Dokidio: Conversational BOT with RAG / KG for Hyper Personalization

Summary

The client needed a highly personalized conversational bot to address inefficiencies in their existing system, which took over 20 seconds for each interaction. They sought a solution that was quick, empathetic, and able to adapt to users' emotional states while offering tailored advice and engaging conversations.

Introduction

The objective was to create a system that could handle interactions in under 2 seconds, ensuring users felt understood and that their needs were anticipated. Additionally, the solution needed to incorporate a Retrieval-Augmented Generation (RAG) system along with a Knowledge Graph (KG) to deliver intelligent and context-aware responses.

Discovery Questions

What are the common user needs or tasks that the bot should anticipate?
How many users interact with the bot simultaneously, and what is the expected scale?
What emotional or behavioral signals should the bot respond to in real-time?
What is the desired accuracy level for personalized recommendations?
What are the performance benchmarks for response times and user satisfaction?
What external systems need to be integrated for real-time data access and responses?

Challenge

The client was experiencing inefficiencies with their current conversational bot, which took more than 20 seconds to respond. This delay caused user frustration and led to drop-offs. Furthermore, the bot's lack of personalized responses made users feel disconnected and unengaged.

Solution

We created a highly efficient, hyper-personalized conversational AI using Retrieval-Augmented Generation (RAG) and Knowledge Graphs. The system offers:

1
Real-Time Personalization:
Utilizes Knowledge Graphs to store and retrieve individual user preferences.
2
Rapid Processing:
The bot now completes interactions in under 2 seconds, significantly improving on the previous system.
3
Empathetic Responses:
Generates AI-driven responses that are tailored to the user’s emotional state and personal interests.