Executive Summary

This business case outlines the implementation of an AI-driven recommender system designed to enhance customer support and operational efficiency for a mid-sized client specializing in business transformation through cloud technologies. By optimizing their all-in-one customer service tool, the client aims to provide seamless support across multiple marketplaces while improving customer satisfaction.

Client Overview

The client focuses on delivering tailored solutions to businesses in APAC and Southeast Asia, enabling them to enhance their competitiveness. Their customer service tool integrates with platforms like Shopee, Lazada, and TikTok, in addition to utilizing Zendesk for managing customer interactions.

Challenges

  1. Resource Quality and Response Time: The client faced issues with inadequate resources, leading to delays in addressing customer queries and negatively impacting customer satisfaction.
  2. Complex Customer Interactions: Managing interactions across multiple platforms created complexity, as each marketplace had its own communication tools and service protocols, making consistent support difficult.

Objectives

  • Build a Recommender System: Develop a model to provide personalized product recommendations based on shop data, integrated with Zendesk for various marketplaces.
  • Integrate Marketplace Communication with Zendesk: Create a centralized solution to manage customer tickets from different shops, enabling comprehensive support and personalized recommendations.

Proposed Solution

  1. Marketplace Integration:
    • Custom Python Applications: Developed for each marketplace (Shopee, Shopify, Lazada) to synchronize shop and seller data with a MongoDB database.
    • Recommender System: Built using Python and TensorFlow to train models on synchronized shop data for personalized product recommendations.
    • REST API Development: Created a REST API using Python and FastAPI for interaction with the recommender app.
  2. Data Management:
    • CRUD Operations API: Developed using JavaScript and NestJS to handle data operations efficiently.
  3. Amazon Chat Integration:
    • Email-Based Solution: Created a custom solution for Amazon using email messages, set up with Dovecot and Postfix.
    • Mail Server Interaction API: Developed using Python and FastAPI for reading and replying to email messages.

Expected Outcomes

  • Enhanced Customer Support: By integrating the recommender system with Zendesk, the client can provide personalized product suggestions, improving customer engagement and satisfaction.
  • Operational Efficiency: The centralized management of customer tickets across multiple marketplaces will streamline workflows and reduce response times.
  • Increased Customer Satisfaction: Personalized recommendations and efficient support will enhance the overall customer experience, potentially leading to higher retention rates.

Conclusion

The implementation of the AI-driven recommender system positions the client to overcome current challenges and achieve their business objectives. By leveraging AI, data, and analytics, the client will enhance their operational efficiency, provide superior customer support, and ultimately drive business growth across multiple marketplaces.