Withdraw Chatbot

Project Overview

  • User groups reflected desire for more digital capabilties such as chatbots for common
  • Wanted to explore chatbot capabilities within global wealth asset management

My Contribution

  • Product manager managing the initiative, working 2 engineers and 1 designer and 1 analyst

Discovery

  • Performed secondary research on chatbot usages and implementation guidelines through academic papers and competitor analysis
  • Worked closely with product and operations teams to explore areas that are suitable for chatbot
  • Job shadowed call center representatives to understand existing pain points
  • Technical proof of concept to explore various frameworks and capabilities

Pain Points

  • Participants can only find out the status of their withdraw requests by phoning the call center
  • Wait time is over 20 minutes to get through to a call representative, and they are only available Monday to Friday during work hours
  • Need guidance to understand the different types of withdraws, but the website is difficult to navigate
  • Anticipate high number of tax inquiries at the end of the year

Ideation

  • Given there was no direct access to participants, the team worked closely with call center representatives who frequently interacted with participants to design the dialog flow through a series of ideation sessions
  • Created low fidelity prototypes for Q&A vs. open-text chatbot, created scenarios and posted on usertesting.com for feedback. Result showed that users preferred the structured approach of Q&A chabots given its limited scope
  • Worked closely with product teams to design and integrate the chatbot into the existing website

Pilot

  • Limited scope with withdraw inquiries and basic tax questions
  • Question-and-answer based implementation instead of open dialogue
  • Survey for user feedback on experience and areas of improvement
  • Tracking statistics on usage and areas of breakdown

Outcome

  • Over 1000 participants have accessed the chatbot, with 70% repeated users
  • Demographics spread across different age groups and geographical locations
  • Mixed feedback, users wanted more transactional capabilities (e.g. taking out a loan through the chatbot)
  • Analyzed usage statistics, identified and prioritized areas for continuous development

Technology Stack

  • Microsoft Bot Framework
  • NodeJS
  • React
  • PowerBI