Overview
4-WEEK DESIGN SPRINT
Delivered in a fast, four-week sprint alongside a cross-functional team of three (Creative Director, Senior UX Designer, Senior UI Designer). While roles were defined, discovery and ideation were highly collaborative across all disciplines. Rather than being an extremely polished visual prototype, we kept things minimal and somewhat brand-agnostic for user testing purposes.
Week 1: Onboarding, stakeholder alignment, and product ideation
Weeks 2–3: Wireframing, interface design, and high-fidelity prototyping
Week 4: Evaluative user testing, design iteration, and executive presentation
Google Gemini enteprise
Product Design
User Flow Architecture
Interactive Prototyping
Evaluative User Testing
SENIOR UI DESIGNER
Deloitte Digital x Google
2026
Collaboration within AN LLM
SOLUTIONS
• Create the ability to break out into parallel conversations while still leveraging the primary context and resources
• allow users to work and iterate within private conversations/workspaces
• allow users to share selected work with the larger team when ready
• design seamless and intuitive interactions and spaces where users understand what is private and what is shared with the larger group
CHALLENGES
• an LLM interface is inherently limited especially for multiple users interacting with AI in a single conversation
• Multiple prompts can completely shift the AI’s focus and output and lead to conflicting instructions
• The conversation can also become bloated with various user prompts, messages, and ai output
• standard LLM interfaces force a single chronological timeline preventing parallel work
*This case study is a truncated version of a broader end-to-end product initiative.
1. Carlos opens gemini enterprise…
2. CARLOS opens the zeropop campaign conversation…
MAIN MENU
Distinguishing between Group and Personal Conversations was key
3. THE TEAM chats within the primary conversation…
SIDE PANEL AMENDMENTS
Adding “Shared Pins” to the Side Panel was the most inuitive
4. Aisha opens a PARALLEL personal conversation…
5. aisha works with an ai agent…
6. THE AI AGENT GENERATES AN IMAGE…
7. AISHA shares the image with the team…
8. THe team reviews the ad…
USER TESTING RESULTS
Over several days, we evaluated this prototype with four industry participants. The testing validated our core hypothesis regarding the importance of human-AI collaboration within LLM interfaces. Additionally, it highlighted key interface limitations and provided actionable insights to guide future design iterations.
