How to build a Webpage Customer Support Agent

This workflow is designed to answer customer questions by leveraging both public website data and internal company knowledge base files. Here’s how it works:

Nodes and Their Roles

1. Customer Question (Input Node)

   • Purpose: Entry point where a customer submits their question.

   • Details: This is a standard input node, labeled for customer queries.

2. URL (Files/Website Extraction Node)

   • Purpose: Extracts and processes content from the company’s public website.

   • Details: Configured to pull and chunk data from a specified URL (e.g., the company’s investor portal).

3. OneDrive Knowledge Base (Knowledge Base Node)

   • Purpose: Searches and retrieves relevant information from files stored in a OneDrive-based knowledge base.

   • Details: Uses semantic search to find the most relevant chunks of information.

4. OpenAI (LLM Node)

   • Purpose: Generates a response to the customer’s question.

   • Details:

     • Receives the customer’s question, website data, and knowledge base results as context.

     • Uses a system prompt tailored for customer support (polite, clarifying, and suggests follow-up questions).

     • Model: OpenAI GPT (customized for support scenarios).

5. Answer (Output Node)

   • Purpose: Presents the generated answer to the user.

   • Details: Receives the output from the LLM node.