Automated product email support with GPT-4.1, Pinecone RAG, and Gmail
Description This workflow is built for e-commerce store owners, customer support teams, and retail businesses who want to provide instant, intelligent email support without hiring additional staff. If you're tired of manually responding to customer inquiries, searching through product catalogs, and copying information into emails, this automation will transform your support process. It turns your inbox into a smart AI-powered support system that reads, understands, and responds to customer questions about your store products while you focus on growing your business. How it works / What it does:
This n8n automation completely transforms how you handle customer email inquiries using AI and Retrieval-Augmented Generation (RAG) technology.
Here's a simple breakdown of how it works:
- Monitor your Gmail inbox using the Gmail Trigger node, which checks every minute for new customer emails (excluding emails sent by you).
- Assess if a reply is needed with an AI-powered classification system. The workflow uses GPT-4.1 with a structured JSON parser to determine whether incoming emails are genuine customer inquiries about your men's clothing store that require a response.
- Filter relevant emails through the If Needs Reply node, which only passes emails that need attention to the AI Agent, preventing unnecessary processing.
- Generate intelligent responses using an AI Agent powered by GPT-4.1-nano. The agent uses a friendly, professional tone and starts each email with "Dear" and ends with "Best regards" to maintain proper email etiquette.
- Search your knowledge base with a Vector Store RAG tool connected to Pinecone. The AI Agent queries your men's clothing product database using OpenAI embeddings to find accurate, up-to-date information about prices, features, and product details.
- Send personalized replies automatically through the Gmail node, which responds directly to the original email thread with clear, concise, and empathetic answers to customer questions.
Once everything is set up, the system runs on autopilot and provides 24/7 customer support without any manual intervention.
How to set up:
Follow these steps to get your AI-powered email support system running:
-
Import the JSON file into your n8n instance.
-
Add your API credentials:
- Gmail OAuth2 credentials for reading and sending emails
- OpenAI API key for the AI Agent and embeddings
- Pinecone API credentials for vector storage
-
Set up your Pinecone vector database:
- Create a Pinecone index.
- Create a namespace.
- Upload your store data to the vector store
-
Configure the Gmail Trigger node to monitor the correct email account.
-
Customize the AI Agent's system message to match your brand voice and support policies.
-
Activate the workflow to enable automatic monitoring and responses.
Requirements:
Before running the workflow, make sure you have the following:
- An n8n account or instance (self-hosted or n8n Cloud)
- A Gmail account for receiving and sending customer emails
- OpenAI API access for the AI Agent and embeddings (GPT-4.1 and GPT-4.1-nano models)
- A Pinecone account with a configured vector database containing your product information
- Your store data, product catalog prepared and uploaded to Pinecone
How to customize the workflow:
This workflow is flexible and can be customized to fit your business needs. Here's how you can tailor it:
- Adjust the response style by modifying the system message in the AI Agent node. You can make it more casual, formal, or brand-specific.
- Add response length controls by updating the system message instructions.
- Currently set to keep responses short and concise, you can adjust this for more detailed explanations.
- Change the polling frequency in the Gmail Trigger node. The default is every minute, but you can adjust it to check more or less frequently based on your email volume.
- Filter specific types of emails by modifying the filters in the Gmail Trigger and "Assess if message needs a reply" nodes to handle specific subjects, senders, or keywords.
- Connect to different email platforms by replacing the Gmail nodes with other email services like Outlook, IMAP, or customer support platforms.
- Add human-in-the-loop approval by inserting a webhook or notification node before the Gmail reply node, allowing manual review before sending responses.
- Implement response tracking by adding database nodes to log all AI-generated responses for quality control and training purposes.
- Add multi-language support by incorporating translation nodes or configuring the AI Agent to detect and respond in the customer's language.
