AITrip • AI-Powered Travel Planner
AITrip is a smart travel planner that generates personalized itineraries using AI, combining user preferences, interests, and recommendations to simplify trip planning and enhance the travel experience in minutes.
Product Design
Visual Design
Interaction Design
AI Prompt Engineering
Scalable Backend Design
My Role
Product Designer
UX / UI Designer
Interaction Designer
AI Prompt Strategist
Team
Me (Design)
Luvkesh Agarwal (CEO)
Dhiraj Bastwade (Tech Head)
Pratik Patil (Engineer)
Timeline
2023 (Jun - Aug)
Plan your travel
Problem & Context
The Challenge of Trip Planning
Trip planning sounds exciting! but in reality, it’s often overwhelming. Travelers face information overload, biased reviews, and endless options, often leading to stress and wasted time. While AI tools like ChatGPT can help, most people don’t know what to ask or how to structure queries to get useful travel itineraries.
Who We Designed For
AITrip is for anyone who wants a simple, personalized itinerary without spending hours on research. From solo backpackers to families, the app aims to reduce stress, save time, and create a balance between adventure and relaxation.
Why It Matters
Travelers struggle with:
Sifting through countless reviews and endless browser tabs
Fitting must-see spots into their limited time
Exhausting over over-planning trip or wasting days due to under-planning.
Managing logistics, costs, weather and safety.
FOMO and decision fatigue due to the choices.
The result? Lost time, wasted energy, and reduced enjoyment.
The Opportunity
Trip planning desperately needed streamlining and personalization. We saw the potential of AI, and a unique opportunity to bridge the gap between generic suggestions and truly helpful, customized and user-focused travel itineraries. The timing was right.
Research & Insights
Primary research
I started the project with a brief initiation call with CEO Luvkesh Agarwal, he shared some observed pain points among his friends who were planning trips. To validate those points, I also had quick conversations with colleagues and peers who were actively planning their own travel,
Key Findings
Google and TripAdvisor searches were time-consuming and often contradictory.
Travel agencies provided good trip packages but lacked personalization.
Independent planning of piecing activities, curating restaurants, and travel logistics was exhausting.
Some travelers wanted a curated itinerary, while some valued flexibility and control.
Secondary research
While working on this product back in 2023, there were no direct competitors who were offering what we envisioned for the product. So to understand more about market, spot gaps and opportunities, I explored two existing tools.
Roamaround.io (now Layla.ai)
Strengths
Clear breakdown of days into morning, afternoon, evening, bedtime.
Weaknesses
Limited customization, outdated UI, itineraries felt generic and unresponsive to user preferences.
Wanderlog
Strengths
Clean, modern UI with budgeting and trip organization features.
Weaknesses
Expected users to build their own itineraries from scratch rather than generating them.
Key Learnings
Breaking down days into morning, afternoon, evening and night improves readability.
Restaurant and cuisine suggestions adds a touch of authenticity.
A modern, user-friendly, and customizable UI is essential.
Drafting Design Goals
With insights from both research phases in place, I used affinity mapping to organize them into key theme and convert them into “How Might We” questions to guide the design direction.
How Might We…
Simplify trip planning into clear, single itinerary without making users feel overwhelmed?
Enable travelers to personalize their itineraries according to their own travel style, interests and preferences?
Reduce the manual effort involved in planning while still giving users flexibility and control?
Make AI interaction effortless through guided, structured inputs to improve usability?
Design a clear, modern UI that presents itineraries in simple to read segments and minimize decision fatigue?
Build trust and delight by integrating local transport, accessibility and safety into the planning experience?
Ideation & Design
With clear research insights and early inputs from CEO Luvkesh Agarwal, I began shaping the MVP scope and exploring how AI could simplify travel planning. The goal was to move beyond just generating itineraries and create something that travelers would actually trust and enjoy using.
Broadening Value Beyond Itineraries
We realized that simply generating itineraries wouldn't be enough to stand out. To deliver true value, the product needed to think ahead for users and surface relevant contextual recommendations like
Closest airports and best travel routes (inspired by Rome2Rio)
Recommended areas to stay
Hotels, restaurants, nightlife, and shopping options
Packing essentials and must-do activities
This could make the product feel more like a personal travel concierge than a static itinerary generator.
