Optimism ChatBot

Role PM / Developer
Period 2024
Type Internal Tool
smart_toy RAG Architecture & Chat UI

chat Project Overview

Developed an AI chatbot to automate employee inquiries at "Rakuten Optimism," the company's largest internal event. Leveraged Dify (open-source LLM app development platform) to integrate Azure OpenAI Service with internal knowledge base.

architecture Technical Challenges

Needed to immediately respond to event-specific "daily updated information" (preventing hallucinations).

Solutions:
Adopted RAG (Retrieval-Augmented Generation) architecture and built a pipeline that automatically ingests the latest manuals from SharePoint into Vector DB. Answer accuracy reached 81% in pre-release testing. Implemented escalation flow to human support for uncertain answers using Power Automate.

trending_down Business Impact

Resolved approximately 67% of total inquiries during the event period with Bot alone. Dramatically reduced burden on operations team and successfully concentrated resources on higher value-added tasks (VIP support, etc.).

67%

Inquiry Reduction

81%

Answer Accuracy

Tech Stack

Similarity Calc Embeddings DeepQA Fine-tuning