Intelligent Construction
Designing transparent, evidence-based, trustworthy AI-assisted construction workflows with fragmented real-world data, grounded in field research.
type
time
Team
6 Research Scientists,
2 Researchers, 1 Dev
Role
UX Research,
UX Design & Prototyping
Construction teams check progress by piecing together schedules, BIM, drawings and site updates across different . On Autodesk’s INTC research team, I conducted field research with 16 professionals across 6 general-contractor firms and turned the findings into an AI-assisted progress-verification experience. I designed how AI confidence and evidence show up, so users can check what the AI-infused Datahive capabilities conclude. The work was demo'd at Autodesk University 2026, where I helped run feedback sessions with industry experts.
outcomes
Hackathon Challenge
Create an impactful travel, finance, or self-wellness app that integrates AI meaningfully.
Constraints
From the get go, I identified our constraints to inform our product direction.
Considering the limited time for recruitment, we decided to target students as our users to leverage our own experience for user research.
To utilize our developers' strengths in web frameworks, we prioritized desktop use cases and bypassed mobile optimization.
Identifying the Problem Space
To ensure a focused and impactful solution, I facilitated a lightning brainstorming session to identify problem spaces grounded in our own experiences and pain points.
Frictionless Investment
Shopping/e-Commerce
Budget Travel Planning
We voted budget travel planning to be our focus as it resonated the most with all members. From our pain points, we synthesized four directions.
Idea Prioritization
After narrowing down on our core pain points, we synthesized three approaches, and evaluated them against three factors:
This decision then informed our design challenge ↓
The Challenge
How might we help budget-conscious students discover and plan unique travel experiences?
Secondary Research
Using Perplexity AI, I quickly researched some emerging trends and gaps in the market.
"Mature products like Wanderlog has limited AI or financial integrations."
"Many users report that travel apps — especially booking platforms — are difficult to navigate, with unintuitive interfaces and confusing flows."
"Travel apps sometimes operate more as marketing channels than honest guides, filtering out negative or mixed feedback."
To tackle the difficulty of navigation and concern of ad-filled recommendations, our product aims to create an easy browsing experience with trustful AI-assistance.
User Research
For qualitative user research, we reflected on our own experiences which informed our persona.
Wants to avoid tourist traps overhyped on social media.
Research takes a long time to find fun & budget-friendly things to do.
Wants to use LLMs to his advantage but finds prompting a hassle.
Solution Ideation
We synthesized our insights into core features and prioritized three:
Feature 1
Hidden Gems Search
View/search member-reviewed activities based on location, price, and type.
✓
Avoids tourism traps
Feature 2
AI-Recommendations
Get personal suggestions by talking to an AI assistant, like a real person.
✓
Reduce hassle of research
Feature 3
Saving to Trips
Save activities to separate trip folders with customizable totals (i.e. cost, type, map).
✓
View all activities in 1 place
To ensure a well-focused solution, we asked ourselves:
Question 1
Is this AI-integration really necessary?
The AI voice assistant (powered by ElevenLabs) is designed to reduce decision fatigue via a fully automated, hands-free discovery. At the same time, we balance this with a manual exploration mode, where users earn redeemable perks for contributing trustworthy reviews, creating a self-sustaining ecosystem of verified travel intel.
Question 2
How would this differ from market competitors, like TripAdvisor?
Rather than providing a comprehensive acitivity booking site or itinerary builder, we are designing a minimal task flow solution. We prioritized on a 'loose-planning' framework that allows users to find tailored activities quickly with minimal effort.
Task Flow
I mapped two pathways for discovery: manual browsing and with AI-voice-assist.
These two approaches ensured that we design for both user types who prefer granular control via filters and those seeking an effortless, hands-free planning experience.
Task Flow: Manual Exploration
Task Flow: AI-Voice-Assisted Search
Lessons Learned
Be transparent, overcommunicate, align
Right from the get go, we discussed our skills and preferences, and shared our own experiences which became the basis of our design and roadmap. I …
Prioritize even more ruthlessly
Despite our efforts in narrowing scope, our final solution included three complex features. I learned that establishing a clear, ruthlessly prioritized roadmap that strips down into even less features could grant more refined solutions in a time-sensitive environment.
Designing with everyone
During this hackathon, I primarily used Figma Make to generate high-fidelity prototypes only after sketching out low-fidelity wireframes. In retrospect, I learned that onboarding the entire team to gen-AI tools earlier would have allowed us to leverage our collective creativity for quicker iterations.