AI Agents

AI Agents #

AI agents are the heart of Support Unicorn's automation capabilities. They can understand customer questions, search your knowledge base, and provide accurate responses automatically.

How AI Agents Work #

When a customer sends a message:

  1. The AI agent analyzes the question
  2. It searches your knowledge base for relevant information
  3. It generates a response using that context
  4. Depending on your configuration, it either sends the response automatically or suggests it to your team

Creating an AI Agent #

To create a new AI agent:

  1. Navigate to AI Agents in the sidebar
  2. Click Create New Agent
  3. Fill in the required information:
    • Name: A descriptive name for your agent
    • Description: What this agent is responsible for
    • System Prompt: Instructions that guide the AI's behavior

Example System Prompt # 

You are a helpful customer support agent for Acme Corp.
Always be polite, professional, and concise.
If you don't know the answer, say so and offer to connect the customer with a human agent.

Training Your Agent #

The quality of your AI agent depends on the quality of your knowledge base. You can train your agent using:

Documentation #

Upload your product documentation as PDF, Markdown, or plain text files. The AI will automatically index and search these documents.

Q&A Pairs #

Add specific question-answer pairs for common customer inquiries. This is the fastest way to get your agent up and running.

Website Content #

Point the AI to your website or help center, and it will crawl and index the content automatically.

Text Snippets #

Add custom text blocks with important information like company policies, troubleshooting steps, or product specifications.

AI Response Modes #

Choose how the AI interacts with customers:

Auto Response #

The AI responds to all customer messages automatically. This is best for:

  • High-volume, repetitive questions
  • Off-hours support (nights and weekends)
  • Tier 1 support that can be fully automated

Hybrid Mode #

The AI generates a suggested response, but your team must approve it before it's sent. This is ideal for:

  • Teams new to AI support
  • Situations requiring human judgment
  • Learning what questions the AI can handle well

Fallback Mode #

The AI only responds when no human agents are available. Perfect for:

  • Teams who want to maintain a human-first approach
  • Providing 24/7 coverage with a small team
  • Handling overflow during peak times

Testing Your Agent #

Before enabling your AI agent for customers, test it thoroughly:

  1. Go to the agent's page
  2. Click Test Agent
  3. Ask it various customer questions
  4. Review the responses for accuracy
  5. Adjust your knowledge base or system prompt as needed

Monitoring AI Performance #

Track how your AI agent is performing:

  • Response Rate: Percentage of messages the AI handled
  • Escalation Rate: How often the AI had to escalate to humans
  • Customer Satisfaction: Ratings for AI responses
  • Common Gaps: Questions the AI couldn't answer

Use these insights to continuously improve your knowledge base.

Best Practices #

  1. Start Small: Begin with a limited knowledge base and expand based on actual questions
  2. Review Regularly: Check AI responses weekly to catch any issues
  3. Set Clear Boundaries: Configure when the AI should escalate to humans
  4. Keep Content Fresh: Update your knowledge base as your product evolves
  5. Use Feedback: Pay attention to customer reactions and adjust accordingly

Advanced Configuration #

Custom Actions #

Configure your AI agent to perform actions beyond just responding:

  • Create support tickets
  • Look up order information
  • Schedule callbacks
  • Update customer records

Multi-language Support #

Support Unicorn AI agents can communicate in multiple languages. Simply add translated versions of your knowledge base.

Context Awareness #

AI agents remember conversation history and can reference previous messages, providing more natural and helpful responses.

Troubleshooting #

Agent Gives Incorrect Answers #

  • Review and update your knowledge base
  • Check if the system prompt is too broad or vague
  • Add specific Q&A pairs for problematic questions

Agent Escalates Too Often #

  • Expand your knowledge base coverage
  • Adjust escalation thresholds
  • Review the types of questions being escalated

Agent Responses Too Verbose #

  • Update system prompt to be more concise
  • Provide examples of ideal response length
  • Enable response length limits

Need more help? Contact our support team at support@supportunicorn.com.