AI Customer Data Query Agent with Google Sheets

Overview

This workflow creates an AI-powered customer data query assistant. It enables users to ask natural language questions about customer information stored in a Google Sheet, such as "Which is our biggest customer?" or "What is the email of customer row 5?". The AI agent (powered by OpenAI) intelligently selects and calls custom tools to retrieve the relevant data from the sheet, making it easy to explore and analyze structured data without writing any queries.

Step-by-Step Breakdown

1. When Chat Message Received (Chat Trigger) — No auth. Listens for incoming messages from the n8n chat interface. This is the entry point for user queries.
2. AI Agent (Agent) — No auth. Orchestrates the conversation, decides which tool to use based on the user's query, and passes the response back to the chat. It uses the OpenAI Chat Model and Simple Memory nodes for context.
3. OpenAI Chat Model (Language Model — OpenAI) — API Key auth. Provides the AI reasoning power. Configured to use gpt-4o-mini for cost-effective performance.
4. Simple Memory (Memory Buffer Window) — No auth. Maintains conversation history within a sliding window, enabling the agent to remember previous interactions.
5. List Columns Tool (Workflow Tool) — No auth. A custom sub-workflow tool that calls the sub-workflow below to return all column names from the Google Sheet.
6. Get Column Values Tool (Workflow Tool) — No auth. Another custom tool that returns all values for a specified column (e.g., all customer names).
7. Get Customer Tool (Workflow Tool) — No auth. Returns a full row of data for a given customer row number.
8. When Executed by Another Workflow (Execute Workflow Trigger) — No auth. Receives the operation (column_names, column_values, or row) and query string from the tools above.
9. Set Google Sheet URL (Set) — No auth. Extracts the sheet URL from an environment variable WEBHOOK_URL (you can change this to any URL or variable).
10. Get Google Sheet Contents (Google Sheets) — OAuth2. Fetches all rows from the sheet named customer_data using the provided document ID (URL). Requires Google Sheets credentials.
11. Check Operation (Switch) — No auth. Routes execution based on the operation field: goes to one of three branches (column_names, column_values, row).
12. Get Column Names (Set) — No auth. Extracts the keys (column headers) from the first row of the sheet data.
13. Prepare Column Data (Set) — No auth. For the column_values branch, it adds a row_number field to each row for later filtering.
14. Filter (Filter) — No auth. For the row branch, filters the sheet data to return only the row where row_number matches the user's query.
15. Prepare Output (Code) — No auth. Wraps the result into a JSON string under the response key, which is sent back to the calling tool.
16. Error Handler (Stop and Error) — No auth. Catches any errors and stops execution with a meaningful message.

Setup Instructions

1. Google Sheets — You need a Google Sheet with a sheet named customer_data. The first row should contain column headers (e.g., Name, Email, Revenue). Create Google Sheets OAuth2 credentials in n8n (see Google Sheets node docs).
2. OpenAI — Obtain an API key from OpenAI. In n8n, create OpenAI credentials (choose API Key auth) and use your key.
3. Environment Variable — By default, the workflow expects the sheet URL in the WEBHOOK_URL environment variable. You can replace this with a static URL or a credential field.
4. Chat Trigger — No special setup; the chat interface works out of the box when you click “Chat” in n8n.

Use Cases and Variations

• Customer Support — Let agents ask about customer accounts without accessing the raw sheet.
• Sales Analysis — Query sales data, e.g., “Which customer has the highest revenue?”
• Adapt to Other Data Sources — Replace the Google Sheets node with Airtable, PostgreSQL, or a REST API to query other databases.
• Different LLM — Swap gpt-4o-mini for another model (e.g., Claude, Gemini) by changing the Language Model node.
• Add More Tools — Extend the sub-workflow with tools for updating or deleting rows, making the agent bidirectional.

This workflow demonstrates a powerful pattern for turning static spreadsheets into conversational data assistants with minimal coding.

{
  "meta": {
    "instanceId": "workflow-de323dd0",
    "versionId": "1.0.0",
    "createdAt": "2025-09-29T07:07:42.480680",
    "updatedAt": "2025-09-29T07:07:42.480694",
    "owner": "n8n-user",
    "license": "MIT",
    "category": "automation",
    "status": "active",
    "priority": "high",
    "environment": "production"
  },
  "nodes": [
    {
      "id": "f3f7546a-8bb3-484c-b0a1-750a8d7d3a74",
      "name": "Sticky Note1",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        0,
        520
      ],
      "parameters": {
        "color": 7,
        "width": 1549,
        "height": 612,
        "content": "### Sub-workflow: Custom tool\nThis can be called by the agent above. It returns three different types of data from the Google Sheet, which can be used together for more complex queries without returning the whole sheet (which might be too big for GPT to handle)"
      },
      "typeVersion": 1,
      "notes": "This stickyNote node performs automated tasks as part of the workflow."
    },
    {
      "id": "a5afaa40-0b68-4d7c-8f78-5cb176fde81d",
      "name": "Sticky Note2",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        0,
        -40
      ],
      "parameters": {
        "color": 7,
        "width": 1068,
        "height": 547,
        "content": "### Main workflow: AI agent using custom tool"
      },
      "typeVersion": 1,
      "notes": "This stickyNote node performs automated tasks as part of the workflow."
    },
    {
      "id": "2dc0ce2f-a09c-4804-9962-f542ec78eb87",
// ... truncated (copy to see full JSON)