Improving an existing AI experience within real product constraints
MyExpert is ASUS’s AI-powered productivity platform for commercial PCs. I joined after the first version was already in place, so most of my work started with understanding what was already there before deciding what needed to change.
Responsibilities:
AI Interaction Design
UX & Interaction Design
Design System
( About this case )
This case study uses publicly available product information and focuses on my design approach and learnings. Internal product strategy, technical implementation, and confidential design details have been omitted.
AI INTERACTION
Helping users understand what “ask anything” actually meant
MyExpert Chat gives users an open input and asks them to start typing. The harder part was making its actual scope visible before users ran into a limit.
What MyExpert Chat supports
CHALLENGE
An open-ended chat with a defined task scope
“Ask anything” sets a big expectation. MyExpert couldn’t reliably support every request that wording invited, so the problem wasn’t only what the assistant could do. It was what users thought it could do.
01 | DESIGNING FOR PREDICTABILITY
What happens when “ask anything” meets a boundary?
The tricky part wasn’t getting MyExpert to answer more questions. It was deciding what should happen when it couldn’t. If the assistant kept trying anyway, the experience became harder to predict. I’d rather make the boundary clear than make the AI sound smarter than it was.
02 | LOOKING BEYOND THE INTERFACE
Chat was only one part of the experience
Fixing individual responses only got me so far. MyExpert already had several tools, each with its own entry point. At some point I started wondering whether the bigger problem was Chat itself, or how users knew where to start in the first place.
03 | VALIDATING THE PRODUCT STRUCTURE
Testing how users made sense of the product
Before pushing the Chat idea further, I wanted to know whether the product structure made sense outside the team that built it. The research started with a simple question: did users group and navigate these capabilities the way we expected them to?
04 | REFLECTION
Finding where AI is actually useful
MyExpert made me think more carefully about where AI adds the most value. Chat was useful when people needed help finding a direction. For people who already knew what they wanted, a direct path often made more sense.
＿
DESIGNING WITH AI
A thinking partner, not a decision maker
I used AI to work through the research faster, especially when comparing interview notes and looking for recurring patterns. One pattern made me rethink the question I was asking. I started with: “How can Chat be a better entry point?” The research led me to ask: “When is Chat actually the better entry point?” AI helped me get through the information faster. Deciding that the question itself needed to change was still my call.