Building a Customer Support Agent in Nan

Introduction

• Demonstration of building a customer support agent in Nan.
• Automation pulls in tickets from tools like Zenesk, Gorgeous, or Freshesk.
• Retrieves all ticket information and comments.
• Two main functionalities:
  • Generate responses for open tickets.
  • Create FAQs from solved tickets.

Workflow Overview

• Open tickets are processed to generate responses via a response agent, sending replies back to CRM (e.g., Zenesk).
• Suggested starting with internal notes for safety.
• Can later automate direct responses to customers via SMS or email.
• Logs all responses in a Google Sheet for easy overview.

Creating Responses

• Automation sends customer inquiries to Nan agent which generates a response.
• Responses can initially be internal notes for review.
• Eventually automate live replies to customers.

Logging in Google Sheets

• Every interaction logged in Google Sheet.
• Logs include date, ticket ID, subject, body, and AI-generated response.
• Easy overview of responses for multiple tickets.

FAQ Generation from Solved Tickets

• Utilizes solved tickets to generate FAQs.
• Helps build a knowledge base by extracting information from human agents.
• FAQs stored in a separate tab in Google Sheet.

Technical Setup

Webhook and Trigger Configuration

• Requires a webhook to connect Zenesk and Nan agent.
• Triggers in Zenesk to send ticket information upon creation or comment update.
• Webhooks notify Nan agent to process ticket information.

API Calls and Authentication

• Use Zenesk API to retrieve ticket information with two calls:
  • Basic ticket details.
  • All comments on the ticket.
• Authentication with Zenesk required for API calls.

Ticket Processing

• Extract ticket comments filtering only public comments.
• Distinguish between customer and staff responses using author IDs.
• Flatten ticket information into markdown for AI processing.

AI Agent and Response Generation

• AI agent generates JSON output formatted for Zenesk API.
• Creates email responses with structured format using AI.

Google Sheets Integration

• Logs AI-generated responses for review.
• Possible to evaluate response accuracy and check for errors.

Creating FAQs from Solved Tickets

• AI agent tasked to generate a succinct FAQ from solved ticket conversations.
• FAQs logged into Google Sheets for easy access and updating of support databases.

Enhancements and Customization

• Ability to customize the workflow for different CRM tools by adjusting API calls.
• Use Perplexity API for web search restricted to specific domains.
• Maintain support articles on the website instead of a separate database.

Conclusion

• Workflow setup is beginner-friendly and can be operational quickly.
• Potential for further customization and enhancement.
• Apply workflow principles across different CRM tools.

Final Steps

• Activate the workflow by configuring webhooks and triggers in Zenesk.
• Ensure workflow is set to active mode before proceeding.

Call to Action

• Encourage viewers to like the video and share their usage and modifications in comments.