Humanice

Creating Engaging Conversational AI to Eliminate Loneliness in GenZ Users

Creating Engaging Conversational AI to Eliminate Loneliness in GenZ Users

ROLE

Product Design Intern

TEAM

1 Product Designer (Mentor)
1 Product Manager


6 Developers
the CEO & Me

TIMELINE

3 months (May-Aug 2023)

SKILLS

Interaction Design


Product Thinking
Visual Design
User Research

Overview

Over the summer, I worked as a product design intern at Warp Engine (later Viggle AI), a Gen AI start-up specializing in digital human technology (aka digital avatars that look like humans) and LLM, to develop "Humanice," a conversational AI app on IOS mobile, in preparation for its internal launch.

*example digital human tech from web

Impact

+40%

increase in customer satisfaction score

+~250%

increase in registered users

+25%

increase in user session time

What was the problem?

It started with a business goal:

The team aims to adapt the company's digital human AI from B2B to B2C, reaching a broader audience.

*example digital human tech from the web

Thus, the team identified an opportunity:

Gen Z's growing loneliness and desire for constant, readily available companionship.

*A survey conducted for Cigna U.S. Loneliness Index suggests that 79% of GenZ participant reported to be lonely.

As the new intern, I asked myself the HMW question:

The Solution

An AI chatbot app that's basically your AI friends 🧑‍🤝‍🧑…

Customized Onboarding

Users will experience a 3-stage onboarding process, customizing desired language and trending preferences.

Chat with AI friends

Users can communicate freely, anytime, anywhere. A companion always prepared to provide a listening ear at a keypress distance.

Share the joy

Users can join communities with similar interests and share fun clips of their conversations with friends and family outside the app.

Users can select memorable AI conversations, export them as video clips, and share them with others.

Users can join communities based on their preferences and connect with others who share similar interests.

Initial scope from the CEO and PM:

BUT how could we differentiate ourselves? I looked at 4+ competitors:

BUT how could we differentiate ourselves?

To acknowledge, the space of AI chatbot companionship is saturated... I conducted a competitive analysis of related AI chat applications popular among the GenZ community.
💡 I observed that competitors offer limited customization for building personal connections, which presents an opportunity for us to gain a competitive edge.
I also realized our initial scope was TOO BROAD—it’s nearly impossible to build and customize a product for everyone, especially in its early stages.

"If you design for everyone, you design for no one"

Thus, I took the initiative to gather insights from user interviews, redefining pain points and personas.🖊️

I recruited participants that fit within these criteria from multiple platforms:
  • GenZ
  • Tech-savvy, always online
  • Keen on cutting-edge technology, and might be interested in interacting with an AI chatbot.

Users are complaining. Here are some of their pain points:

01

😡👎

Users dislike mechanic responses and want more organic, human-like conversations with the AI.

02

😵‍💫❓

Users found the interface and navigation of current AI chat systems to be confusing and non-intuitive.

03

🥱💤

Users lack of motivation to talk to a random “person” irrelevant to their interests.

Gen Z, but maybe a more specific persona?

After user interviewing, I realized how Gen Z users can be different despite being placed in the same age group. Thus, the team shifted the user persona to a segment of Gen Z, the Entertainment Fanatics.
Entertainment Fanatics
n. Someone who has an extreme level of enthusiasm for an entertainer or other entertainment-related subject.

There are people who are interested in talking to:
1)Characters in the fantasy world 🧚
2)People who share the niche knowledge🧑‍🎓

🎬 🎮 🍿️ 🤩

🎬 🎮 🍿️ 🤩

Design goals

Based on the research result, I established 3 design goals that aid the design process.

Foster Comfortable Conversations
Enhance Usability and Intuitive Navigation
Provide Personalized Experiences

Realigning with stakeholders and exchange ideas 👭👬

I facilitated a 7 people brainstorming workshop with my mentor designer, involving the PM and developers.

PHASE 1

Building the Fundamental Features for MVP 🧱

Collaborating with the PM, I prioritized the three most important features for the MVP user flow based on value and effort:
1. Onboarding (mid-value, low-effort)
2. Chat with AI (high-value, high-effort)
3. Character Selection (mid-value, mid-effort)
STEP 1

Onboarding

STEP 2

Chat with AI

STEP 3

Select Character

EXPLORATION #1

Mimicking Human-like Conversations

To enhance user comfort with AI chat, I experimented with seamlessly integrating digital human technology into familiar contexts, such as text, audio, and video chat.
EXPLORATION #2

The Magical (Audio) Button

After settling on an audio chat format, I discussed with my mentor how users can interact with a simple audio button and receive feedback. I experimented with some simple motion designs using After Effects.

