AI-Powered Customer Feedback Analysis and Routing

This n8n workflow demonstrates a foundational setup for processing and analyzing customer feedback using AI, with a focus on structured output. It provides a starting point for more complex workflows that could integrate with services like Gmail, Zendesk, Slack, or Pipedrive for automated routing and action.

What it does

This workflow showcases the core components for an AI agent to process input and generate structured output:

1. Manual Trigger: Initiates the workflow upon manual execution. This is a placeholder for actual triggers like new emails, support tickets, or form submissions.
2. Edit Fields (Set): A placeholder node to simulate incoming customer feedback or to pre-process data before AI analysis.
3. Basic LLM Chain: Defines a chain of operations for the Language Model.
4. OpenAI Chat Model: Utilizes an OpenAI Chat model to perform the AI analysis. This is where the "intelligence" of the workflow resides, processing text and extracting information.
5. Simple Memory: Provides a basic memory buffer for the AI agent, allowing it to retain context within a conversation or a series of interactions.
6. Structured Output Parser: Ensures that the AI's output is formatted according to a predefined schema (e.g., JSON), making it easy for subsequent nodes to consume and act upon the extracted information.
7. AI Agent: Orchestrates the interaction between the LLM chain, memory, and output parser to perform a specific task, such as analyzing sentiment, categorizing feedback, or extracting key entities.

Prerequisites/Requirements

• n8n Instance: A running n8n instance (cloud or self-hosted).
• OpenAI API Key: An API key for OpenAI to use the OpenAI Chat Model.
• Basic understanding of LangChain concepts: Familiarity with LLM chains, agents, memory, and output parsers will be beneficial for extending this workflow.

Setup/Usage

1. Import the workflow: Copy the provided JSON and import it into your n8n instance.
2. Configure Credentials:
   • Locate the "OpenAI Chat Model" node.
   • Configure your OpenAI API credentials within this node.
3. Customize Input:
   • Modify the "Edit Fields (Set)" node to simulate the actual customer feedback you expect (e.g., email body, support ticket description).
4. Define AI Task:
   • Configure the "AI Agent" and "Basic LLM Chain" nodes with the specific prompts and instructions for analyzing your customer feedback (e.g., "Analyze the sentiment of this feedback," "Extract the customer's issue and suggested solution," "Categorize the feedback as bug, feature request, or general inquiry").
5. Define Output Structure:
   • Adjust the "Structured Output Parser" node to define the exact JSON schema you expect the AI to return (e.g., { "sentiment": "positive/negative/neutral", "category": "bug/feature", "summary": "..." }).
6. Extend the Workflow: After the "AI Agent" node, you can add further nodes to:
   • Route feedback: Use an IF node to route feedback based on category or sentiment to different services (e.g., Slack for urgent issues, Zendesk for support tickets, Pipedrive for sales leads).
   • Store data: Save analyzed feedback to Google Sheets, a database, or a CRM.
   • Send notifications: Post summaries to Slack or send internal emails.
7. Execute: Run the workflow manually using the "When clicking ‘Execute workflow’" trigger to test, or replace the manual trigger with an actual event-based trigger (e.g., a Gmail New Email trigger, a Zendesk New Ticket trigger).