Early Sketches & First Drafts
We initially scoped two user types:
Users with a destination in mind
Explorers who wanted suggestions based on interests (adventure, cuisine, shopping, etc.)
Refining the Flow Through Iteration
Through frequent discussions and whiteboarding sessions with stakeholders, the design evolved through key refinements:
Before
Two user scopes (Those who knew their destination, Those who wanted to explore new places)
Two date selection fields (Depart and Return date)
After
Auto-fetch source location based on IP address / Location permission
Remove Return date and reduce form to trip duration and start date, this reduced the cognitive load for the users.
Removed customize itinerary to include it later in stepper approach.
Simplified Input Flow
We shifted from an overwhelming “all-in-one form” to a stepper approach
Before
Long customize itinerary pop-up.
After
Implemented stepper approach which collects
Travel Details:
Travel Style (Solo, Couple, Friends or Family) + Traveler count details + Arrival and Departure timingsBudget & Interest:
Budget per person + Trip Interests
Earlier, we used a FAQ-style layout, where users could expand questions like “Book flights to Manali’s nearest airports” or “Best ways to reach Manali from Mumbai” to reveal details. But as the content grew, users had to scroll bit too much and may get stuck in all the expanded FAQ sections.
To improve clarity and navigability, I replaced it with a new a section called "Your Smart Destination Guide". This made the experience feel more app-like and intentional, Thereby. turning AI outputs into clean, user-friendly travel guide.
Before
Progressive disclosure was done using FAQ styled sections.
After
A left navigation menu with sections such as “How to Get There,” “Getting Around”, etc.
Clicking any category dynamically opens its content on the right panel, allowing users to explore at their own pace without feeling overwhelmed.
AI Prompting & UX Challenges
Designing with AI wasn't just about aesthetics, we had to make sure that the AI generated responses could be mapped into a structured and user-friendly format while making sure that it was accurate, useful and personalized.
Key Challenges
Unstructured AI Responses
ChatGPT generated useful content but it lacked a consistent structure making it hard for engineers to map it onto UI components.
Personalization vs. Simplicity
We needed to make sure that itinerary was tailored as per the user's preferences, without asking them to "prompt engineer" or fill too many input fields.
Incomplete or Generic Itineraries
While testing early outputs, there were lot of gaps in regards to time based segments and contextual details like meal and evening plans.
Scaling Beyond Itineraries
We wanted to provide smart destination guide that could provide suggestions for restaurants, hotels, nightlife, shopping, etc, for the destination.
💡 My Solution
After testing numerous ChatGPT queries and hitting its daily usage limits for over a week, I leveraged my IT engineering background and my acquired knowledge of prompt engineering to propose the following solution
Structuring the Chaos
I rewrote the queries such that the responses always returned in JSON (RFC8259 compliant) format with consistent key-value pairs.
I structured the queries in a manner that the responses are always around the travel domain, thereby eliminating AI hallucinations and unrelated outputs.
This made it easier for the developers to parse and map the responses directly onto UI components, making the system scalable, reliable and developer-friendly.
Making Personalization Effortless
I proposed {{variables}} which collect and store user inputs progressively in the entire form.
Each variable was appended in the query dynamically making sure that all user's style and preferences were present in the requesting query.
Ensuring Complete, Trustworthy Outputs
I created the JSON query in a manner that it always returned each day in time segments while also taking users arrival and departure time into consideration.
I also made sure that the time segments had relevant activity + restaurant suggestions, this ensured that the itinerary received in the response was detailed and relevant.
Extending Beyond Itineraries
Once the core of the product was stable, I created a modular query architecture that powered the MVP.
Interest & Budget Query → Captures user preferences and suggest interests and budget based on the destination.
Title & Overview Query → Generates trip overview and suggests best months to visit, ideal trip duration and expected weather.
Itinerary Query → Delivers structured itinerary divided into days and time segments with suggestions.
Destination Guide Queries → For restaurants, shopping places, nightlife, etc.
Future-Proofing with Engineering Collaboration
From my testing, I realized that prompts are not static. The queries that work today might not work tomorrow.
To address this, I designed a complete backend editor that allowed designers and engineers to
Add / Remove / Update any prompt templates.
Switch between different AI models.
This ensured that the system remained flexible, scalable, and future-ready.
