Progressive Disclosure
Lead with editorial (Star Tribune wins), surface AI for power users (ScreenGeni.us wins).
Strib Bot
AI-Powered Search for Minnesota State Fair
Project Scope
Platform: Responsive Web
Timeline: 2024
Current Stage: Live
Role & Team
Product Designer (Contract)
Star Tribune Team
ScreenGenius Partnership
Technical Details
React/Next.js
OpenAI GPT-4
Vercel
Key Metrics
6,900 User Sessions
12,000 AI Searches
1.9M Fair Visitors
PROBLEM SPACE
Two clients. Competing goals. One product. The challenge wasn't solving discovery—it was deciding whose version ships.
THE CHALLENGES
AI vs. Editorial
ScreenGeni.us wanted to showcase semantic search. Star Tribune wanted to showcase journalism. Same product, competing goals.
Trust vs. Utility
AI could surface great matches. Users with dietary restrictions couldn't verify the system wasn't hallucinating ingredients.
Depth vs. Speed
Star Tribune had rich content worth exploring. Users had 12 seconds of attention in a hot, crowded environment.
THE OPPORTUNITY
How do we balance editorial showcase, AI demonstration, and user needs—when each pushes against the others?
Transparent Attribution
Sparkle emoji distinguished AI from journalism. Users trusted both more when they knew which was which.
Designed for the Environment
12 seconds of attention in crowds and heat. Scannable cards first, deep content for users who want it.
Home Page: From AI-First to Editorial Navigation
Star Tribune wanted badges. ScreenGeni.us wanted search. Users just wanted to find food fast.
Voice Removed
Budget cut voice. Placeholder text ("Something cool on a hot day") teaches semantic search behavior instead.
9 Editorial Badges
Star Tribune requirement to showcase 10 years of editorial curation. Each badge maps to their content categories.
Search Deprioritized
Search bar stays above fold but badges dominate visual hierarchy. Editorial wins the home screen.
Search Results: Building Trust Through Transparency
AI mistakes become publisher mistakes. Star Tribune cut any data they couldn't validate—price, distance, ingredients.
✨ Attribution Note
Header explains sparkle = AI-generated. Users know which descriptions to trust for dietary accuracy.
Editorial Badges
"Strib Classics," "Best Value"—Star Tribune's curation visible on each card. Adds credibility.
Sparkle on AI Text
AI-generated descriptions marked inline. Editorial descriptions unmarked. Clear source distinction.
13 Results
Pared down. Filters, price, distance, favorites all cut. Focused on what shipped.
Vendor Detail: From Prototype to Ship
V1 explored what's possible. V2 shipped what timeline, data, and scope allowed.
Vague Location
"West End Market" based on available data. Precise navigation descoped.
Editorial Snippet
Star Tribune review excerpt with link to full article. Drives traffic to journalism.
Editorial Badges
"Worth the Wait," "One-of-a-Kind"—Star Tribune's curation visible and tappable.
Menu Tabs
Food & Drink and Alcohol separated. Cleaner scanning for users.
INSIGHTS
What I Learned
1
Publisher Stakes Are Different
News organizations can't treat AI like a startup. One wrong ingredient claim is a credibility crisis. Star Tribune cut anything editorial couldn't verify—that's not scope creep, that's brand protection.
2
Prototypes Explore, Products Ship
V1 asked "what's possible?" V2 asked "what can we build, verify, and support?" The gap between those questions is the actual job.
3
Engineering Tradeoffs Are Design Decisions
What engineers can scrape, verify, and ship shapes the product as much as wireframes do. Design with build constraints, not against them.
Ready for more?
Aura Finance
A comprehensive financial wellness platform targeting Gen Z and Millennials.
B2CGamification
Maro
A mental health support platform for parents.
B2CMental Health
Mixed Signals
A platform for exploring and learning from ethical decisions through structured scenarios and reflection.
0-1 ProductB2B2C