AI Configuration

Fine-tune how Cothon's AI features work. Choose models, adjust behavior, and optimize for your use case.

AI Models

Available Models

Cothon uses state-of-the-art AI models:

Selecting Default Model

1. Go to Settings > AI > Models
2. Choose your preferred default model
3. Applied to all AI features unless overridden

Feature-Specific Models

Override models for specific features:

Extraction Settings

Extraction Mode

Focus Areas

Customize what the AI looks for:

• Technical Requirements - Specifications, standards
• Compliance - Certifications, regulations
• Timeline - Deadlines, milestones
• Financial - Budget, pricing constraints
• Experience - Past performance, qualifications

Extraction Accuracy

Balance speed vs. accuracy:

• Fast - Quick results, may miss edge cases
• Balanced (Default) - Good accuracy, reasonable speed
• Thorough - Maximum accuracy, slower processing

Analysis Behavior

Strictness Level

How strictly the AI evaluates compliance:

Capability Matching

Configure how capabilities match requirements:

• Exact Match - Must match explicitly
• Semantic Match (Default) - Understands similar meanings
• Broad Match - More flexible interpretation

Confidence Thresholds

Set minimum confidence levels:

Proposal Generation

Writing Style

Choose the tone for generated proposals:

• Professional - Formal, business-appropriate
• Technical - Detail-oriented, precise
• Conversational - Friendly, approachable
• Concise - Brief, to-the-point

Section Preferences

Enable/disable proposal sections:

1. Go to Settings > AI > Proposals
2. Toggle sections on/off
3. Reorder sections as needed

Default sections include:

• Executive Summary
• Technical Approach
• Past Performance
• Management Plan
• Pricing Schedule

Auto-Enhancement

Automatically improve generated content:

• Grammar Check - Fix errors automatically
• Consistency - Maintain terminology
• Compliance Language - Include required phrases

AI Assistant

Response Length

• Concise - Brief, direct answers
• Detailed - Comprehensive explanations
• Adaptive - Matches question complexity

Conversation Memory

• Full Context - Remembers entire conversation
• Recent Only - Last 10 exchanges
• Minimal - Each question independent

Proactive Suggestions

Enable AI to offer suggestions:

• When viewing requirements
• When writing proposals
• When browsing opportunities

Performance Settings

Processing Priority

• Quality - Best results, may be slower
• Balanced - Good results, reasonable time
• Speed - Fastest processing, may sacrifice detail

Concurrent Operations

Limit simultaneous AI operations:

Caching

Enable AI response caching:

• Faster repeated analyses
• Reduced API usage
• Cache expires after 24 hours

Cost Management

Usage Tracking

Monitor AI usage:

1. Go to Settings > AI > Usage
2. View:
   • Operations this month
   • Estimated cost
   • Usage by feature

Budget Limits

Set monthly limits (organization admins):

1. Go to Settings > Organization > AI Budget
2. Set maximum monthly spend
3. Get alerts at 80% and 100%

Custom Instructions

Global Instructions

Add instructions applied to all AI operations:

Example:
- Always use Canadian English spelling
- Reference ISO 9001:2015 standards when relevant
- Avoid acronyms without definitions

Feature-Specific Instructions

Add instructions for specific features:

1. Go to Settings > AI > Custom Instructions
2. Select feature (extraction, proposals, etc.)
3. Add your instructions
4. Save changes

Testing Configuration

Preview Mode

Test AI settings before applying:

1. Click Test Configuration
2. Upload a sample document
3. See results with new settings
4. Apply or adjust

A/B Comparison

Compare different configurations:

1. Enable A/B Mode
2. Set two configurations
3. Run same document through both
4. Compare results side-by-side

Resetting AI Settings

Restore defaults:

1. Go to Settings > AI
2. Click Reset to Defaults
3. Confirm reset
4. All AI settings return to factory defaults

Next Steps

• Notifications - Configure alerts
• Bid Analysis - Use AI for analysis
• Proposals - Generate AI proposals