Prompts hidden due to privacy reasons.
Visual Design
With the MVP scope and AI query system in place, I moved into designing the interface and visual identity for the product. I ensured that the front-end and back-end worked seamlessly together and every visual and interaction design decision was made with future growth and technical scalability in mind.
Human-Centered Visuals
Instead of generic icons for Solo, Couple, Friends, and Family trip types, I used AI-generated vector illustrations that made the interface friendlier and more relatable. This simple switch brought warmth and personality to the experience.
Interest Selection
While AI effectively suggested relevant interests for the destination, some users wanted to add specific or personal interests. So, I designed a flow allowing users to add their own interests, which were then appended directly to the Itinerary query prompt.
However, through testing, I found out that adding too many interests, especially for short 3-5 day trips, make the itineraries unrealistic and overwhelming. After stakeholder discussions and validation, we limited selections to five interests, balancing creativity with quality results.
Designing for AI Limitations
Back in 2023, AI responses were noticeably slower. To avoid user frustration during the wait, I created Lottie-ready animations in After Effects to add personality and delight through motion on top of the already existing skeleton loading animation.
Extending Functionality Beyond the Core UI
I designed additional features that made the MVP more usable, shareable, and personal:
Flask icon for feedback, users could report bugs or submit ideas.
Share, allowing users to share itineraries with friends and family seamlessly.
Smart Download System, a context-aware download experience through which users can download their itinerary (including Smart Destination Guide sections that they had actually viewed).
Identifying Business Touchpoints
Although the product was completely free, I saw opportunities to drive business without disrupting the user experience and create a natural, non-intrusive funnel for engagement.
Login to Plan: Users must log in to start planning, allowing us to capture engagement and personalize experiences.
Request a Callback: Travelers could connect with the AerTrip Holidays team to refine and execute their plans.
Visa Guidance: A “Know More” link before the itinerary to redirect users to the visa page, adding value while connecting them to relevant services.
Smart Destination Guide Integrations:
How to Get There: A single click on “Book Now” used traveler count, source/destination, and dates to redirect users seamlessly to the flights page.
Where to Stay: Hotels followed a similar workflow, linking directly to hotels page.
Testing & Iteration
As an MVP, speed was critical. Formal usability testing with travelers wasn’t feasible, so I relied on rapid feedback loops with stakeholders and engineers, updating screens and queries every few days. What we didn’t validate with users upfront, we compensated for with tight collaboration, adaptability, and an eye on the future, ensuring that the product was launch-ready and prepared to evolve once in the hands of real travelers.
Insights & Impact
Design Trade-offs & Future Vision
During the MVP design, we made decisions that balanced short-term goals with long-term vision:
Daily Activity Level Slider
We shipped a slider that let users set their preferred daily activity level, from 4 to 12 hours. During testing, I realized it had little impact on the generated itineraries, which could confuse users if their choices didn’t reflect in the results. However, stakeholders decided to keep it as it gave travelers a sense of control and acted as a placeholder for future refinement.
Visual Enrichment with Photos
I envisioned itineraries as more than text, proposing photo carousels of key attractions to make them visually engaging. Though time constraints kept it out of the MVP, I was thrilled to see the team later bring this idea to life exactly as I had envisioned.
Product Impact
Scalable backend design helped engineers edit / remove / update dynamic queries.
The JSON-based prompt responses made it easier for developers to map it onto the UI easily.
Stakeholders saw how AI could solve a real travel-planning pain point in a structured, scalable way.
A unique product that combined itinerary generation with contextual recommendations, something the competitors lacked at that time.
Limitations & Learnings
No real user testing before launch, as speed-to-market was the top priority.
MVP didn’t cover budgeting, scams, or hyperlocal recommendations, we marked it for future versions.
Designing for AI products isn’t just about the UI, it also requires engineering-aware design decisions (like structured prompts).
In rapid MVPs, collaboration and continuous feedback can sometimes serve as effective alternatives for user validation.
Success lies in balance, keeping MVPs focused and shippable, and leaving room for future growth.
🌱 My Reflection
This project taught me how to blend design, AI, and engineering thinking to create something new. It sharpened my ability to not only craft interfaces but also solve complex technical challenges like dynamic AI prompts and structure AI responses.