Automated Product Email Support with GPT-4 and Pinecone RAG
This n8n workflow automates the process of providing email support for product-related queries using advanced AI capabilities. It leverages GPT-4 for intelligent responses and Pinecone for Retrieval Augmented Generation (RAG) to ensure accurate and context-aware replies, all triggered by incoming Gmail messages.
What it does
This workflow streamlines your product support by:
- Monitoring Gmail: It continuously listens for new emails in a specified Gmail inbox.
- Filtering Emails: It checks the subject line of incoming emails. If the subject is "Product Support," the email proceeds for AI processing. Other emails are ignored.
- AI-Powered Response Generation:
- It uses an OpenAI Chat Model (likely GPT-4 based on the directory name) to understand the user's query.
- It employs OpenAI Embeddings to convert the query into a vector representation.
- It queries a Pinecone Vector Store (acting as a knowledge base) to retrieve relevant product information using the generated embeddings. This is the RAG component, ensuring responses are grounded in your product documentation.
- A Vector Store Question Answer Tool is used to formulate answers based on the retrieved information.
- An AI Agent orchestrates these steps, planning and executing the necessary actions to generate a comprehensive and accurate response.
- Structured Output Parsing: It processes the AI-generated response to ensure it adheres to a predefined structure.
- Sending Replies via Gmail: Once a suitable response is generated, the workflow automatically sends a reply to the original sender using Gmail.
Prerequisites/Requirements
To use this workflow, you will need:
- n8n Instance: A running n8n instance (self-hosted or cloud).
- Gmail Account: A Google account with Gmail enabled, configured as a credential in n8n for both triggering and sending emails.
- OpenAI API Key: An API key for OpenAI to access their chat models (e.g., GPT-4) and embedding services.
- Pinecone Account: An account with Pinecone, and an existing index populated with your product documentation or relevant knowledge base. This will be used for the Retrieval Augmented Generation (RAG) component.
Setup/Usage
- Import the Workflow: Download the provided JSON and import it into your n8n instance.
- Configure Credentials:
- Gmail Trigger & Gmail Node: Set up a Google OAuth2 credential for your Gmail account. Ensure it has permissions to read and send emails.
- OpenAI Chat Model & Embeddings OpenAI: Configure your OpenAI API Key credential.
- Pinecone Vector Store: Configure your Pinecone API Key and Environment credentials.
- Customize Nodes:
- Gmail Trigger: Specify the mailbox you want to monitor (e.g., "INBOX"). You might want to add further filters if needed.
- If Node: The current condition checks for "Product Support" in the subject. Adjust this condition (
{{ $json.subject.includes('Product Support') }}) to match the specific subject lines or other email properties you want to trigger this workflow. - Pinecone Vector Store: Ensure the "Index Name" and "Namespace" (if used) are correctly configured to point to your Pinecone index containing product information.
- OpenAI Chat Model: You can adjust the model (e.g.,
gpt-4,gpt-3.5-turbo) and other parameters like temperature as needed. - Gmail Node (Send Email): Customize the "To" field (likely
{{ $json.fromEmail }}), "Subject" (e.g., "Re: {{ $json.subject }}"), and the "Body" of the email to include the AI-generated response.
- Activate the Workflow: Once all configurations are complete, activate the workflow. It will start monitoring your Gmail for new emails matching the specified criteria.
