Build a Smart Chat Agent with Dumpling AI Agents That Finds Local Businesses and Auto-Saves Them to Airtable

Introduction

This workflow transforms chat messages into real business data. Instead of manually searching Google or Yelp, your AI assistant can now:

• Understand chat questions using GPT-4o
  
• Use Dumpling AI to find local business info
  
• Save all structured business data directly into Airtable
  

Why this is powerful:

• It connects language understanding with real-time data gathering
  
• Saves time by fully automating lead collection
  
• Makes business outreach, research, and CRM building much faster
  

Let’s walk through this system step-by-step.

Step 1: Trigger – When a Chat Message Is Received

What happens here:
This node listens for incoming chat messages. It could be integrated with a chat app, form, or custom interface.

• Node: When chat message received
  
• It starts the workflow when a user sends a message.

Trigger:

Pro Tip:
You can connect this to a frontend app using a webhook or embed it into any internal tool.

Step 2: Process Input with GPT-4o

What happens here:
The assistant uses GPT-4o to understand the user’s request. For example, if the user says, “Find top coffee shops in Chicago,” GPT interprets the goal: location, category, intent.

• Node: Process Input with GPT-4o Mini
  
• Connects to OpenAI API to process natural language input.

Input:

Node:

• 

Pro Tip:
You can tweak the prompt to be more domain-specific, like “Only show restaurants with more than 100 reviews.”

Step 3: Track the Last 10 Messages for Memory

What happens here:
This memory buffer lets the assistant remember the last 10 messages. It ensures continuity and keeps the conversation flowing naturally.

• Node: Track Last 10 Messages
  
• Helps GPT-4o understand user context across multiple messages.

Node:

Output:

Pro Tip:
Use this to personalize results. For example, if the user asks for “Italian restaurants” and later says “same area”, the bot will remember.

Step 4: Route the AI Tasks with GPT, Memory, and Tools

What happens here:
This central node connects all logic: chat understanding, memory, and tool usage. It decides whether to run a local business search or a news search.

• Node: Route AI Tasks with Dumpling + GPT + Memory
  
• System message: “Use Local Business Finder for business search. Use News Agent for news.”

Node:

Output:

Pro Tip:
You can route based on keyword intent. For example, “find”, “search”, “near me” → Local Business Agent. “latest”, “headline”, “breaking” → News Agent.

Step 5: Use Dumpling AI – Local Business Finder Agent

What happens here:
This is the business intelligence engine. It sends the processed query to Dumpling AI, which scrapes Google Maps and local directories.

• Node: Local Business Finder Agent
  
• POSTs the input to Dumpling AI using the agent ID for business lookup.
  

JSON Payload Example:

{

  “messages”: [

    {

      “role”: “user”,

      “content”: “{{ $json.chatInput }}”

    }

  ],

  “agentId”: “7a8cad79-2d74-4ff4-9688-072f74f3d764”,

  “parseJson”: “false”

}

Output:

Node:

Pro Tip:

You can modify the agent in Dumpling AI to focus on specific industries, like lawyers, gyms, or retail shops

Step 6: Use Dumpling AI – News Agent 

What happens here:
If the query is about current events, this node handles that request.

• Node: News Agent
  
• Calls another Dumpling AI agent for news summaries.
  

Output:

Node:

Pro Tip:
This step is optional. You can disable it if you want your bot focused only on local businesses.

Step 7: Save Structured Business Data to Airtable

What happens here:
The assistant now takes the structured response from Dumpling AI (business name, address, phone number, etc.) and auto-saves it into your Airtable base.

• Node: Save Business Results
  
• Airtable operation: Upsert
  
• Maps data from Dumpling AI to the following fields:
  
  • Name of Restaurant
    
  • Address
    
  • Phone Number
    
  • Rating
    
  • Total Reviews
    
  • Website
    
  • Category

Output:

Node:

Pro Tip:
Use “Upsert” to avoid duplicates. You can use “id” or “phone number” as the unique identifier.

Final Workflow Structure

Your final automation flow should look like this:

• Chat Message Triggered →
  
• Message Understood by GPT-4o →
  
• Memory Tracks Conversation →
  
• Request Routed →
  
• Dumpling AI Agent Finds Local Businesses →
  
• Results Saved to Airtable

Complete Setup:

Conclusion

This system turns your AI assistant into a real-time lead collector and data logger. It can handle conversations, perform research, and save business insights without lifting a finger.

What makes this workflow powerful:

• Combines GPT’s intelligence with Dumpling AI’s real-time scraping
  
• Removes the manual effort of researching and logging leads
  
• Scales effortlessly for any region or business category
  

✅ Best Use Cases:

• Local lead generation
  
• Competitor research
  
• Market discovery
  
• Automated CRM population
  

Pro Tip:
You can expand this automation by:

• Notifying your team on Slack or email when new leads are added
  
• Filtering businesses based on ratings or number of reviews
  
• Automatically assigning leads to sales reps using Airtable automations
  

Download the blueprint used in this blog post

Click here to access the blueprint. Whether you’re just starting out or looking to refine your AI Automation skills, our community has got the tools and support to help you succeed.