Intelligence Settings

Configure AI models, analysis features, and intelligence capabilities.

Accessing Intelligence Settings

1. Go to Workspace Settings (or Settings → Intelligence)
2. Select Intelligence tab
3. Configure AI and analysis options

AI Provider Configuration

Selecting an AI Provider

Choose from supported AI providers:

Provider	Models Available	Notes
OpenAI	10 models including GPT-5.2	Default provider
Mistral AI	3 models	European option
Google	4 models	Gemini family
Cohere	2 models	Enterprise-focused

Changing Provider

1. Go to Intelligence → AI Provider
2. Select provider from dropdown
3. Choose specific model
4. Save changes

📝 Note: Model changes affect all AI features workspace-wide.

Default Model

GPT-5.2 by OpenAI is the default:

• Highest accuracy
• Best general performance
• Recommended for most use cases

Model Considerations

Factor	Consideration
Accuracy	Some models perform better on specific tasks
Speed	Smaller models may be faster
Cost	Different models have different pricing
Privacy	Some prefer specific providers

Transcription Settings

Language Configuration

Setting	Description
Auto-detect	Harmony identifies the spoken language
Primary language	Default language for transcription
Secondary languages	Additional languages to support

Supported languages:

• English (primary)
• Spanish
• French
• German
• Portuguese (beta)

Custom Vocabulary

Add industry-specific terms for better transcription:

1. Go to Intelligence → Custom Vocabulary
2. Add terms, names, and jargon
3. Save vocabulary list
4. Harmony learns these terms

Examples:

• Product names
• Technical terminology
• Customer names
• Industry jargon

Speaker Identification

Configure speaker detection:

• Enable/disable speaker labels
• Train on known speakers
• Improve accuracy over time

Analysis Configuration

Default Metrics

Choose which metrics to calculate for all meetings:

• Engagement score
• Talk ratio
• Sentiment analysis
• SPIN distribution
• Keywords detection
• Competitor mentions

Scorecard Templates

Create default scorecards for meeting evaluation:

1. Go to Intelligence → Scorecards
2. Create or edit templates
3. Define criteria and weights
4. Apply to meeting types

Learn more about Scorecards →

Competitor Configuration

Define competitors to track:

1. Go to Intelligence → Competitors
2. Add competitor names
3. Include name variations
4. Enable automatic detection

Keyword Tracking

Set up automatic keyword detection:

1. Go to Intelligence → Keywords
2. Add terms to track
3. Categorize by type
4. Configure alerts

Insight Generation

Automatic Insights

Configure what insights are generated automatically:

Insight Type	Auto-Generate
Meeting summary	✅ Default on
Action items	✅ Default on
Key topics	✅ Default on
Coaching tips	⚙️ Configurable
Risk flags	⚙️ Configurable

Custom Insight Prompts

Create workspace-default custom insights:

1. Define insight prompts
2. Set as default for all meetings
3. Users see these automatically

Insight Accuracy

Improve insight accuracy over time:

• Review and correct insights
• Provide feedback on AI outputs
• Harmony learns from corrections

Processing Settings

Processing Priority

📝 ASSUMPTION: Priority settings may vary by plan.

Priority	Description
Standard	Normal queue processing
High	Faster processing (additional cost)
Real-time	Live transcription (Max/Enterprise)

Notification on Complete

Configure processing notifications:

• Email notification
• In-app notification
• Slack/Teams notification

Quality Settings

Transcription Quality

Setting	Description
Standard	Default accuracy
Enhanced	Additional processing passes
Professional	Highest accuracy (Enterprise)

Audio Enhancement

📝 ASSUMPTION: Audio enhancement options.

• Noise reduction
• Echo cancellation
• Audio normalization

Integration with AI Features

Companion Integration

Intelligence data feeds Companion:

• Transcripts enable query responses
• Insights inform suggestions
• Context enhances accuracy

Agent Integration

Agents access intelligence outputs:

• Process meeting data
• Trigger on insights
• Generate reports

Advanced Settings

Confidence Thresholds

Set minimum confidence for various features:

Feature	Default Threshold
Action items	80%
Sentiment labels	70%
Speaker identification	85%
Custom insights	75%

Processing Rules

Define rules for processing:

• Skip certain meeting types
• Apply different settings by calendar
• Custom processing pipelines

Enterprise Features

Self-Hosted AI

For Enterprise self-hosted deployments:

• Use your own AI infrastructure
• Connect to internal models
• Air-gapped processing

Custom Models

• Fine-tune on your data
• Industry-specific optimization
• Specialized vocabulary

Monitoring

AI Usage Dashboard

Track AI resource consumption:

• Transcription minutes
• AI queries
• Processing time
• Model usage breakdown

Quality Metrics

Monitor transcription and analysis quality:

• Accuracy scores
• User corrections
• Feedback trends