AI Trustpilot Review Analyzer & Insights Agent

Unlock Deep Customer Understanding with this AI Agent

This powerful AI Agent automates the entire process of gathering, analyzing, and deriving insights from your Trustpilot reviews. It's designed as a two-part system: first, it scrapes reviews for your specified company, processes them using OpenAI's text-embedding-3-small model to understand their semantic meaning, and stores these vectorized reviews in a Qdrant database.

Once data ingestion is complete, the second part of the agent is triggered. This analysis phase employs a K-means clustering algorithm (via a Python node) to group similar reviews, identifying recurring themes and topics. For each significant cluster, it then leverages an OpenAI LLM (gpt-4o-mini) through the Langchain Information Extractor node to generate concise insights, determine overall sentiment, and suggest actionable improvements. All this valuable data, including raw review snippets, is then neatly organized and exported to a Google Sheet, empowering you to act on customer feedback swiftly.

Key Features & Benefits

• Automated Review Acquisition: Continuously scrapes the latest reviews from Trustpilot for a specified company, ensuring your insights are always based on fresh data.
• AI-Powered Semantic Analysis: Utilizes OpenAI's text-embedding-3-small to create vector embeddings of reviews, capturing nuanced meaning beyond simple keywords.
• Efficient Data Management: Stores and organizes review embeddings and metadata in a Qdrant vector database, optimized for similarity search and advanced filtering.
• Intelligent Theme Discovery: Employs a K-means clustering algorithm (Python) to automatically group reviews by common topics and sentiment, revealing underlying patterns in customer feedback.
• Actionable AI Insights: Leverages OpenAI's gpt-4o-mini to analyze review clusters, generating clear summaries, overall sentiment (strong negative to strong positive), and practical suggestions for improvement.
• Streamlined Reporting: Consolidates all findings, including AI-generated insights and raw review data, into a Google Sheet for easy access, sharing, and tracking.

• An n8n instance (Cloud or self-hosted) with Python execution capabilities (scikit-learn and numpy packages required for the clustering node).
• OpenAI API Key with access to text-embedding-3-small and gpt-4o-mini (or a compatible chat model).
• Qdrant instance accessible by n8n (e.g., http://qdrant:6333). Ensure the collection 'trustpilot_reviews' is created or can be created by the workflow.
• Google Sheets API credentials configured in n8n.
• The target company's Trustpilot page must be publicly accessible for scraping (e.g., www.yourcompany.com).

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Initial Configuration: a. In the 'Set Variables' node (ID: f0ea6b63-c96d-4b3f-8a21-d0f2dbb4efc3), update the companyId value with the domain of the company whose Trustpilot reviews you want to analyze (e.g., www.yourcompany.com). This company ID is used to construct the Trustpilot URL.
4. Credential Setup: a. OpenAI: Configure your OpenAI API credentials in the 'Embeddings OpenAI' node (ID: e22d92b8-e8e9-42aa-9d02-2e70234f11ed) and the 'OpenAI Chat Model' node (ID: f21369b9-1dd5-4b35-a1f3-00fd67794051). b. Qdrant: Set up your Qdrant API credentials for the nodes: 'Clear Existing Reviews', 'Qdrant Vector Store', 'Find Reviews', and 'Get Payload of Points'. Ensure the URLs in these nodes correctly point to your Qdrant instance (default is http://qdrant:6333). The collection name used is trustpilot_reviews. c. Google Sheets: Configure your Google Sheets API credentials in the 'Export To Sheets' node (ID: d77daa23-6acf-4daa-bf4c-33da4d05a54c). Update the documentId to your target Google Sheet and sheetName as needed.
5. Python Environment: The 'Apply K-means Clustering Algorithm' node executes Python code requiring scikit-learn and numpy. Ensure your n8n instance can run Python and these libraries are installed or can be installed. The first execution might take longer if packages need to be downloaded.
6. Scraping Scope: The 'Get TrustPilot Page' node is set to scrape a maximum of 3 pages (maxRequests: 3). Adjust this if you need more comprehensive data.
7. Workflow Execution: The workflow is designed in two stages. The manual trigger initiates data scraping and storage. The 'Trigger Insights' node (ID: 61c3117c-757c-45dd-b9d5-1122b793be30) then calls the analysis part of the workflow using an 'Execute Workflow Trigger' node (ID: 64af9cc7-a194-4427-ba78-d9a1136b962f).
8. Activate the workflow. Monitor the execution, especially the first run for Python package installations.