HelpScout Integration

The HelpScout integration allows you to deploy your AI chatbot directly into your HelpScout inbox, enabling automatic responses to customer emails. The AI agent appears as a team member and can handle conversations, create drafts for human review, or fully automate responses.

Key Features

• Automatic AI responses to customer emails
• Draft mode for human review before sending
• Customizable response tone (professional, friendly, casual, etc.)
• Configurable response modes (all messages, first only, or assisted)
• Smart message batching with configurable delays
• Conversation merging for customers who send multiple emails
• AI disclaimer to inform customers they're interacting with AI
• Customizable email signatures
• Automatic handoff to human agents when needed
• Conversation tracking in your dashboard
• Mailbox-specific configuration
• Playground for testing AI responses before going live

Quick Setup

How It Works

1. A customer sends an email to your HelpScout inbox
2. You (or a workflow) assign the conversation to the AI Agent user
3. The AI analyzes the conversation and your knowledge base
4. A response is generated and either published or saved as a draft
5. If the AI can't help, it hands off to a human agent

Configuration Options

Response Mode

Control when the AI agent responds to conversations and whether responses are published immediately or saved as drafts:

Response Tone

Choose how your AI communicates with customers:

Reply Delay

Configure how long the AI waits before responding (0-300 seconds). This allows the AI to batch multiple customer messages sent in quick succession into a single response. Default is 30 seconds.

• Set to 0 for immediate responses
• Set to 30-60 seconds to handle customers who send multiple messages

Mailbox Selection

Choose which mailbox the AI monitors:

• Specific Mailbox: Only conversations in the selected inbox trigger AI responses
• All Mailboxes: The AI can respond to conversations across all your inboxes

AI Disclaimer

Add a disclaimer to the first AI-generated response in each conversation to inform customers they're receiving an automated response:

• Enabled by default: A disclaimer is automatically added to the first AI response
• Customizable text: Modify the disclaimer to match your brand voice (max 500 characters)
• Default text: "This response was generated by an AI assistant and may require verification."

The disclaimer appears at the end of the first AI response in each conversation, giving customers transparency while maintaining a natural conversation flow.

Email Signature

Add a consistent signature to the end of every AI response:

• Disabled by default: Enable to add a signature to all responses
• Customizable text: Enter your preferred signature text (max 500 characters)
• Examples: "Best Regards, Support Team" or company-specific signatures

The signature appears at the end of every AI-generated response, helping maintain brand consistency across all communications.

Conversation Merging

When customers send multiple emails in quick succession (e.g., adding information or follow-up thoughts), the AI can consolidate them into a single response:

• Enabled by default: Multiple rapid emails are merged before the AI responds
• Registration Period: How long to wait for additional emails (10-600 seconds, default 150)
• Conversation Oldness Limit: Maximum age of conversations eligible for merging (10-600 seconds, default 150)

When merging occurs:

1. The earliest conversation becomes the "primary" conversation
2. Secondary conversations receive a "Conversation Linked" note explaining the merge
3. Messages from secondary conversations are imported as "Imported Message" notes
4. Secondary conversations are automatically closed
5. The AI responds once to the combined context

AI Conversation Tagging

All conversations handled by the AI are automatically tagged with a special "quassum" tag in HelpScout. This helps you:

• Filter and view all AI-handled conversations
• Track AI performance in HelpScout reporting
• Identify which conversations have received AI assistance

Automatic Handoff

The AI automatically hands off to human agents in these situations:

• Customer explicitly requests to speak with a human
• Conversation exceeds 50 customer messages
• AI determines it cannot adequately answer the question
• Conversation requires actions outside the AI's capabilities

When handoff occurs, the AI:

1. Posts a message informing the customer that a human will follow up
2. Unassigns itself from the conversation (or assigns to a specific user if configured)

Security & Permissions

• OAuth authentication - Secure connection without sharing credentials
• Read-only where possible - The integration only requests necessary permissions
• Encrypted tokens - OAuth tokens are encrypted at rest using AES-256-GCM
• Webhook verification - All incoming webhooks are verified using HelpScout's signature
• Disconnect anytime - Remove the integration from your dashboard instantly

Testing with the Playground

Before going live, use the built-in Playground to test how the AI responds to customer messages:

Accessing the Playground

1. Go to Integrations in your dashboard
2. Click Settings on the HelpScout integration
3. Click Playground (or find it in the integration menu)

Playground Features

• Simulated conversations: Send test messages as if you were a customer
• Real AI pipeline: Uses the same response generation as live HelpScout conversations
• Customizable context: Set the customer email and conversation subject
• Full conversation history: Build multi-turn conversations to test context handling
• Copy transcript: Export the conversation for review or documentation
• Save for review: Automatically saves conversations when cleared for later analysis

Debug Mode

When debug mode is enabled (configured by administrators), the Playground shows additional information:

• Model name: Which AI model generated the response
• Tool calls: Any tools the AI used (e.g., knowledge base lookups)
• System prompt: The full instructions given to the AI
• Input messages: How the conversation was formatted for the AI
• Duration: Response generation time in milliseconds

Troubleshooting

Best Practices

Getting Started

• Start with Assisted Mode to review AI responses before sending
• Use a test conversation before going live
• Train your knowledge base with common support questions

Optimizing AI Responses

• Keep your knowledge base updated and comprehensive
• Use clear, structured content for better AI comprehension
• Include FAQs and common troubleshooting steps

Managing Handoffs

• Monitor conversations regularly for smooth handoffs
• Set up HelpScout workflows to route unassigned conversations
• Train your team on when to reassign to the AI agent

Scaling Up

• Switch to All Messages mode once you're confident in AI responses
• Use First Only mode for high-touch customers
• Review conversation metrics to measure AI effectiveness

Frequently Asked Questions