Generate response suggestions

The core function of the AI Copilot: It analyzes the conversation and suggests an appropriate response.

How it works

1. Open a conversation in the inbox
2. Click on "AI Suggestion" (or the Copilot icon)
3. The Copilot generates a suggestion
4. You can use the suggestion:
   • Accept: Text is inserted into the input field
   • Edit: Modify the text before sending
   • Ignore: Write your own response

What the Copilot uses as context

The Copilot includes the following information:

Chat history

The last 30 messages of the conversation (Visitor + Agent + Bot).

Knowledge Base (RAG)

The Copilot searches for relevant KB articles and snippets based on the last visitor message — just like the chatbot, but for the agent.

Canned Responses

Your saved text modules are provided as additional context. The Copilot can use matching text modules as a basis or combine them.

Tenant Persona

The system prompt and custom instructions from the AI configuration are included so that the suggestion matches the tone of your company.

Improve the quality of suggestions

• Maintain the Knowledge Base: The better your KB, the better the suggestions
• Snippets for common questions: Snippets provide more precise answers than long articles
• Optimize the system prompt: A good prompt also improves the Copilot suggestions
• Keep canned responses up to date: The Copilot uses them as a reference

Token consumption

A response suggestion typically consumes 500–2000 tokens (depending on chat length and KB context). The consumption is tracked in the AI Usage.