Reply to Chat with AI

How to Use the “Reply to Chat With AI” Action in Botassium

If you want your WhatsApp automation to respond intelligently to incoming messages using AI, the “Reply to Chat With AI” action in Botassium is built for that. It lets you set up a fully autonomous assistant that can reply naturally based on user input—powered by the AI model of your choice.

This guide walks you through configuring this action for conversational replies, including model options, instructions, and custom training.

What is the “Reply to Chat With AI” Action?

The Reply to Chat With AI node is an action that uses a selected AI model to generate a response based on the last message from the user. The reply is sent immediately to the user without expecting additional input.

You can use it for:

• Handling frequently asked questions

• Replacing static replies with dynamic, human-like responses

• Creating a lightweight AI support assistant

• Responding with personalized or context-aware answers

Key Features

1. AI Model Selection

You can choose from:

• Your own OpenAI model (e.g., GPT-3.5, GPT-4)

• Your own OpenRouter model (e.g., Claude, Mixtral, etc.)

• The built-in free model mistralai/mistral-saba (no key required)

To use OpenAI or OpenRouter, you must provide your own API key in the Settings page.

2. Custom Instructions

Optionally, you can define specific behavior or tone for the AI.
Examples:

• “Act as a friendly sales assistant.”

• “Only reply with short summaries.”

• “Speak formally and avoid emojis.”

These instructions are sent as system-level prompts to shape how the model responds.

3. Optional Custom Training

You can attach a pre-built knowledge base to the AI response. This training can include:

• Uploaded PDF, DOCX, or TXT documents

• Website links you’ve added to your knowledge source

When training is selected, the model answers based not just on general knowledge, but also on your custom content, making it useful for:

• Company-specific FAQs

• Policy documents

• Product manuals

• Internal knowledge bases

You may also choose to skip training, in which case the model replies based on the latest user message and your optional instructions.

Example Use Case

You run a travel agency and want to auto-reply to general questions about your tour packages:

1. Use the Reply to Chat With AI action

2. Choose gpt-4 (via your OpenAI key)

3. Add custom instruction: “Act as a travel expert from XYZ agency. Always promote available packages politely.”

4. Attach a custom training that includes your travel brochure PDF

5. When a user asks, “Do you have tours to Dubai in November?”, the AI responds with up-to-date details from your training content

The Reply to Chat With AI action is supported across WhatsApp automation flows, and is especially helpful in smart customer support, FAQ bots, multilingual replies, lead qualification, and natural language understanding scenarios where static replies fall short.