Project Overview
I worked with the Economic Development Partnership of North Carolina to create a consumer-facing AI product that assisted users with booking trips around the state of North Carolina. The project was the start of a larger initiative to incorporate more artificial intelligence into the NC trip planning experience through a mobile app.
This case study focuses on the research and design involved with striking a pleasant balance of AI involvement in the trip planning process.
Understanding How Humans Plan Trips
Before starting any design work, I studied how people plan trips today and where AI fits into this planning process. Three questions guided the majority of my research:
Where are people hesitant to use AI?
How do people normally explore their options?
How involved should AI be in the process?
The EDPNC provided me with several volunteers to speak with, along with access to some of their current site analytics to get a better understanding of where and how users look for ideas for their trip.
In order to balance AI involvement and self-guided browsing, these were the flow chart and findings that I came up with to help guide the design of the app:
Finding 1: People are curious about AI, but not convinced
Finding 2: PLANNING IS LONG AND EXPLORATORY
The average path to booking a trip spans 71 days (about 33 days of inspiration + 38 days of research and planning)
Travelers view about 140 pages of travel content in the 45 days leading up to booking
Finding 3: AI is welcome for research, not deciding
Among travelers who used generative AI, 73% used it for itinerary ideas and 67% for general research
Data provided by Global Rescue and Expedia
Designing a Hybrid AI Experience
Ultimately, the balance I came away with was a hybrid self-exploratory and AI-supported experience.
AI is an option, not required
The homepage couldn’t just be a chat window. Even though the core of the product is artificial intelligence, it needed to act as a browsable collection of recommendations with AI accessible at all times.
AI suggests, but users decide
The result of human-AI conversations should be ideas and inspiration rather than actual plans or itineraries.
Answers support browsing
Responses aren’t walls of text or loops of a dozen follow-up questions. Users needed to tell the AI what they wanted, and the AI needed to provide cards/pages as early as possible in the conversation.
Commitments belong to the user
Likes, adds, or any other future mechanism should be done by the user, not by the AI.
Lessons Learned
AI is an extremely powerful tool that can be used to enhance B2C products in really exciting ways. Upon speaking with users and doing research, the need for balance in any AI product became increasingly clear. This is not a "more is better" situation.
Users that want to interact with AI will seek that part of the product out, almost instantly. Those who are hesitant need to enjoy the product prior to any AI interaction. At a point, most users will eventually give it a try. An overbearing AI quickly scares these users away, but an AI that fits perfectly into the UX of the rest of the application is often accepted with open arms.
Let's connect
Looking to have a conversation? Reach out!
Whether you're a hiring manager looking to chat, a junior designer looking for mentorship, or an event coordinator interested in me speaking… drop a line!