Skip to main content

​
Overview

Twenty integrates powerful AI capabilities to help you automate tasks, generate content, and enrich your data. These features use advanced language models to understand context and provide intelligent assistance throughout your workflow.

​
Available AI Features

AI Email Generation

Generate professional email drafts with AI assistance

Data Enrichment

Automatically enrich contact and company records with AI

AI Agents

Create intelligent agents to automate complex workflows

AI Chat

Ask questions about your data and get instant answers

​
How AI Works in Twenty

Twenty’s AI features are powered by large language models (LLMs) that can:
  • Understand context - Analyze your data and provide relevant suggestions
  • Generate content - Create emails, summaries, and descriptions
  • Extract information - Pull data from unstructured text
  • Make predictions - Suggest next best actions
  • Automate decisions - Execute workflows based on intelligent rules

​
AI Capabilities

​
1. Email Generation

AI can help you write emails faster:
  • Generate professional email drafts
  • Suggest responses based on context
  • Adjust tone and style
  • Create follow-up sequences
Learn more: AI Email Generation

​
2. Data Enrichment

Automatically enhance your records with AI:
  • Find company information from minimal data
  • Enrich contacts with job titles and social profiles
  • Discover company size, industry, and location
  • Update records with latest information
Learn more: AI Data Enrichment

​
3. AI Agents

Build intelligent automation with AI agents:
  • Create custom agents for specific tasks
  • Give agents access to your CRM data
  • Configure agent behavior and permissions
  • Execute agents in workflows
AI agents can:
  • Research prospects
  • Qualify leads
  • Update records based on criteria
  • Generate reports and summaries

​
4. AI Chat

Ask questions about your CRM data:
  • “How many deals closed this month?”
  • “Show me contacts at companies in San Francisco”
  • “What’s my pipeline value?”
  • “Find all high-priority opportunities”
The AI chat understands natural language and can query your data without writing code.

​
Enabling AI Features

​
For Cloud Users

Twenty Cloud includes AI features by default. Simply:
  1. Navigate to Settings > AI
  2. Enable the features you want to use
  3. Start using AI in your workflows
AI features on Twenty Cloud are subject to usage limits based on your plan. Check your plan details for specific quotas.

​
For Self-Hosted Users

To enable AI features on self-hosted instances:
1

Choose an AI Provider

Twenty supports multiple AI providers:
  • OpenAI (GPT-4, GPT-3.5)
  • Anthropic (Claude)
  • Open-source models (via APIs)
2

Configure Environment Variables

Add AI configuration to your .env file. The exact variables depend on your chosen provider.Example for OpenAI:
# AI Configuration
AI_ENABLED=true
AI_PROVIDER=openai
OPENAI_API_KEY=your_api_key_here
3

Set Usage Limits (Optional)

Configure cost controls:
AI_BILLING_ENABLED=true
AI_MONTHLY_BUDGET=100
4

Restart Twenty Server

Apply the configuration:
docker-compose restart
5

Verify Setup

Check AI features are available in Settings > AI.

​
AI Models

Twenty uses different AI models for different tasks:

​
Smart Models

For complex reasoning and generation:
  • GPT-4 - Advanced understanding and generation
  • Claude 3 Opus - Long context and detailed analysis
Use for:
  • Email generation
  • Data enrichment
  • Complex agent tasks

​
Fast Models

For quick, simple tasks:
  • GPT-3.5 Turbo - Fast and cost-effective
  • Claude 3 Haiku - Quick responses
Use for:
  • Simple classifications
  • Quick suggestions
  • Chat responses
Twenty automatically selects the appropriate model for each task to balance quality and cost.

​
AI Billing and Usage

​
How AI Usage is Measured

AI features consume tokens:
  • Input tokens - Your prompts and context
  • Output tokens - AI-generated content
Costs vary by model:
  • Smart models: Higher cost, better quality
  • Fast models: Lower cost, good for simple tasks

​
Monitoring Usage

Track AI consumption:
  1. Navigate to Settings > AI > Usage
  2. View usage by feature, model, and time period
  3. Set budget alerts

​
Cost Control

Manage AI costs:
  • Set budgets - Configure monthly spending limits
  • Usage alerts - Get notified when approaching limits
  • Rate limiting - Control requests per user/workspace
  • Model selection - Choose cost-effective models
AI features can incur significant costs at scale. Monitor usage regularly and set appropriate limits for your organization.

​
Privacy and Security

Twenty takes AI privacy seriously:
  • Data encryption - All data sent to AI providers is encrypted
  • No training - Your data is not used to train AI models
  • Opt-in - AI features are opt-in only
  • Audit logs - Track all AI operations
  • Access controls - Restrict AI features by role

​
Self-Hosted Privacy

For maximum privacy:
  1. Use self-hosted AI models
  2. Configure on-premise LLM providers
  3. Keep all data within your infrastructure

​
Implementation Details

AI features are implemented in Twenty’s backend:
  • AI modules: packages/twenty-server/src/engine/metadata-modules/ai/
  • Agent execution: packages/twenty-server/src/engine/metadata-modules/ai/ai-agent-execution/services/agent-async-executor.service.ts:9
  • Workflow integration: packages/twenty-server/src/modules/workflow/workflow-executor/workflow-actions/ai-agent/ai-agent.workflow-action.ts:29
  • Billing: packages/twenty-server/src/engine/metadata-modules/ai/ai-billing/services/ai-billing.service.ts

​
Best Practices

  1. Start small - Enable one feature at a time
  2. Monitor costs - Track AI usage regularly
  3. Set limits - Configure appropriate budgets
  4. Train your team - Educate users on AI capabilities
  5. Review outputs - Always verify AI-generated content
  6. Iterate prompts - Refine agent instructions for better results

​
Limitations

Be aware of AI limitations:
  • Not always accurate - AI can make mistakes
  • Requires context - More data = better results
  • Cost considerations - Can be expensive at scale
  • Response time - Some operations take seconds
  • Language support - Best for English content

​
Future AI Features

Upcoming AI capabilities:
  • Voice transcription for calls
  • Sentiment analysis
  • Automated lead scoring
  • Predictive analytics
  • Custom model fine-tuning

Share Feedback

Have ideas for AI features? Share them on GitHub Discussions.

​
Getting Help

For AI feature support:
Experiment with AI features in a test workspace before rolling out to your team.

Build docs developers (and LLMs) love

Get started for freeTalk to us