AI Website Chatbot with CRM Lead Collection Using GPT and Google Sheets

This n8n training workflow demonstrates how to connect a **sub-workflow** as a **tool** to an AI Agent. In this example, the main workflow is a **Website Chatbot** that engages visitors, collects contact information, and sends that data to a CRM process. The CRM process itself is a separate sub-workflow, connected to the agent as a tool via the **Tool Workflow** node. --- ## Step-by-Step Setup Instructions ### 1. Create the Sub-Workflow (CRM Tool) This sub-workflow will be triggered by the AI agent to process collected information. It will: 1. Receive inputs (email, description) from the main chatbot workflow. 2. Format the data into a structured JSON format. 3. Append the data to a Google Sheet (acting as the CRM database). 4. Send a confirmation message back to the main workflow. **Steps inside the sub-workflow:** - **When Executed by Another Workflow** - Triggered by the main workflow's tool node. - **Convert Conversation (Agent)** - Uses OpenAI to extract and format the input into a JSON structure: ```json { email: "jane.doe@example.com", description: "Wants help automating lead intake and sending Slack notifications." } ``` **Structured Output Parser** - Ensures the extracted data matches the expected JSON schema. **Append row in sheet (Google Sheets)** - Adds the new lead data to your CRM sheet. **Code Node** - Returns a simple text confirmation like `Thanks for the info, we will be in touch soon`. ### Required setup for Google Sheets: - Enable the Google Sheets API and connect your Google account in n8n. - Create a sheet with at least the columns `email` and `description`. - Use the sheet's Document ID and tab name in the Google Sheets node. --- ## 2. Create the Main Workflow (Website Chatbot) This workflow acts as the main AI Agent handling incoming chat messages. ### Steps in the main workflow: 1. **When chat message received** - Starts the workflow whenever a visitor sends a message via your chatbot integration. 2. **Website Chatbot (Agent Node)** - Configured with a **System Message** that: - Briefly explains your services. - Asks the visitor what processes they want to automate. - Requests their name and email. - Sends collected data to the CRM tool once email and description are available. 3. **OpenAI Chat Model** - Connects to the AI agent as its language model. 4. **Simple Memory** - Stores short-term context for the ongoing chat. 5. **CRM Tool (Tool Workflow Node)** - Points to the sub-workflow created in Step 1, allowing the chatbot to trigger it directly. --- ## 3. Connecting the Sub-Workflow to the AI Agent 1. Add a **Tool Workflow** node to the main workflow. 2. Select `Parameter` as the source. 3. Paste in your sub-workflow JSON or select it from your n8n workflows. 4. Connect the **Tool Workflow** node to your AI Agent using the `ai_tool` connection. 5. Give the tool a clear description (e.g., `CRM tool to store lead information`) so the agent knows when to use it. --- ## 4. How It Works in Action 1. A visitor sends a message through the chatbot. 2. The AI Agent engages, asks questions, and collects their name, email, and request. 3. Once collected, the agent triggers the **CRM Tool**. 4. The sub-workflow formats the data, stores it in Google Sheets, and sends a confirmation. 5. The chatbot confirms with the visitor that their request was received. --- ## 5. Customization Ideas - Replace Google Sheets with your actual CRM API. - Add validation to ensure the email format is correct before saving. - Expand the CRM tool to send a Slack or email notification after storing the lead. --- **Created by Robert A. - Ynteractive** Website: https://ynteractive.com Email: robert@ynteractive.com