Related Templates
Automate interior design lead qualification with AI & human approval to Notion
Overview This automated workflow intelligently qualifies interior design leads, generates personalized client emails, and manages follow-up through a human-approval process. Built with n8n, Claude AI, Telegram approval, and Notion database integration. ⚠️ Hosting Options This template works with both n8n Cloud and self-hosted instances. Most nodes are native to n8n, making it cloud-compatible out of the box. What This Template Does Automated Lead Management Pipeline: Captures client intake form submissions from website or n8n forms AI-powered classification into HOT/WARM/COLD categories based on budget, project scope, and commitment indicators Generates personalized outreach emails tailored to each lead type Human approval workflow via Telegram for quality control Email revision capability for rejected drafts Automated client email delivery via Gmail Centralized lead tracking in Notion database Key Features ✅ Intelligent Lead Scoring: Analyzes 12+ data points including budget (AED), space count, project type, timeline, and style preferences ✅ Personalized Communication: AI-generated emails reference specific client details, demonstrating genuine understanding ✅ Quality Control: Human-in-the-loop approval via Telegram prevents errors before client contact ✅ Smart Routing: Different workflows for qualified leads (meeting invitations) vs. unqualified leads (respectful alternatives) ✅ Revision Loop: Rejected emails automatically route to revision agent for improvements ✅ Database Integration: All leads stored in Notion for pipeline tracking and analytics Use Cases Interior design firms managing high-volume lead intake Architecture practices with complex qualification criteria Home renovation companies prioritizing project value Any service business requiring budget-based lead scoring Sales teams needing approval workflows before client contact Prerequisites Required Accounts & API Keys: Anthropic Claude API - For AI classification and email generation Telegram Bot Token - For approval notifications Gmail Account - For sending client emails (or any SMTP provider) Notion Account - For lead database storage n8n Account - Cloud or self-hosted instance Technical Requirements: Basic understanding of n8n workflows Ability to create Telegram bots via BotFather Gmail app password or OAuth setup Notion database with appropriate properties Setup Instructions Step 1: Clone and Import Template Copy this template to your n8n instance (cloud or self-hosted) All nodes will appear as inactive - this is normal Step 2: Configure Form Trigger Open the Client Intake Form Trigger node Choose your trigger type: For n8n forms: Configure form fields matching the template structure For webhook: Copy webhook URL and integrate with your website form Required form fields: First Name, Second Name, Email, Contact Number Project Address, Project Type, Spaces Included Budget Range, Completion Date, Style Preferences Involvement Level, Previous Experience, Inspiration Links Step 3: Set Up Claude AI Credentials Obtain API key from https://console.anthropic.com In n8n: Create new credential → Anthropic → Paste API key Apply credential to these nodes: AI Lead Scoring Engine Personalized Client Outreach Email Generator Email Revision Agent Step 4: Configure Telegram Approval Bot Create bot via Telegram's @BotFather Copy bot token Get your Telegram Chat ID (use @userinfobot) In n8n: Create Telegram credential with bot token Configure Human-in-the-Loop Email Approval node: Add your Chat ID Customize approval message format if desired Step 5: Set Up Gmail Sending Enable 2-factor authentication on Gmail account Generate app password: Google Account → Security → App Passwords In n8n: Create Gmail credential using app password Configure Client Email Delivery node with sender details Step 6: Connect Notion Database Create Notion integration at https://www.notion.so/my-integrations Copy integration token Create database with these properties: Client Name (Title), Email (Email), Contact Number (Phone) Project Address (Text), Project Type (Multi-select) Spaces Included (Text), Budget (Select), Timeline (Date) Classification (Select: HOT/WARM/COLD), Confidence (Select) Estimated Value (Number), Status (Select) Share database with your integration In n8n: Add Notion credential → Paste token Configure Notion Lead Database Manager with database ID Step 7: Customize Classification Rules (Optional) Open AI Lead Scoring Engine node Review classification criteria in the prompt: HOT: 500k+ AED, full renovations, 2+ spaces WARM: 100k+ AED, 2+ spaces COLD: <100k AED OR single space Adjust thresholds to match