Use Cases Pricing Docs
EN SL HR
Log in Start Free

AI Suggestions

AI-powered response suggestions to help you reply faster

AI-Powered Response Suggestions

Hal's AI suggestions help you respond to customer messages faster and more consistently. The AI analyzes the conversation context and your knowledge base to suggest relevant responses.

How It Works

  1. Visitor sends a message

    When a visitor sends a message through the chat widget, Hal captures the full conversation context.

  2. AI analyzes the context

    The AI reviews the conversation history, visitor metadata, and searches your knowledge base for relevant information.

  3. Get suggested responses

    Click the "AI Suggest" button in the message input area to generate a response suggestion based on the conversation context.

  4. Review and send

    Review the suggested response, edit if needed, and send. The AI learns from your edits to improve future suggestions.

Enabling AI Suggestions

To start using AI suggestions:

  1. Go to Settings > AI Configuration
  2. Add your OpenAI API key
  3. Choose your preferred model (GPT-3.5 Turbo or GPT-4)
  4. Enable AI suggestions for your app
Note: You need to provide your own OpenAI API key. Hal uses your key to generate suggestions, so you're billed directly by OpenAI for usage.

Knowledge Base Integration

AI suggestions become more accurate and helpful when integrated with your knowledge base:

  • Automatic Context - The AI searches your articles for relevant information
  • Consistent Answers - Responses align with your documented policies and procedures
  • Source Citations - Suggestions include references to knowledge base articles when applicable
  • Multi-Language Support - AI can suggest responses in the visitor's language if your knowledge base has translations

To maximize AI effectiveness, create comprehensive knowledge base articles covering your most common support topics.

Custom Prompts

Pro Business

Customize the AI's behavior with custom system prompts:

You are a helpful customer support agent for Acme Inc.
Our tone is friendly and professional.
Always offer to escalate complex issues to a human agent.
Include relevant links to our documentation when appropriate.

Custom prompts allow you to:

  • Define your brand voice and tone
  • Set response guidelines and policies
  • Specify formatting preferences
  • Include industry-specific terminology

Usage Monitoring

Track your AI usage in Settings > AI Configuration:

  • Suggestions Generated - Total number of AI suggestions requested
  • Tokens Used - OpenAI token consumption for billing estimates
  • Average Response Time - How long AI takes to generate suggestions
  • Acceptance Rate - How often suggestions are used vs. edited or discarded

Best Practices

1. Build a Comprehensive Knowledge Base

The more information in your knowledge base, the better the AI suggestions. Include:

  • Common questions and answers
  • Product documentation
  • Troubleshooting guides
  • Company policies

2. Review Before Sending

Always review AI suggestions before sending. The AI is helpful but not perfect. Check for:

  • Accuracy of information
  • Appropriate tone for the situation
  • Relevance to the specific question
  • Sensitive information that shouldn't be shared

3. Provide Feedback

When you edit AI suggestions, you're teaching the system what works best for your use case. Over time, suggestions will better match your style.

4. Use Custom Prompts

Define your brand voice and guidelines in custom prompts. This ensures consistency across all AI-generated suggestions.

5. Monitor Token Usage

Keep an eye on your OpenAI token usage to manage costs. GPT-3.5 Turbo is more cost-effective for most use cases, while GPT-4 provides higher quality for complex queries.

Pricing and Limits

AI suggestions are available on all plans, but you need to provide your own OpenAI API key:

  • Free Plan - AI suggestions available with your OpenAI key
  • Pro Plan - Custom prompts, advanced usage analytics
  • Business Plan - Priority support, custom model fine-tuning (coming soon)

OpenAI charges based on token usage. See OpenAI's pricing page for current rates.

Privacy and Security

Hal takes privacy seriously:

  • Your OpenAI API key is encrypted at rest
  • Conversation data sent to OpenAI follows their data usage policies
  • You can disable AI suggestions at any time
  • No conversation data is stored by OpenAI (using non-training API)

Troubleshooting

AI suggestions not appearing

  • Verify your OpenAI API key is valid and active
  • Check that AI is enabled in Settings > AI Configuration
  • Ensure your OpenAI account has sufficient credits

Suggestions are not relevant

  • Add more content to your knowledge base
  • Create custom prompts to guide the AI
  • Try using GPT-4 for better context understanding

High token usage

  • Use GPT-3.5 Turbo instead of GPT-4
  • Keep knowledge base articles concise and focused
  • Limit conversation history sent to AI (configurable in settings)

Next Steps