Notion Feedback Processing with GPT-4 Sentiment & Insight Generation - n8n Workflow

Markdown formatted string required:

Product Managers & Analysts: Those who need to quickly process large volumes of user feedback stored in Notion.
Automation Enthusiasts: Users looking for advanced examples of integrating Langchain AI Agents within an n8n workflow.
Customer Success Teams: Professionals aiming to categorize feedback instantly based on sentiment (positive, negative, neutral).
Technical Users: Individuals seeking robust n8n templates for complex data processing and decision-making.

Handling user feedback manually is time-consuming and often leads to missed opportunities. This advanced n8n workflow solves this by instantly analyzing every piece of feedback added or updated in a Notion database. It leverages the power of OpenAI's GPT models via specialized n8n node tools to achieve two critical goals: first, rapid sentiment analysis to categorize the feedback; and second, invoking a sophisticated AI Agent to consult existing data and generate structured, new insights. This specific n8n workflow is a prime example of high-value AI automation, ensuring that product teams get actionable data the moment feedback arrives. By deploying this n8n template, you move beyond simple data storage to intelligent, proactive feedback management.

This powerful n8n workflow operates in two concurrent paths initiated by a single n8n trigger:

1. Trigger & Data Preparation: The process starts with the 'on database item update' n8n trigger, which monitors a specified Notion database for new or modified feedback items.

1. Path 1: Instant Sentiment Analysis: The raw feedback content is routed to the Langchain 'Sentiment Analysis1' n8n node. This node uses the 'OpenAI - Sentiment Analysis' LLM to quickly determine the overall sentiment (e.g., positive or negative). Finally, the result updates the original item in Notion using the 'update feedback sentiment analysis' n8n node, providing immediate categorization.

1. Path 2: Deep Insight Generation: Concurrently, the workflow prepares context for the AI Agent. It first uses a Notion n8n node ('Get database structure') and a Code n8n node ('Clean the database propertied') to understand the database schema and product context. This metadata, along with the feedback, is compiled into a detailed prompt using the 'Set Prompt' n8n node.

1. AI Agent Execution: The data flows into the 'AI Agent' n8n node. This agent is configured with 'Window Buffer Memory' and the 'OpenAI - AI Agent' LLM. Crucially, the agent has access to several Notion Tool n8n nodes ('Get all feedback', 'Get all insights', 'Create New Insight'). The agent decides if the feedback is similar to existing insights or if it warrants creating a brand new, structured insight using its tools. It utilizes an auto-fixing output parser to ensure the final output is correctly formatted.

1. Final Update: Whether the AI Agent generated a new insight or simply processed the data, the workflow concludes by executing the 'Update Feedback' n8n node on the original item, often marking the process as complete or adding any relevant AI-generated tags.

Markdown formatted string required:

1. Download & Import: Download the provided JSON code. In your n8n instance, navigate to 'Workflows' and click 'New'. Import the JSON file directly. This will load the complete n8n workflow structure.


2. Notion Credentials: You will need to set up Notion credentials for all Notion-related n8n node steps. Ensure the Notion integration has access permissions to the specific Feedback and Insights databases you plan to monitor and update.


3. OpenAI Credentials: Set up credentials for OpenAI/GPT-4 access. All Langchain Chat n8n node steps ('OpenAI - Sentiment Analysis', 'OpenAI - AI Agent', 'OpenAI - Parser fixing') require an active API key.


4. Configure Trigger: In the 'on database item update' n8n trigger, select the Notion database ID that holds your raw user feedback. The webhook will monitor this database for changes.


5. Database Mapping: Crucially, review the properties within the Notion n8n node steps (especially those retrieving or updating data) to ensure they correctly map to the columns (properties) in your Notion database (e.g., 'Feedback Text', 'Sentiment', 'Processed Status').


6. Activate: Save the n8n workflow and toggle it to 'Active' to start automated processing.

Markdown formatted string required:

on database item update (n8n trigger): The starting point of this automated flow. It is an n8n trigger specifically configured to watch for updates within a chosen Notion database, initiating the downstream AI processes.
Sentiment Analysis1 (Langchain Sentiment Analysis n8n node): Orchestrates the process of classifying the mood of the incoming text. It relies on the linked OpenAI LLM for natural language processing.
update feedback sentiment analysis (Notion n8n node): Updates the original Notion page item. Function: Pushes the result from the sentiment analysis back into a property field of the Notion database item.
AI Agent (Langchain Agent n8n node): This is the core intelligence hub. Function: It uses complex reasoning over a set of provided tools to decide the best action (consulting past data, generating a new insight). Key Configuration: It links to three Notion Tool n8n node inputs and uses Memory and Structured Output Parsers.
Get all feedback / Get all insights / Create New Insight (Notion Tool n8n node): These custom tools provide the AI Agent with the ability to interact dynamically with the Notion database, allowing it to search for relevant existing data and write back structured results.
Clean the database propertied (Code n8n node): A custom n8n node used to transform or clean the metadata received from Notion's schema before it is passed to the AI Agent for context setting.

• Auto-fixing Output Parser (Langchain n8n node): Ensures that even if the AI Agent returns a slightly malformed JSON object, the system attempts to fix it using an auxiliary LLM ('OpenAI - Parser fixing'), guaranteeing reliable output for subsequent n8n node operations.