your business requirements Modify currency if not using AED Step 8: Personalize Email Templates Open Personalized Client Outreach Email Generator node Customize: Company name and branding Signature placeholders ([Your Name], [Title], etc.) Tone and style preferences Alternative designer recommendations for COLD leads Step 9: Test the Workflow Activate the workflow Submit a test form with sample data Monitor each node execution in n8n Check Telegram for approval message Verify email delivery and Notion database entry Step 10: Set Up Error Handling (Recommended) Add error workflow trigger Configure notifications for failed executions Set up retry logic for API failures Workflow Node Breakdown Client Intake Form Trigger Captures lead data from website forms or n8n native forms with all project details. AI Lead Scoring Engine Analyzes intake data using structured logic: budget validation, space counting, and multi-factor evaluation. Returns HOT/WARM/COLD classification with confidence scores. Lead Classification Router Routes leads into three priority workflows based on AI classification, optimizing resource allocation. Sales Team Email Notifier Sends instant alerts to sales representatives with complete lead details and AI reasoning for internal tracking. Personalized Client Outreach Email Generator AI-powered composer creating tailored responses demonstrating genuine understanding of client vision, adapted by lead type. Latest Email Version Controller Captures most recent email output ensuring only final approved version proceeds to delivery. Human-in-the-Loop Email Approval Telegram-based review checkpoint sending generated emails to team member for quality control before client delivery. Approval Decision Router Evaluates reviewer's response, routing approved emails to client delivery or rejected emails to revision agent. Email Revision Agent AI-powered editor refining rejected emails based on feedback while maintaining personalization and brand voice. Client Email Delivery Sends final approved personalized emails demonstrating understanding of project vision with clear next steps. Notion Lead Database Manager Records all potential clients with complete intake data, classification results, and tracking information for pipeline management. Customization Tips Adjust Classification Thresholds: Modify budget ranges and space requirements in the AI Lead Scoring Engine prompt to match your market and service level. Multi-Language Support: Update all AI agent prompts with instructions for your target language. Claude supports 100+ languages. Additional Routing: Add branches for special cases like urgent projects, VIP clients, or specific geographic regions. CRM Integration: Replace Notion with HubSpot, Salesforce, or Airtable using respective n8n nodes. SMS Notifications: Add Twilio node for immediate HOT lead alerts to mobile devices. Troubleshooting Issue: Telegram approval not received Verify bot token is correct Confirm chat ID matches your Telegram account Check bot is not blocked Issue: Claude API errors Verify API key validity and credits Check prompt length isn't exceeding token limits Review rate limits on your Anthropic plan Issue: Gmail not sending Confirm app password (not regular password) is used Check "Less secure app access" if using older method Verify daily sending limits not exceeded Issue: Notion database not updating Confirm integration has access to database Verify property names match exactly (case-sensitive) Check property types align with data being sent Template Metrics Execution Time: ~30-45 seconds per lead (including AI processing) API Calls: 2-3 Claude requests per lead (classification + email generation, +1 if revision) Cost Estimate: ~$0.05-0.15 per lead processed (based on Claude API pricing) Support & Community n8n Community Forum: https://community.n8n.io Template Issues: Report bugs or suggest improvements via n8n template feedback Claude Documentation: https://docs.anthropic.com Notion API Docs: https://developers.notion.com License This template is provided as-is under MIT license. Modify and adapt freely for your business needs. --- Version: 1.0 Last Updated: October 2025 Compatibility: n8n v1.0+ (Cloud & Self-Hosted), Claude API v2024-10+
Automated UGC video generator with Gemini images and SORA 2
