DisneyNow

Baymax Chatbot

The Baymax Chatbot project was an enhancement feature for the Disney Now App - A cross-platform tool for kids to enjoy the latest Disney TV episodes and play games. The goal was to leverage Disney IP(Baymax - a helpful companion) for testing R&D discovery work for a machine learning initiative.


my role

Interaction Designer

Responsibilities: Discovery, Research, UX Wires, Visual Design, Motion Design

team

1 Interaction Designer, 1 Front-end Developer, 1 R&DEngineer, 1 Product Manager, 1 VP of Design(Stakeholder)

Q1-2018 - Jan-Mar

TOOLS

Sketch, After Effects, Zeplin

PLATFORM

iOS


01

What if Baymax could be a real companion?

Background: Because of my enthusiasm to work on different types of projects, the VP of Design tasked me with creating a prototype concept for a Disney R&D AI initiative. What if we could capitalize on Disney IP and test the concept out on one of our platforms for children - DisneyNOW? After some initial set up conversations with cross-functional teams, we settled on Baymax as the perfect character to be a friendly companion and useful chatbot tool.

Challenge:

How can we serve up promotional content to users in a personalized and more interactive way?

 

Short-Term Goals:

  1. Provide more avenues for a younger audience to search for relevant content efficiently.

  2. Interactive, fun and more gamified experience for attention grabbing

  3. Business incentive to have users stay in the platform and gather qualitative insights for what content to serve up via AI machine learning

Long-Term Goals:

  • Content promotion by building a conversational agent


02

Conversational Flow

One of the first exercises I started out with was empathizing with the user. The average user age on our app ranged from 5-10 years old or the other user type would be their parents putting content on for their little one. Understanding that we must have a simple way to communicate, I studied the character of Baymax (watching the movie and listening through audio clips) in order to combine his friendly tone with simple text

One exercise I created was an artifact that we could refer to. I created  a conversational flow and synthesized how different key words could help trigger the stored content, using different types of phrasing based on the parameters of the project.


03

Iterations

The main entry points for the chatbot were based on the higher traffic areas for search: Search Home, Shows Detail Page, and where the search bar was tappable.

I started off with some sketches and initial wires, and then started creating multiple visual iterations to help showcase the blue sky concept approach for Baymax’s character.

Our team has weekly check-ins to go over feasibility and technology constraints, so my designs morphed from an ideal state into a more reasonable MVP. Initially we had both voice and typing triggers, but we ended up cutting it to just the typing triggers because of technical constraints and high user error. The success rate of kids being able to formulate sentences well enough with their voices for a request was pretty low.


04

Design States

After presenting multiple iterations in front of the design team and Research & Development team, we were able to focus on one version for a development prototype. Throughout the process, I was working directly with the front-end developer and having conversations about content specs, while the design of the overarching chatbot behavior was under review.

Below is the developed prototype for multiple conversational states based on different content request patterns. This allows the user to get the same results with different wording. This would happen by data being stored from the users account and behavior on the app to help with suggesting the content relevant to them.


User Actions:

  • User initiates DisneyNow App

  • User goes to search area

  • Baymax Chatbot pops up

  • User requests “Please help me”

  • Baymax provides 1st results


User Actions:

  • User exclaims they are satisfied with their content

  • User asks bot to pinpoint a specific episode


User Actions:

  • User asks Baymax to recommend them shows

  • User swipes through recommended content carousel

  • User asks Baymax specifically for comedy shows

  • Baymax send through a comedy focused carousel


User Actions:

  • User initiates Baymax from idle state

  • User reminds Baymax what time of day it is

  • User says they’re satisfied with their content

  • Baymax ends the chat


05

Moments of Delight

To include the goal of making this experience more interactive for a younger audience, I took the visualization of the feature farther by creating motion studies for Baymax’s facial expressions. This was a concept mockup utilizing emotion to convey the different states of Baymax’s help, and also combining audio clips to provide more story-telling of his character.

Searching State

Low Battery

Idle State

Validation


06

Challenges

  1. The project didn’t have any time constraints and we worked off of the availability of the R&D department, so solutioning took a little bit of time.

  2. Not a lot of dedicated resources

  3. Full scope of capabilities was being learned in the moment, but that was the fun part!



07

Impact

This project has received development award accolades. Currently, the software is under a patent at Disney.