B2B2C Chatbot Projects

Overview

AI Chatbot projects for B2C Enterprise partners

Providing 24/7 customer service has become essential for B2C companies. Based on our experience in operating AI Assistants, we conducted several projects with partners in the finance and retail sectors. These projects involved implementing Kakao i Connect technology, an NLP-based AI chatbot, and customer service solutions.

Company

Role

  • Product Manager

  • Product/Conversation Designer

Deliverables

  • Product Policy

  • UX Design

  • API Architecture

  • Conversation Design

Project Date

  • Feb 2021 - Aug 2022
    6 Month Each Project

Context & Background

The customer wait time had surged, and demand for 24/7 connectivity with customers has increased.

During the pandemic, customer contact centers shut down, disrupting businesses and limiting customer engagement.

This led to a surge in demand for chatbots on KakaoTalk, South Korea’s leading messenger with 50 million users, and on partner company apps.

However, many partners faced challenges due to a lack of chatbot implementation expertise.

Hyphothesis : HMW Approach

“How Might We deliver AI chatbots optimized for clients and ensure sustainable operations?”

A chatbot isn’t a one-time deliverable. To optimize AI chatbots for clients, we held workshops to define the MVP, plan the roadmap, and outline key features.

Since chatbots require ongoing monitoring and learning, we focused on designing a sustainable operational system. Additionally, we developed tools to help clients manage and maintain their chatbots effectively.

User Research

Stakeholders Interviews

  • NH Investment Customer care team staff(Contact center agent)

  • NH Investment IT engineer

  • Korea Investment Customer care team staff(Contact center agent)

  • Korea Investment IT Engineer

  • Kakao i Connect Skill Builder Product Owner(On Premise Bot Data & Response Builder)

  • GS Retail IT/Infra Staffs

  • GS Retail customer care staffs

10,000 Voice of Customer(VOC) Data analysis

  • Analyzed and preprocessed an average of 10,000 VOC (Voice of Customer) data points

  • Designed conversation scenarios by defining top 30 identified user query types as intents to improve recognition

  • Integrated partial generative AI to handle open-ended user inputs beyond the 30 predefined intents, incorporating features such as small talk, web-based response generation, and a financial knowledge bot

  • Generated and trained 200,000 utterances for on-premise NLU, based on actual VOC input data

Outcomes

Kakao i Connect Chatbot

We built chatbots by defining requirements of partner company and understanding needs of their users

Welcome Response User Experience

When users first interact with the chatbot, they are often unsure of what they can do.
To address this, we designed a welcoming message and used card buttons to guide users for quick access to services, allowing them to easily understand the tasks they can perform within the chatbot.

Guide User Experience

Since customers tend to make ambiguous statements, we initially provide responses at a broad category level. These responses include detailed feature guidance and examples to encourage accurate usage by the user.

Knowledge search bot features using generative AI

During the pandemic, stock trading surged, bringing in many users with limited financial literacy and straining customer support.

To ease this burden, we designed a GPT-2-powered knowledge search bot that provides essential information on financial terms, IPOs, and more.

This feature helps beginner traders better understand financial transactions and promotes financial literacy.

Impact

Through 6 B2B2C projects, we achieved an average of 1 million utterance transactions in the first month of service launch, replacing an average of 20% of contact center inquiries. Additionally, these projects contributed $10 million, or 15% of the company’s revenue in 2021.

NEXT Projects

Workmate.AI - AI Work Assistant

UX Research & Strategy Project
at Royal college of Art