This workflow automates the creation of user-generated-content-style product videos by combining Gemini's image generation with OpenAI's SORA 2 video generation. It accepts webhook requests with product descriptions, generates images and videos, stores them in Google Drive, and logs all outputs to Google Sheets for easy tracking. Main Use Cases Automate product video creation for e-commerce catalogs and social media. Generate UGC-style content at scale without manual design work. Create engaging video content from simple text prompts for marketing campaigns. Build a centralized library of product videos with automated tracking and storage. How it works The workflow operates as a webhook-triggered process, organized into these stages: Webhook Trigger & Input Accepts POST requests to the /create-ugc-video endpoint. Required payload includes: product prompt, video prompt, Gemini API key, and OpenAI API key. Image Generation (Gemini) Sends the product prompt to Google's Gemini 2.5 Flash Image model. Generates a product image based on the description provided. Data Extraction Code node extracts the base64 image data from Gemini's response. Preserves all prompts and API keys for subsequent steps. Video Generation (SORA 2) Sends the video prompt to OpenAI's SORA 2 API. Initiates video generation with specifications: 720x1280 resolution, 8 seconds duration. Returns a video generation job ID for polling. Video Status Polling Continuously checks video generation status via OpenAI API. If status is "completed": proceeds to download. If status is still processing: waits 1 minute and retries (polling loop). Video Download & Storage Downloads the completed video file from OpenAI. Uploads the MP4 file to Google Drive (root folder). Generates a shareable Google Drive link. Logging to Google Sheets Records all generation details in a tracking spreadsheet: Product description Video URL (Google Drive link) Generation status Timestamp Summary Flow: Webhook Request → Generate Product Image (Gemini) → Extract Image Data → Generate Video (SORA 2) → Poll Status → If Complete: Download Video → Upload to Google Drive → Log to Google Sheets → Return Response If Not Complete: Wait 1 Minute → Poll Status Again Benefits: Fully automated video creation pipeline from text to finished product. Scalable solution for generating multiple product videos on demand. Combines cutting-edge AI models (Gemini + SORA 2) for high-quality output. Centralized storage in Google Drive with automatic logging in Google Sheets. Flexible webhook interface allows integration with any application or service. Retry mechanism ensures videos are captured even with longer processing times. --- Created by Daniel Shashko
Track personal finances in Google Sheets with AI agent via Slack
Who's it for This workflow is perfect for individuals who want to maintain detailed financial records without the overhead of complex budgeting apps. If you prefer natural language over data entry forms and want an AI assistant to handle the bookkeeping, this template is for you. It's especially useful for: People who want to track cash and online transactions separately Anyone who lends money to friends/family and needs debt tracking Users comfortable with Slack as their primary interface Those who prefer conversational interactions over manual spreadsheet updates What it does This AI-powered finance tracker transforms your Slack workspace into a personal finance command center. Simply mention your bot with transactions in plain English (e.g., "₹500 cash food, borrowed ₹1000 from John"), and the AI agent will: Parse transactions using natural language understanding via Google Gemini Calculate balance changes for cash and online accounts Show a preview of changes before saving anything Update Google Sheets only after you approve Track debts (who owes you, who you owe, repayments) Send daily reminders at 11 PM with current balances and active debts The workflow maintains conversational context using PostgreSQL memory, so you can say things like "yesterday's transactions" or "that payment to Sarah" and it understands the context. How it works Scheduled Daily Check-in (11 PM) Fetches current balances from Google Sheets Retrieves all active debts Formats and sends a Slack message with balance summary Prompts you to share the day's transactions AI Agent Transaction Processing When you mention the bot in Slack: Phase 1: Parse & Analyze Extracts amount, payment type (cash/online), category (food, travel, etc.) Identifies transaction type (expense, income, borrowed, lent, repaid) Stores conversation context in PostgreSQL memory Phase 2: Calculate & Preview Reads current balances from Google Sheets Calculates new balances based on transactions Shows formatted preview with projected changes Waits for your approval ("yes"/"no") Phase 3: Update Database (only after approval) Logs transactions with unique IDs and timestamps Updates debt records with person names and status Recalculates and stores new balances Handles debt lifecycle (Active → Settled) Phase 4: Confirmation Sends success message with updated balances Shows active debts summary