What I did

UX Strategy

Led the UX strategy, conducting stakeholder and user interviews, performing UX audits, and analyzing analytics to identify pain points and opportunities.

Human-Centric Design

Integrated user feedback and business goals to deliver high-quality solutions that resonate with users. Implemented a design system to ensure consistency and scalability

Project Overview

I had the opportunity to work with Mathieu Bayou and Nicolas Jellab, the co-founders of ai_licia, to completely redesign their innovative AI assistant for streamers. Ai_licia enhances live streaming by offering real-time interaction, audience engagement, and stream management support, with the goal of bonding with audiences and assisting streamers in a simple and fun way. Through extensive research, I addressed key challenges in managing interactions, customizing AI behaviors, and streamlining platform integration.

What I did

UX Strategy

Spearheaded the UX strategy, conducting stakeholder and user interviews, performing UX audits, and analyzing analytics to identify pain points and opportunities

Human-Centric Design

Integrated user feedback and business goals to deliver high-quality solutions that resonate with users. Implemented a design system to ensure consistency and scalability

Project Overview

I had the opportunity to work with Mathieu Bayou and Nicolas Jellab, the co-founders of ai_licia, to completely redesign their innovative AI assistant for streamers. Ai_licia enhances live streaming by offering real-time interaction, audience engagement, and stream management support, with the goal of bonding with audiences and assisting streamers in a seamless, fun way. Through extensive research, I addressed key challenges in managing interactions, customizing AI behaviors, and streamlining platform integration. The user-centered redesign has already made a notable impact on user engagement and stream quality, showcasing the power of UX design and AI in live streaming.

Design Process

Before / After

Research

I conducted in-depth user interviews with streamers to identify their key challenges and goals, focusing on AI customization, stream management and audience engagement. Our primary objective was to uncover pain points and refine ai_licia's interactions to better serve both streamers and their audiences. This research laid the foundation for a user-centered redesign that directly addressed their needs.

Research

I conducted in-depth user interviews with streamers to identify their key challenges and goals, focusing on AI customization, stream management and audience engagement. Our primary objective was to uncover pain points and refine ai_licia's interactions to better serve both streamers and their audiences. This research laid the foundation for a user-centered redesign that directly addressed their needs.

Participant 1 - Arya

Twitch Streamer

“I had a bit of difficulty trying to navigate the core personality to get it fit quite right”

Participant 2 - Bass

Twitch Streamer

“I knew that I answered that somewhere but I had to search where that was when I want to change it”

Participant 3 - Arya

Twitch Streamer

“In my head, it's like it still looks like it's a work in progress in a bit. (...) it reminds me when I used to make my own website in wordpress for some reason”

Participant 4 - Rob

Twitch Streamer

“It also kind of annoys me because I tend to use the dark mode and sometimes colors make me miss certain things in the light mode. (...) but it's still kind of messy”

UX Audit & Analytics

Before jumping into solutions, I ran a detailed UX audit alongside real usage data using Microsoft Clarity. We looked at heatmaps, dead clicks, scroll depth, and quick backs to understand where streamers were dropping off or struggling. We also noticed a high bounce rate in key settings screens and confusion around profile configuration.

UX Audit & Analytics


The UX enhancements implemented in the TalkToo app, including the new onboarding process, Discover page, bottom navigation, and the bespoke backend management platform, have had a profound impact on both user engagement and business metrics. These initiatives demonstrate the powerful synergy between AI technology and user-centered design, underscoring the importance of strategic UX in driving app success.

+62%

Avg. Session Duration

Avg. Session Duration

+25%

Performance Score Improvement

Performance Score Improvement

-75%

Quick Exit Rate

Quick Exit Rate

90%

Return User Rate

Return User Rate

Information Architecture Rework

One of the first challenges was the IA. Settings were scattered across multiple menus, and users didn’t know where to go for specific tasks like customizing ai_licia’s behavior or reviewing chat memory.

Information Architecture Rework

One of the first challenges was the IA. Settings were scattered across multiple menus, and users didn’t know where to go for specific tasks like customizing ai_licia’s behavior or reviewing chat memory.

Key Improvements

Quick adjustments

Easy sliders for changing chattiness, interaction balance, and talking style during streams.


Quick adjustments

Easy sliders for changing chattiness, interaction balance, and talking style during streams.


Quick adjustments

Easy sliders for changing chattiness, interaction balance, and talking style during streams.


Memory hub

A new dashboard showing what ai_licia has learned and how she’s interacting with the audience.


Memory hub

A new dashboard showing what ai_licia has learned and how she’s interacting with the audience.


Memory hub

A new dashboard showing what ai_licia has learned and how she’s interacting with the audience.


Live interaction tools

Easy sliders for changing chattiness, interaction balance, and talking style during streams.


Live interaction tools

Easy sliders for changing chattiness, interaction balance, and talking style during streams.


Live interaction tools

Easy sliders for changing chattiness, interaction balance, and talking style during streams.


Impact & Feedback

Returning users increased

from 87.46% to 89.75%, meaning more streamers kept coming back to use ai_licia.

Returning users increased

from 87.46% to 89.75%, meaning more streamers kept coming back to use ai_licia.

Returning users increased

from 87.46% to 89.75%, meaning more streamers kept coming back to use ai_licia.

Average session time jumped

from 3.7 minutes to 6.0 minutes, people spent more time using ai_licia

Average session time jumped

from 3.7 minutes to 6.0 minutes, people spent more time using ai_licia

Average session time jumped

from 3.7 minutes to 6.0 minutes, people spent more time using ai_licia

Quick exits dropped

from 11.15% to 2.61%, which told us users were finding value and staying engaged instead of bouncing out.

Quick exits dropped

from 11.15% to 2.61%, which told us users were finding value and staying engaged instead of bouncing out.

Quick exits dropped

from 11.15% to 2.61%, which told us users were finding value and staying engaged instead of bouncing out.

Dead clicks went down

by over 25%, making it clear that the new UI made navigation easier and more intuitive.

Dead clicks went down

by over 25%, making it clear that the new UI made navigation easier and more intuitive.

Dead clicks went down

by over 25%, making it clear that the new UI made navigation easier and more intuitive.

Performance score improved

from 66 to 87, the new interface wasn’t just cleaner, it was faster and smoother to use.

Performance score improved

from 66 to 87, the new interface wasn’t just cleaner, it was faster and smoother to use.

Performance score improved

from 66 to 87, the new interface wasn’t just cleaner, it was faster and smoother to use.

User intent shifted positively

with high-intent sessions going up and low-intent ones dropping significantly.

User intent shifted positively

with high-intent sessions going up and low-intent ones dropping significantly.

User intent shifted positively

with high-intent sessions going up and low-intent ones dropping significantly.

Impact & Feedback

Conclusion

Designing for live streamers taught me a lot about fast-paced environments, real-time feedback loops, and designing for people who need to stay in flow. Balancing AI power with user control was tricky—but worth it.

Conclusion

Designing for live streamers taught me a lot about fast-paced environments, real-time feedback loops, and designing for people who need to stay in flow. Balancing AI power with user control was tricky—but worth it.