Bankwest New 'Bee' Chatbot for web and in-app

Bankwest Case Study: Revamping the Bee Bot

Project Overview

The previous Bankwest bot faced several challenges: poor containment, weak Natural Language Understanding (NLU), a lean scope, and a lack of persona. To address these issues, I designed a new Bee bot for both web and in-app platforms. The revamped bot comprised 48 guided flows with disambiguation, an extensive knowledge base with NLU-trained knowledge articles, and a total of 138 intents.

Design Process

The design process began with on-site workshop whiteboard sessions with the Bankwest team in Perth. These sessions were crucial for understanding the team's needs and brainstorming solutions. Using the insights gathered, I created wireframes to design guided flows in Lucidchart. I then rewrote all the content to give Bee a distinct persona, aiming to reduce abandonment by making the experience far more engaging.

Implementation and Results

The result was two sophisticated bots: one for authenticated internet banking and another for the authenticated Bankwest app. This comprehensive redesign led to a significant increase in containment, which jumped from 7% to 40%. Although a project of this scale typically takes 12 months to deliver, I completed it in just 7 weeks. Additionally, I managed the development, tuning, and QA teams to ensure the project's success.

Key Achievements

• Designed and implemented 48 guided flows with disambiguation.

• Developed an extensive knowledge base with 138 intents.

• Enhanced the bot's persona to create a more engaging user experience.

• Increased containment from 7% to 40%.

• Delivered the project in 7 weeks, well ahead of the usual 12-month timeline.

• Managed the development, tuning, and QA teams to ensure seamless execution.

• This case study highlights the successful overhaul of the Bankwest bot, demonstrating the impact of effective conversation design and project management.