Suggest Reply
Generate AI-powered response suggestions based on conversation context
Suggest Reply generates a context-aware response based on the conversation history, helping operators reply faster and more consistently.
How to Use
- Open AI Actions → Suggest Reply
- Optionally add extra instructions to guide the AI
- The suggestion streams into the editor in real-time
- Review and edit before sending
How It Works
The AI analyzes the conversation history — including the contact's messages, previous operator replies, and the contact's profile — to craft a relevant response. The suggestion follows the tone, identity, and knowledge base configured in the active assistant.
The response is automatically formatted for the channel:
| Channel | Format |
|---|---|
| Telegram | HTML tags |
| WhatsApp markdown | |
| Viber / Instagram | Plain text |
Any {{variable}} placeholders in the conversation are preserved exactly as-is in the suggestion.
History Message Count
Configure how many recent messages the AI reads via the AI Preferences bar:
| Option |
|---|
| Auto |
| 10 messages |
| 20 messages |
| 30 messages |
| 50 messages |
| 75 messages |
| 100 messages |
When set to Auto, the system uses the organisation's inbox settings with a minimum of 7 messages.
Context Included
- Chat history — Recent messages labeled as customer or operator (system messages excluded)
- Contact profile — Customer profile fields are always included for context
- Knowledge base — Documents uploaded to the assistant (PDF, DOCX, TXT)
- Media — Audio transcripts as text, documents as text, images if the model supports vision
Empty Conversation
If the chat has no messages (e.g. deleted by message retention):
- With contact profile — The AI generates a reply based on profile context alone
- Without contact profile — Returns a "no context" error
Use Cases
- Respond quickly to common inquiries
- Maintain consistent tone across the team
- Draft initial replies for complex questions to refine