Skip to main content

AI Fallback

Let AI handle customer conversations when you're away or busy.

How It Works​

  1. Customer sends a message
  2. If you don't respond within X minutes, AI takes over
  3. AI responds based on your custom prompt and context
  4. You can take over anytime by sending a message
  5. All AI responses are marked so customers know

Configuration​

SaaS Users​

Go to your AI Settings to configure:

  • Enable/disable AI fallback
  • Set response delay (how long to wait before AI responds)
  • Customize the AI system prompt
  • Set business hours (AI only active outside hours)

Self-Hosted Users​

Add to your .env file:

.env
# OpenAI API key
OPENAI_API_KEY=sk-...

# AI Configuration
AI_ENABLED=true
AI_RESPONSE_DELAY=60  # seconds
AI_MODEL=gpt-4o-mini  # or gpt-4o

# Custom system prompt
AI_SYSTEM_PROMPT="You are a helpful customer support agent for Acme Inc. Be friendly and concise. If you don't know something, say so."

Custom System Prompt​

Write a prompt that defines your AI's personality and knowledge:

You are a friendly customer support agent for PocketPing, a customer chat tool.

Key information:
- PocketPing costs $19/month for Pro, $49/month for Team
- Free tier includes 100 sessions per month
- We support Telegram, Discord, and Slack

Guidelines:
- Be concise and helpful
- If asked about pricing, provide the above information
- If you don't know something, say "I'll have a team member get back to you"
- Never make up features or pricing

Context Variables​

Include dynamic context in your prompts:

You are helping a visitor on {{page_url}}.

Visitor info:
- Name: {{visitor_name}}
- Email: {{visitor_email}}
- Plan: {{visitor_plan}}

Previous messages in this conversation are provided below.

Available variables:

  • {{page_url}} - Current page URL
  • {{visitor_name}} - Visitor name (if identified)
  • {{visitor_email}} - Visitor email (if identified)
  • {{visitor_plan}} - Customer plan (if identified)
  • {{company_name}} - Your company name
  • {{current_time}} - Current time

Business Hours​

Configure AI to only respond outside business hours:

.env
AI_BUSINESS_HOURS_ONLY=true
AI_BUSINESS_START=09:00
AI_BUSINESS_END=17:00
AI_BUSINESS_TIMEZONE=America/New_York
AI_BUSINESS_DAYS=1,2,3,4,5  # Monday to Friday

Manual Override​

From Telegram​

In a conversation topic, use:

/ai off   # Disable AI for this conversation
/ai on    # Re-enable AI for this conversation

From Discord​

/ai off
/ai on

From the API​

POST /api/sessions/{sessionId}/ai
{
  "enabled": false
}

Monitoring AI Responses​

All AI responses include metadata:

{
  "id": "msg_xxx",
  "content": "Hello! I'd be happy to help...",
  "type": "ai",
  "metadata": {
    "ai_model": "gpt-4o-mini",
    "ai_latency_ms": 1234
  }
}

In Telegram/Discord, AI responses are prefixed with 🤖.

Best Practices​

  1. Start with a delay - Give yourself 1-2 minutes to respond before AI kicks in
  2. Be specific in prompts - Include pricing, features, and common questions
  3. Set boundaries - Tell AI what it shouldn't do (e.g., don't promise refunds)
  4. Monitor responses - Review AI conversations periodically
  5. Use business hours - Let AI handle off-hours inquiries

Example Prompts​

E-commerce Support​

You are a customer support agent for an online store.

Handle these common questions:
- Shipping: We ship within 2-3 business days. Free shipping over $50.
- Returns: 30-day return policy. Items must be unused.
- Order status: Direct customers to check their email for tracking.

For refund or complaint issues, say "I'll have our team look into this right away."

SaaS Support​

You are a support agent for a SaaS product.

Pricing:
- Free: Up to 5 users
- Pro ($29/mo): Up to 25 users
- Enterprise: Contact sales

For technical issues, collect details and say a developer will follow up.

Next Steps​