Connecting/ Waiting for Response

Loading Spin

Connected Ready to Speak

Button background not rotated

User Speaking

Button background rotation, sound wave appears

STEP 1

Onboarding

STEP 2

Chat with AI

STEP 3

Select Character

EXPLORATION #3

Flexibility in Character Options for the Users

Character selection is key to providing a customized experience for users. I tried out different layouts to display these characters and landed on the first solution.

Contextual Observation with Potential Users 🔭

To validate our design, the team conducted 30+ contextual observations and user testings at Bilibili World and ChinaJoy, two of China's most renowned digital entertainment expos.

PHASE 2

Adding New Features to Encourage Social Interaction and Engagement 🧑🤝🧑

In Phase 2, our goal is to address urgent technical issues, enhance the social features, and improve the feedback loop as we continue to rapidly grow as a Gen AI product, according to the user testing results.

Technical🔩

46% of the users fail to unsend voice message during their chat

Social🧑‍🤝‍🧑

83% of the users expressed interest in connecting with other users with similar interest

Feedback📃

93% proposed character and community ideas and hope to be added to the app
01 Technical

Adding signifiers to guide the unsend voice message

Users are now guided to slide up and release when holding the audio button and are provided with timely feedback once the unsend action is completed.

User Pain Point
Users are unsure about the action to unsend the voice messages 🫤
02 Social

Discovering Fun Communities

Users can join circles aligned with their interests and engage in discussions with like-minded individuals.

User Pain Point
Users get bored easily merely talking to the AIs 🥱
02 Social

Sharing Your Joy to Others

Enabling users to select and share chat history in video format, either within the in-app community or through other applications, enhancing user engagement.

03 Feedback

Improving Feedback Suggestion

Users can provide feedback through the settings, whether they want to request a new character or feature, or report any issues. Future improvements will be made based on this feedback.

And then... it was INTERNAL LAUNCH 🚀

Humanice was tested by over 200 users on TestFlight and is about to launch on the App Store.

Impact

As indicated by analytics

As indicated by analytics

As indicated by analytics

As indicated by analytics

The platform experienced an exponential growth of 250% in registered users.

According to post-launch surveys

According to post-launch surveys

According to post-launch surveys

According to post-launch surveys

Customer satisfaction scores improved by ~40%, reflecting increased user confidence and delight in the product

As per usage data

As per usage data

As per usage data

As per usage data

User session time increased significantly, showcasing improved usability and sustained user interest.

First internship, my takeaways:

01

Navigating ambiguity by breaking it down into manageable parts.

02

One effective way to gather feedback is by directly engaging with users.

Thanks for visiting :)

Feel free to reach out at yl3262@cornell.edu!


I'm more than happy to chat about design, art, and my cats 🐱🐱🐱

PROJECTS

Gen AI Start-up Internship Project
Wiley Internship Project
Queue Me In UX project
Duolingo Redesign Case Study

Copyright © 2024 Happy Li

Thanks for visiting :)

Feel free to reach out at yl3262@cornell.edu!


I'm more than happy to chat about design, art, and my cats 🐱🐱🐱

PROJECTS

Gen AI Start-up Internship Project
Wiley Internship Project
Queue Me In UX project
Duolingo Redesign Case Study

Copyright © 2024 Happy Li

Thanks for visiting :)

Feel free to reach out at yl3262@cornell.edu!


I'm more than happy to chat about design, art, and my cats 🐱🐱🐱

PROJECTS

Gen AI Start-up Internship Project
Wiley Internship Project
Queue Me In UX project
Duolingo Redesign Case Study

Copyright © 2024 Happy Li

Thanks for visiting :)

Feel free to reach out at yl3262@cornell.edu!


I'm more than happy to chat about design, art, and my cats 🐱🐱🐱

PROJECTS

Gen AI Start-up Internship Project
Wiley Internship Project
Queue Me In UX project
Duolingo Redesign Case Study

Copyright © 2024 Happy Li