Includes logging timestamp Requirements Essential Services: n8n instance (self-hosted or cloud) Slack workspace with admin access Google account Google Gemini API key PostgreSQL database Recommended: Claude AI model (mentioned in workflow notes as better alternative to Gemini) How to set up Google Sheets Setup Create a new Google Sheet with three tabs named exactly: Balances Tab: | Date | CashBalance | OnlineBalance | Total_Balance | |------|--------------|----------------|---------------| Transactions Tab: | TransactionID | Date | Time | Amount | PaymentType | Category | TransactionType | PersonName | Description | Added_At | |----------------|------|------|--------|--------------|----------|------------------|-------------|-------------|----------| Debts Tab: | PersonName | Amount | Type | Datecreated | Status | Notes | |-------------|--------|------|--------------|--------|-------| Add header rows and one initial balance row in the Balances tab with today's date and starting amounts. Slack App Setup Go to api.slack.com/apps and create a new app Under OAuth & Permissions, add these Bot Token Scopes: app_mentions:read chat:write channels:read Install the app to your workspace Copy the Bot User OAuth Token Create a dedicated channel (e.g., personal-finance-tracker) Invite your bot to the channel Google Gemini API Visit ai.google.dev Create an API key Save it for n8n credentials setup PostgreSQL Database Set up a PostgreSQL database (you can use Supabase free tier): Create a new project Note down connection details (host, port, database name, user, password) The workflow will auto-create the required table n8n Workflow Configuration Import the workflow and configure: A. Credentials Google Sheets OAuth2: Connect your Google account Slack API: Add your Bot User OAuth Token Google Gemini API: Add your API key PostgreSQL: Add database connection details B. Update Node Parameters All Google Sheets nodes: Select your finance spreadsheet Slack nodes: Select your finance channel Schedule Trigger: Adjust time if you prefer a different check-in hour (default: 11 PM) Postgres Chat Memory: Change sessionKey to something unique (e.g., financetrackeryour_name) Keep tableName as n8nchathistory_finance or rename consistently C. Slack Trigger Setup Activate the "Bot Mention trigger" node Copy the webhook URL from n8n In Slack App settings, go to Event Subscriptions Enable events and paste the webhook URL Subscribe to bot event: app_mention Save changes Test the Workflow Activate both workflow branches (scheduled and agent) In your Slack channel, mention the bot: @YourBot ₹100 cash snacks Bot should respond with a preview Reply "yes" to approve Verify Google Sheets are updated How to customize Change Transaction Categories Edit the AI Agent's system message to add/remove categories. Current categories: travel, food, entertainment, utilities, shopping, health, education, other Modify Daily Check-in Time Change the Schedule Trigger's triggerAtHour value (0-23 in 24-hour format). Add Currency Support Replace ₹ with your currency symbol in: Format Daily Message code node AI Agent system prompt examples Switch AI Models The workflow uses Google Gemini, but notes recommend Claude. To switch: Replace "Google Gemini Chat Model" node Add Claude credentials Connect to AI Agent node Customize Debt Types Modify AI Agent's system prompt to change debt handling logic: Currently: IOwe and TheyOwe_Me You can add more types or change naming Add More Payment Methods Current: cash, online To add more (e.g., credit card): Update AI Agent prompt Modify Balances sheet structure Update balance calculation logic Change Approval Keywords Edit AI Agent's Phase 2 approval logic to recognize different approval phrases. Add Spending Analytics Extend the daily check-in to calculate: Weekly/monthly spending summaries Category-wise breakdowns Use additional Code nodes to process transaction history Important Notes ⚠️ Never trigger with normal messages - Only use app mentions (@botname) to avoid infinite loops where the bot replies to its own messages. 💡 Context Awareness - The bot remembers conversation history, so you can reference "yesterday", "last week", or previous transactions naturally. 🔒 Data Privacy - All your financial data stays in your Google Sheets and PostgreSQL database. The AI only processes transaction text temporarily. 📊 Backup Regularly - Export your Google Sheets periodically as backup. --- Pro Tips: Start with small test transactions to ensure everything works Use consistent person names for debt tracking The bot understands various formats: "₹500 cash food" = "paid 500 rupees in cash for food" You can batch transactions in one message: "₹100 travel, ₹200 food, ₹50 snacks"