AI Assistant Discovery Dashboard
Summary
I redesigned an AI assistant dashboard for Reality AI Labs, streamlining educators' access to innovative AI tools and improving their search experience.
I owned the Favorite Assistants workflow, the History section, responsive desktop and mobile layouts, and interaction design across the dashboard.
- My Role
- Lead Product Designer
- Timeline
- Nov 2024 - Apr 2025
- Team
- Cameron MoralesJosh GrossJulio RamirezKamal RameshSarah Nguyen
- Skills
- User researchCompetitive AnalysisWireframingPrototypingProject Management
Background
Target users
- Educators (K-12)
- Educators need fast access to the right AI tools for them between lessons, grading, and planning.
- School Administrators
- School admin needs to roll out shared AI tools at scale, and to trust that every educator can find and use them.
Problem
Discovery dashboard that didn’t support discovery
Educators couldn't quickly find tools that fit their needs. Access was scattered, categories were confusing, and recommendations felt generic.
How might we make AI tool discovery fast, intuitive, and personal for educators?
Auditing the previous design
Limited search
No advanced filters, so finding the right tool takes longer.
Slide 1 of 4: Limited search
Research
Competitive Analysis
Our team reviewed educational platforms to see how they organize and surface content.
I analyzed Google Career Certificates, Khan Academy's Khanmigo, and Duolingo, focusing on card layouts and how each product highlights favorites.
We combined our findings into a competitive analysis grid, surfacing common patterns that shaped our design direction.
Survey Results
I chose to conduct a survey to hear from educators at different career stages quickly. I surveyed 12 K–12 teachers, with 1 to 10+ years of experience, about how they find and use AI tools.
Design process
Collaborating across a team
Mapping user flows
I designed the Favorites user flow, mapping each step an educator takes to save and return to an assistant.
Individual wireframes
Each team member wireframed layouts independently. My design reduced vertical scrolling and gave assistants clear sections: Trending, Favorites, and categories.
A/B Testing
We split into two groups and combined our individual wireframes into two versions. My group's version used a horizontally scrolling grid, my stacked-card design, and a new Recently Used section.
We tested both versions with 3 stakeholders to decide which features to keep, including carousels, filter buttons, drag-and-drop, and a trash button.
final product
Desktop View
Onboarding guide
A short walkthrough introduces first-time users to key features, so they can start finding tools right away.
Discovery options
Grid-based sections and filters organize assistants by category, making a large library easy to browse.
Choose favorite assistants
Tapping the star on any assistant adds it to Favorites. Recently Used keeps go-to tools one click away.
final product
Mobile View
Midway through, the team asked for a mobile version, with one week left before the deadline. I adapted the desktop design for smaller screens while keeping the core features.
Adapting for mobile
Trending assistants appear in a swipeable carousel, and Assistants and Favorites each get their own page to keep screens uncluttered. Notifications alert educators when their assistants are updated.
Impact
Feedback from stakeholders
“This dashboard clearly supports our growing user base and AI tools.”
“I love the quick access to your most useful tools”
“I appreciate the usability for both first-time and seasoned users”
Key takeaways
Aligning many perspectives
With six designers, we had more strong ideas than we could use.
I set up a shared system for idea sharing and task assignment, which kept us organized and strengthened my communication and project management skills.
For the user, not the interaction
Stakeholders pushed back on some of our interaction ideas, like drag-and-drop for Favorites.
That feedback shifted my focus from impressive interactions to what users need most. I began asking: What's easiest for the user, and how can they find the right tool fastest?
