations per minute
  • Daily limits: Based on plan tier
  • Monthly quotas: See table above

If you hit a limit:

  1. Wait for rate limit window to reset (1 minute)
  2. Upgrade plan for higher limits
  3. Contact support for temporary increase

Privacy and Compliance​

Data Protection​

GDPR Compliance:

  • ✅ Right to deletion (AI results deleted with documents)
  • ✅ Data minimization (only necessary data sent to AI)
  • ✅ Purpose limitation (AI only used for explicit features)
  • ✅ Transparency (all AI operations logged)

HIPAA Compliance (Healthcare customers):

  • ✅ Business Associate Agreement (BAA) available
  • ✅ Zero-data-retention AI providers
  • ✅ Encrypted PHI in transit and at rest
  • ✅ Audit logs for all AI access

SOC 2 Compliance:

  • ✅ Security controls for AI operations
  • ✅ Availability monitoring
  • ✅ Confidentiality protections
  • ✅ Annual third-party audit

AI Provider Information​

Google Gemini (Primary provider):

  • Model: gemini-2.0-flash-exp
  • Region: US (configurable for EU customers)
  • Data retention: Zero (immediate deletion after response)
  • Compliance: SOC 2, GDPR, HIPAA-eligible

Groq (Speed-optimized provider):

  • Model: llama-3.3-70b-versatile
  • Region: US
  • Data retention: Zero
  • Use case: Email triage, case triage (high volume)

Technical Details​

API Integration​

Developers can access AI features via API:

# Classify document
POST /api/v1/ai/classify
{
  "documentId": "doc_abc123",
  "vaultId": "vault_xyz789"
}

# Response
{
  "category": "contract",
  "confidence": 0.95,
  "summary": "Employment agreement...",
  "keyPoints": ["Salary: $150K", "Start date: Jan 15"],
  "entities": {
    "names": ["TechCorp", "Jane Smith"],
    "dates": ["2024-01-15"],
    "amounts": ["$150,000"]
  }
}

for full documentation.

Monitoring​

AI Operations Dashboard (Enterprise):

  • Real-time AI usage tracking
  • Confidence score distributions
  • Error rates and retry counts
  • Latency percentiles (p50, p95, p99)
  • Cost analytics

Access at Settings → AI Monitoring (Enterprise only)

FAQ​

Is my data used to train AI models?​

No. We use zero-data-retention providers that immediately delete your data after processing. Your documents never become training data for AI models.

Can I use AI features offline?​

No. AI features require internet connection to access AI providers (Google Gemini, Groq). Document storage and viewing work offline.

What happens if AI makes a mistake?​

AI is not perfect. Always review AI-generated results, especially for:

  • Legal documents
  • Financial information
  • Security decisions

Report inaccurate results to help us improve: Settings → Report AI Issue

Can I disable AI features?​

Yes. Go to Settings → Advanced Features → Disable AI Assistant. Your existing documents are unaffected.

How accurate is the AI?​

Accuracy varies by use case:

  • Document classification: 85-95% accuracy
  • Entity extraction: 80-90% accuracy
  • Sentiment analysis: 75-85% accuracy
  • Report generation: Requires human review

Always check confidence scores and verify results.

Does AI work in languages other than English?​

Currently, AI features work best in English. Limited support for:

  • Spanish
  • French
  • German
  • Portuguese

Other languages: AI will attempt processing but accuracy may be lower.

What if I exceed my AI quota?​

Options:

  1. Wait: Quota resets monthly
  2. Upgrade: Higher plan = more quota
  3. Contact Support: Request temporary increase

AI operations fail gracefully—no data loss if quota exceeded.

Can I export AI analysis results?​

Yes. Export options:

  • JSON: Structured data for programmatic use
  • CSV: Import to Excel/Sheets
  • PDF: Human-readable reports
  • API: Access via REST API

Go to Documents → Select document → AI Analysis → Export

Getting Help​

AI Support​

Technical Issues: ai-support@torvussecurity.com Inaccurate Results: Report via Settings → Report AI Issue Feature Requests: feature-requests@torvussecurity.com Enterprise Support: Available 24/7 for Enterprise customers

Resources​

Last Updated: October 15, 2025