AI Agents

What is an AI Agent?

An AI Agent is an intelligent assistant trained on your content that can automatically answer customer questions, provide support, and handle FAQs. Agents use advanced language models to understand context and provide accurate, helpful responses based on your knowledge base.

AI Agents are perfect for:

• Customer Support: Handle common questions and support requests
• FAQ Automation: Answer frequently asked questions instantly
• Information Retrieval: Provide accurate information from your documents and website
• 24/7 Availability: Offer round-the-clock customer assistance

Creating an AI Agent

Follow these steps to build your first AI Agent:

Step 1: Navigate to Agents

1. Click on the Agents tab in the sidebar menu
2. You'll see a list of all your existing agents (if any)

Step 2: Create New Agent

1. Click the Create Agent button
2. You'll be taken to the agent configuration page

Step 3: Configure Your Agent

Fill in the basic configuration:

• Name: Give your agent a descriptive name (e.g., "Customer Support Agent", "Product FAQ Bot")
• Description: Add an optional description to help identify the agent's purpose
• System Instructions (Prompt): Define how your agent should behave, its role, tone, and guidelines
  • Example: "You are a helpful customer support agent. Be friendly, concise, and always try to solve customer problems."
• AI Model Provider: Choose from available providers:
  • OpenAI (GPT-4, GPT-3.5)
  • Claude (Anthropic)
  • Gemini (Google)
  • And more...

Step 4: Add Knowledge Sources

Train your agent by adding knowledge sources. Your agent will use this content to answer questions accurately.

Supported Knowledge Sources:

• Website URLs: Add your website URL to train the agent on your website content
  • The agent will crawl and index your website pages
• PDF Documents: Upload PDF files containing information your agent should know
  • Product manuals, documentation, guides, etc.

To add knowledge sources:

1. Click Add Knowledge Source
2. Choose the type (URL, PDF, or Text)
3. For URLs: Enter the website URL
4. For PDFs: Upload your PDF file(s)

You can add multiple knowledge sources to create a comprehensive knowledge base.

Step 5: Configure System Tools (Optional)

Enhance your agent's capabilities by adding system tools. Tools allow your agent to perform actions beyond just answering questions.

Available Tools:

• Google Calendar: Schedule events, check availability, create appointments
• Google Sheets: Read and write data to spreadsheets
• Google Docs: Create and update documents
• Gmail: Send emails, read messages
• Custom Tools: Add your own integrations via webhooks

To add tools:

1. Scroll to the System Tools section
2. Enable the tools you want your agent to use
3. Configure tool-specific settings
4. Add guidelines for when and how to use each tool

Step 6: Test Your Agent

Before deploying, test your agent to ensure it responds correctly:

1. Use the Test Agent interface on the agent page
2. Type test questions your customers might ask
3. Verify the agent provides accurate, helpful responses
4. Check that it uses knowledge sources correctly
5. Test tool usage if you've configured any

Testing Tips:

• Ask questions that should be answered by your knowledge sources
• Test edge cases and unclear questions
• Verify the agent's tone matches your instructions
• Check that tool calls work as expected

Step 7: Deploy Your Agent

Once your agent is tested and ready:

1. Deploy to Website Widget: Add the agent to your website chat widget

   • Copy the widget code
   • Embed it on your website
   • The agent will handle conversations on your site

2. Deploy to WhatsApp: Connect your agent to WhatsApp

   • The agent will automatically respond to WhatsApp messages
   • Integrate with your existing WhatsApp workflows

Key Features

• Multi-Source Training: Train agents on websites, PDFs, and custom text
• Multiple AI Models: Choose from various AI providers (OpenAI, Claude, Gemini, etc.)
• Tool Integration: Connect with Google services and custom tools
• Customizable Behavior: Define agent personality and response style through prompts
• Real-Time Testing: Test agents before deployment
• Multi-Language Support: Agents can respond in multiple languages
• Context Awareness: Agents maintain conversation context for better interactions

Best Practices

• Clear Instructions: Write detailed system instructions to guide agent behavior
• Quality Knowledge Sources: Use well-structured, accurate content for training
• Regular Testing: Test with real customer questions before going live
• Monitor Performance: Review agent conversations to identify improvements
• Update Knowledge: Keep knowledge sources up-to-date with latest information
• Set Boundaries: Define what the agent should and shouldn't do
• Use Tools Wisely: Only enable tools your agent actually needs