Infilla Forum Search
Transformed Infilla’s Forum search experience to help city planners access information more efficiently and reliably.
Role
Product Design Intern
Timeline
October 2025 - January 2026
Context
Who is Infilla
Civic-tech company that partners with U.S. cities’ planning departments to provide modern tools that help them work faster, collaborate better, and serve their communities.
What is Forum
Forum is Infilla’s Q&A and search platform used by planners, zoning staff, and citizens to…
Ask questions to experts
Review past discussions
Access verified regulatory sources
Primary Users
City and county zoning staff who work at permit counters, talking directly with applicants under constant time pressure to provide accurate information.
Core Needs
These planners need quick access to accurate, authoritative information that gives them confidence before sharing answers on zoning, land use, and development procedures with applicants.
The Problem
Infilla tasked us with designing an AI integration into Forum to leverage their regulatory knowledge base. Existing user research conducted by the company showed the following.
Search variation
Keyword search fails on terminology variations.
Extreme result volumes
Results would either show too much or too little.
Diverse tech fluency
Users range from tech-savvy planners to professionals nearing retirement.
Initial Research
Competitive Analysis
To piece how AI might fit into the Forum product, we looked to how competitors handle search, trust, and usability.
Affinity Mapping
Based off the competitive analysis we affinity maps some themes.
Early Explorations
Grounded themes into early concepts of an opened ended Q&A AI chatbot that would answer questions with citations.
Survey Insights
We also surveyed 10 available planners to understand how users currently use Forum in their workflow, their search behavior, as well as their stance on AI.
80%
Of participants prioritized fast and accurate results
50%
Of participants wanted to review official code themselves
Most Participants
All Participants
Displayed low to medium confidence in interpretations generated from AI
Concept Design Remodel
To move fast we designed chatbot-esque solutions that would differ from the current search model, however survey results showed users just wanted better tools, not another GPT wrapper.
How might we…
Design an AI-enhanced search experience that works with users' keyword habits, addresses varying tech literacy and AI skepticism, and integrates seamlessly into the existing platform?
Enhanced Search Input
Give planners control through flexible and guided filtering
Suggested Filters
Automatically surface the most relevant zoning categories, code sections, and use types based on the user’s query, reducing manual filtering and improving results before search.
File Upload
Users can upload planning documents and site photos to add regulatory and physical context to a search, enabling results that reflect how similar codes and site conditions have been interpreted and applied.
Add Address
Use a specific address to return the zoning code, or past cases that apply to that location.
Research Loop
Support deep efficient research with, different types of sources, preview, and highlights.
Side Drawer Preview
To enhance the result browsing experience we designed a side drawer preview tool that displays Forum answers and AI parsed external document highlights.
Before
Questions are the only results, external Documents hidden within dropdown.
After
External documents included in search, source navigation to sort.
Unified results
Search results combine forum discussions with supporting code and documents, organized by source to keep research clear and manageable.
Concept Testing & Iterations
Before the end of the internship, we conducted five 30-minute concept tests with planners to validate designs and make changes accordingly.
Insight
Participants liked side panels but expressed they may be text heavy
Implemented highlights for Forum Previews (user generated content) and improved hierarchy for External Documents.
Insight
When search fails, participants went to external sources rather than posting on Forum
Updated microcopy on CTA's to encourage users to an unhappy path.
Insight
Some participants were unsure of how new features functioned
Tooltips existed on the platform were commonly ignored, provided toasts to pause and explain
Project Wrap Up
Final Deliverables
As the internship concluded, we delivered a comprehensive design document to set the team up for implementation, including annotated Figma files, user flows, research insights, and design rationale to guide development forward.
Learnings
The iterative process
While we waited for survey feedback, we explored designs that misaligned with what users wanted, but we still came out of it with findings that influenced final concepts.
Listen for nuance
Learned to dig into the why and greater contexts behind user interviews rather than taking certain reactions at face value. Sometimes
Overcommunicate in startups
Early hesitation to reach out to stakeholders slowed progress, but once we initiated regular check-ins, designs aligned faster and meetings became more productive