Customer Inquiry Analysis and Response Generation Automation: AI Customer Center

A system that builds a chatbot-style customer center to efficiently handle FAQs and option/chat-type inquiries.

Background

• The client operating a large shopping mall had at least 7 customer service agents directly responding to various customer inquiries daily. On high-volume days or when complex inquiries extended consultation time, customer satisfaction was significantly declining.
• Although FAQs and inquiry boards were utilized, users preferred chat-based consultations. However, the existing KakaoTalk channel-based option selection customer center had relatively low customer satisfaction.
• The client was continuously expanding the shopping mall, meaning customer inquiries would increase further, requiring an effective solution.
• Hiring additional customer service staff for temporarily increasing inquiries was burdensome, and utilizing existing staff for other tasks would be more effective.

Project Description

1. Past Customer Inquiry Analysis and Key Scenario Identification
   • Categorized FAQs and situations requiring/not requiring chat
   • Organized and classified newly needed response situations
2. User Interface Design (Including Design)
   • Another key aspect of AI chatbot projects is "efficiently exchanging Input/Output within a defined chat space" - performed scenario analysis and design work for this
3. Backend System Design and Development
   • Determined which generative AI solution to use and when/how to call APIs
   • Comprehensive design and implementation of how to connect with shopping mall member/product databases
4. Question-Answer Database Construction
   • Built Q&A DB for FAQs, option/chat-type inquiries classified earlier
   • Determined and executed cost-efficient service operations
5. Real Agent and AI Chatbot Collaboration Scenario Design and Implementation
   • Even with AI chatbots, real agent intervention is needed in certain exception situations
   • Implemented features for real agents to collaborate with AI chatbot for quality management of early responses
6. Continuous Learning Updates
   • Performed ongoing Prompt Engineering and Fine Tuning for response optimization

Key Achievements

• Customer inquiries that were only handled during agent working hours (Mon-Fri, 10AM-5PM) can now be processed regardless of day/time, greatly improving user satisfaction
• With only a few given options like "Order inquiry," "Shipping policy," "Payment methods," the probability of customers finding desired answers was very low. The chatbot-style customer center achieved 80-90% "first response" completion rate
  • (10-20% of inquiries were directly handled by agents while strengthening chatbot capabilities)
• Handled more inquiries with only 20-30% of the human resources previously allocated for customer service
• Continuously updated Vector DB (RAG) after commercialization to improve response rates

Development Process

1. Step 1: Service Requirements Definition
   • Established detailed customer center requirements (including AI response and agent collaboration features)
   • Collected sales policies, shipping policies, FAQs
   • Organized and cleaned existing inquiry/response records
2. Step 2: User Scenario Confirmation
   • Confirmed chatbot response scenarios (flow)
   • Basic Prompt Engineering design
   • Confirmed chatbot response feedback verification and human agent collaboration methods
3. Step 3: User Interface and Web UX Design
   • Reviewed and adjusted service requirements and user scenarios
   • Determined User Interface capable of handling scenarios
   • Designed Web Application features
4. Step 4: Architecture Design, AI Solution Integration
   • Designed backend system and database configuration
   • Built RAG system using VectorDB (Weaviate)
   • Integrated LLMs including OpenAI, Gemini
5. Step 5: Full Feature Development and Commercialization
   • Front-End development
   • Back-End development
   • Collected user feedback through commercialization
6. Step 6: Continuous Tuning (AI)
   • Prompt Engineering to improve response rates
   • Updates based on user feedback analysis

Our Strengths

• Development company with 10+ years of experience
• Data/algorithm experts (Seoul National University ECE Bachelor's/Master's graduates)
• Rich experience planning/designing/developing/commercializing "conversational chatbot" service with 1.2+ million downloads
• Proven capability in user analysis and excellent UX/UI, as demonstrated by Google Play Best Award (selected once per year by Google)
• Extensive project experience in automation/content generation using generative AI solutions