Designing an emotionally responsive AI companion
through prompts, characters, and interaction.
Average session depth
Return rate
Funding stage
The Problem
Despite its strong technology, user engagement was low
— most left after day one or two.
The Challenge
We had only a few weeks to diagnose the problem
and redesign the experience.
I quickly identified what mattered most and focused on high-impact changes.
Research approach
To understand the drop-off, we spoke directly with
users and studied their feedback.
I led user research together with the CEO, focusing on qualitative insights.
Research Insight
What we learned from users:
The research revealed clear patterns in where users disengaged.
To move quickly and strategically, I narrowed the scope to two high-impact flows.
This sharper focus guided every design decision that followed.
Problem statement
How might we help emotionally distressed users share more naturally and feel understood, to increase long-term engagement and retention?
Design Process 1
Prompt
Lowering the barrier to start.
Boopie’s previous version, Clapy, had a blank interface, users had to know what they wanted to talk about,
which created a high barrier to entry.
Competitive Analysis
While analyzing competitors, we noticed a common pattern:
Through competitive benchmarking, I identified a shared UX strategy across successful AI tools. They lowered the activation barrier by offering guided entry points.
hypothesis
This informed our hypothesis:
Design goal
Before jumping into solutions,
we set a simple framework of design principles.
Prompt Delivery
I developed two directions and assessed them using our design principles.
Lower the barrier to start with
approachable prompts.
Foster emotional engagement to motivate
users to explore more topics.
Lower the barrier to start with
approachable prompts.
Foster emotional engagement to motivate
users to explore more topics.
Iteration
Okay, we picked Option B. Now what?
I recommended Option B, as it encouraged focus despite showing only a few cards per day, and the team aligned.
But selecting the format was just the start. The real challenge was making the cards engaging long-term and scalable for growth.
Iteration
I iterated on the topic cards to align business and UX needs.
Final design below.
Iteration
Added Category Tags
We introduced themed categories to diversify daily topics.
This also enabled light monetization by letting users unlock preferred categories.
Iteration
Daily Quest & XP System
We integrated an XP system to encourage exploration.
Less-picked topics rewarded more points, and completing five cards unlocked bonus XP.
Iteration
Typographic Emphasis
Because the prompts were full sentences, I highlighted key words
to reduce cognitive load and make each card feel intentional.
|
Design Process 2
Response
Making the AI feel alive.
Research
Behavioral & Emotional AI Research
I reviewed behavioral and emotional AI research and found strong evidence that character-first approaches improve emotional connection, trust, and retention.
Competitive Benchmarking
I also analyzed leading products in the space and saw a consistent pattern.
Apps like Duolingo, Wysa and Finch use expressive mascots to humanize the experience and drive engagement.
Design Directions
Based on these insights,
I proposed three character-first design concepts.
Simple
Flat
Playful, friendly, and scalable, but needed
multiple variations and full-body animations.
Playful
Full animation
A cute, bust-up character that’s expressive and relatable, but typically resource-heavy to produce and maintain.
Expressive
Resource-heavy
Stakeholder review
However, when I shared this with the CEO and engineers..
To meet the MVP timeline, the team prioritized resource efficiency and minimal animation.
As the sole designer, building and maintaining a large customization library would have been too time-intensive.
recommendation
Minimal abstract character
with expressive face & simple customization
Through rapid iterations with the team, I combined the strengths of the Human-like and Abstract directions.
Boopie became a minimal abstract character with expressive human touches and focused customization.
recommendation
This direction balanced efficiency and expression
It reduced development overhead, enabled lightweight customization, and kept the character expressive and engaging.
It met technical constraints while leaving room for personalization.
final design
Emotionally responsive character welcomes user
Boopie opens with an emotionally responsive character, setting a seamless and engaging tone.
final design
Daily topic cards make starting a conversation feel effortless.
To reduce hesitation, I introduced daily topic cards that made starting effortless.
The character below subtly reacted as if waiting, reinforcing immersion and connection.
So, when the user swiped the character down, the daily topic cards appeared, ensuring the immersion wasn’t broken.
So… did it actually work?
Impact
Average session depth: 12 → 58 turns
Return rate: 2.1 → 3.7 visits/week
Impact
The tone of feedback completely changed.
In a pilot test with the mental health center and a public exhibition, we heard things like:
Impact
The pitch deck I designed also helped the team secure Pre-A funding
What I learned...
Emotional engagement is a design challenge — solved by crafting
the right signals at the right moments to help users feel understood.
