The Key Components of AI Node

• Important Note: Question and System Prompt always go hand-in-hand,
  for example:

  Question	System Prompt
  “Is this message asking about opening hours? Answer YES or NO.”	“Only output YES or NO.”

  
  

  OR

  Question	System Prompt
  “Read $customer_input and determine if the customer is asking about pricing.”	“You are a strict classifier. No explanations. Answer Yes if $customer_input is about pricing, answer No if $customer_input is not related to pricing.”

  
  

Step-by-step Guide - Basic Setup

This setup is ideal when you want the AI to process a message once and immediately continue the flow (e.g., rewriting text, extracting intent, cleaning data).

Step 1: Start with an “Open Question” Widget

The AI Node requires input text to process.
Use a "Open Question" widget to collect that input:

1. Drag and release a "Open Question" widget,
   configure and click "Update":

   

   Edit Open Question
   Question: This message will be delivered to users
   Type: Define the type of data to be collected, in this scenario, please select "Text".
   Variable: Define the name of this variable, for instance: "customer_message"

   
   

2. Drag the connecting line from Start → Open Question.

3. The Open Question widget will capture any free-text message sent by the customer.

Storing the message as data is required to use an AI Node.

✅ A new variable customer_message acquired!

Step 2: Add an AI Node

This node will read the stored text and process it based on your instructions.

1. Drag and release a "AI Node" widget.

2. Inside the "AI Node" widget, configure:

   • AI Model

   • Knowledge Base (use Empty Knowledge Base)

   • Question (the task)

   • System Prompt (output rules)

   • Variable (for example ai_result )

     (Refer to the guide above for detailed configuration.)

3. When done, click Update.

4. Drag the connecting line from Open Question → AI Node.

✅ The customer_message variable has been processed and AI Node has produced an output ai_result.

Step 3: Add a “Send Message” Widget

This is where you display or use the AI’s output, in this case we are demonstrating with a simple "Send Message" widget:

1. Drag and release to create a "Send Message" widget.

2. Configure the widget by typing in your message.

3. Insert the AI Node’s output variable (i.e., ai_result).

4. When done, click Update.

5. Then connect AI Node → Send Message.

Step 4: Save the AI Agent.

Click Save. Your Basic AI Agent setup is complete.

This setup is best for:

• Rewriting messages

• Extracting clean data

• Identifying customer intent

• Auto-tagging or routing

• Storing clean data into CRM

Use Case: Marketing team wants clean answers from customers, even if they reply with typos or slang.

In this step-by-step user guide, we will walk you through setting up AI Node to clean or transform customer messages:

1. Logon to iSlash AI Platform at https://app.islash.io/

2. Go to AI Agents and click Create.

3. Complete the left panel settings,
   including Name, Channels, Status, Start Condition and Exit Condition.
   
   

4. Drag and release Open Question, connect to Start and update columns:

   

   Edit Open Question	Input Suggestions
   Question : input greeting message	"Hi, thanks for enquiring with us."
   Type : Test / Media	select Text
   Variable : input variable name	“customer_input”

   
   

5. Drag and release AI Node, update columns:

   

   Edit AI Node	Input Suggestions
   AI Model :	select ChatGPT 4o.
   Knowledge Base :	select “Open AI - Empty Knowledge Base (ADA 002)”
   Question : input your intended process	“Rewrite the message $customer_input into proper English.”
   System Prompt : provide instruction to AI	“Do not change meaning. Keep it simple.”
   Error Message
   Send Error Message with Button	Disable
   Must Answer Based on Knowledge Base	Disable
   With Chat History	Disable
   Include sources in the reply	Disable
   Variable : customize the output variable name	“clean_message_1”

   
   

6. Click “Update” to save, connect to the widget before.
   
   

7. Drag and release Send Message, input message to be sent.

   

   You will be able to insert the output content by selecting the variable from the variable list.
   
   

8. Click “Update” to save,
   connect the AI Node earlier to Send Message.
   
   

9. At the top right corner, click Save to complete the AI Agent that is able to process and confirm customer’s message.

The Results

Customer sends a typo-heavy reply
: “i wan the promo code pls ty”

AI Node rewrites
: “I want the promo code, please.”

Reply to customer
: “Just to confirm, you said: Could you please provide the promo code? Thank you.”