Humanice

Background

Over the summer, I interned as a product designer at Warp Engine, a Gen AI start-up specializing in digital human technology (aka digital avatars that look like humans) and LLM. I joined a cross-functional team 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 user base

+26%

increase in engagement time

Problem Statement

User Focus

Generation Z is increasingly experiencing loneliness and needs companionship that is readily available and provides a sense of constant connection.

Business Focus

The company encountered difficulties transforming its digital human AI technology in the B2B market to a B2C product, in order to reach a bigger scope of targeted GenZ users.

What did I encounter as the new intern?

When I joined, the team had gathered foundational user research for the AI assistant app, targeting Gen-Z users who are more likely to embrace and adapt to new AI technology. The design team was nearing completion of the mid-fidelity MVP and was getting ready for user testing.

What do GenZ users look for in GenAI chatbots?

I conducted a competitive analysis of related AI chat applications popular among GenZ, such as Character AI, Pi AI, Call Annie, and ChatGPT.

This helped pinpoint potential opportunities where we could learn from and potentially outperform them.

What did they do well?

✅ Informative
✅ Responsive
✅ Use of diverse LLM models

What can be improved?

❌ Lack of entertainment
❌ Text-heavy
❌ Not human-like

Collecting Insights from User Testing 🖊️

Our team conducted 10+ user testing with potential users including friends and volunteers from platforms like Reddit, Discord, etc.
Here are 3 key insights derived from the testing, showcasing the pain points to be solved.

Users felt uncomfortable interacting with AI, due to the uncanny valley effect and unfamiliarity.

😨

Users found it challenging to navigate how to communicate with the AI and use other features.

😖

Users think its a more informative application than one that’s engaging and entertaining.

🥱

PERSONA

Gen Z, but maybe more specific?

After user testing, 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.

How Might We...

Create Sense of Trust towards AI

Ensure that interactions with the AI feel comfortable and familiar to users, avoiding any unsettling experiences.

Enhance Intuitive Navigation & Usability

Allow users to access and make use of all features easily and maintain a natural, easy to use conversation that mirrors typical human interactions.

Bring Customized Entertainment

Allow users to choose their favorite character and language to personalize and enhance their interaction experience.

Brainstorm Workshop with Stakeholders to Exchange Ideas 👭👬

Based on the 3 HMW questions, I facilitated a 7 people brainstorming workshop with my mentor designer, involving the PM and developers.

Prioritize High Value, Low Effort Features 💰⏰

Considering the tight deadline and limited budget for the MVP stage, a 2x2 matrix was used to evaluate and prioritize features together with the team.

Information Architecture as the Foundation and Roadmap Reference 🗺️

After discussing with the PM, features with high value and low effort were added on to the Information Architecture. According to the product roadmap, some features in purple are planned for phase 1, and others in pink are postponed to phase 2.

PHASE 1

Building the Fundamental Features for MVP 🧱

Derived from IA, this is what the general user flow looks like. There will be an extra onboarding step for first time users.

Creating Wireframes for the MVP Framework

Here is a glance to the wireframe I created referencing the mvp user flow. It acts as a grounded structure for the following explorations and iterations.

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.

Audio chat appears to be a good choice for its high technical feasibility and good use of the technology.

A. Text Chat

B. Audio Chat

C. Video Call

A. Friend Contact List

✅ Pattern provides familiarity; Displays information clearly

❌ too standardized

B. Friend Card

✅ Puts emphasis on photo; Gamified and intriguing layout

❌ Information hard to read

C. Friend Swipe

✅ Focus on the first character

❌ Occupies majority of the space; Blocks other character displays

EXPLORATION #2

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.

EXPLORATION #3

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

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.

Here are the insights we gathered:

REFINE #1

Unsend Voice Message

When the user holds the audio button, guide them to slide up and release to unsend the current voice message. Provide timely feedback once the action is completed.

REFINE #2&3

Add Questions List

The Questions section was added to the bottom right of the screen. Users can use these customized prompt questions to ask the AI character and start a fun conversation. After confirming through user testing that the Language section had a low click-through rate, I moved it to the settings page.

And then... it was Dev Hand-off and INTERNAL LAUNCH 🚀

I collaborated with cross-functional teams, working closely with engineers to perfect every UI detail, ensuring precise padding and spacing.

The app was tested by over 200 users on TestFlight and is about to launch on the App Store. According to a questionnaire, user satisfaction rate increased by around 40%.

PHASE 2 (EXTERNAL LAUNCH PREP)

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

According to the feedback collected from the internal launch, many participants expressed a desire to connect with others who share similar interests. Therefore, in Phase 2, our goal is to enhance the social features and improve the feedback loop as we continue to rapidly grow as a Gen AI product.

Discovering Fun Communities

By integrating a community feature into the app, users can join circles aligned with their interests and engage in discussions with like-minded individuals.

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.

Improve 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.

While NDA prevent me from sharing specific digital human tech details, I am happy to discuss my experience further. Feel free to contact me at yl3262@cornell.edu.

First internship, my takeaways:

The Power of Product Thinking

As the sole intern on the product and design team, I had the unique chance to engage in in-depth research to understand our users, the market, and competitors. Collaborating closely with the PM, I actively participated in brainstorming sessions, adopting a product-centric mindset and creating the most appropriate features for our audience.

Don't Afraid to Speak Up

Initially, I found it challenging to put forth ideas during company-wide meetings, particularly in the presence of experienced colleagues. However, following 1:1s with my mentor designer, I adopted the practice of preparing bullet-point lists and rehearsing my presentations in advance. As my internship progressed, I gained the confidence to present my design solutions, even in front of our CEO.

Communication with Devs

Witnessing the progressive integration of design work was an immensely rewarding experience. Working closely with both front-end and back-end developers has underscored the significance of robust design systems while emphasizing the critical need for thorough evaluation of edge cases.

Creating Engaging Conversational AI to Eliminate Loneliness in GenZ Users