Travel friends to Travel Agents

When our project began I imagined a very different travel platform. I wanted to explore a social space where people, especially solo travelers, could connect while traveling. My idea came from the thought that travel is shaped by the people we meet and not only the places we visit. A platform could help someone find a companion for an activity, for example, exchange advice with another traveler, making you feel less alone in an unfamiliar city.
That idea was supported by my own experience. I have worked in boutique hotels in Amsterdam, where travel often seemed personal and social rather than just transactional. A friend of mine had also considered a similar travel startup idea, which made me curious about whether there was an opportunity to build something around this concept. At that point I expected our project to focus on the social side of a trip.
As our team explored Layla, an AI travel assistant, the direction changed. Layla could already help travelers discover destinations and put together itineraries through conversation. We initially discussed the difficulty of choosing the right option among so many flights, hotels and activities. But the question that stayed on my mind was simpler: once someone has an itinerary, who actually books it?
Your plan can look complete while the traveler still has to check availability, compare prices and terms, book each component, and keep track of confirmations and more. If a flight changes, it may affect your hotel arrival or planned activities. I focused my contribution on explaining this gap between having a plan and having a trip that actually works.
Finding evidence for that problem taught me to be more precise. Expedia Group research showed that travelers viewed an average of 141 travel content pages and spent just over five hours with travel content before booking. Those figures were very similar to my own experience while browsing through Booking.com, Tripadvisor and other sites. Yet, they do not measure the work of making or changing reservations. A more relevant survey found that 54% of travelers were more likely to choose a platform where they could book all trip elements. Respondents also mentioned time savings and easier trip management as reasons to book them together.
Our proposed feature became an agent that could take an approved itinerary, check current options, present prices and terms, make authorized bookings and collect confirmations in one customer journey. If one reservation changed, it would check the effect on the rest. For example, after a later flight arrival, Layla could flag a conflicting activity, propose another option and wait for approval before changing it. Thus, the opportunity was in coordinating connected bookings together, rather than adding another way to chat about travel.
We also had to revise our strategic argument. I initially thought of Layla as an independent platform that could challenge established booking sites through Expedia’s inventory. Expedia’s acquisition of Layla changed that framing. The connection to supply and booking infrastructure became more plausible, but we could no longer describe Layla as an outsider challenging Expedia. Instead, our project explored how Layla could change the booking experience within an established travel company.
The part of the project where I was most involved was creating our Canva prototype. It made me turn a broad idea into screens a traveler could actually follow. I designed a route from Layla’s trip card to the booking decision, including the messages the traveler would see before agreeing with a purchase. Instead of saying that the agent would be “transparent,” I had to decide what information and what amount belonged on the screen and when it should appear. A message about a flight being £100 over budget needed to be visible before approval instead of being buried in a general description of the itinerary, but at the same time could not be too overwhelming for the travel to not demotivate them from the booking.
The prototype used green, yellow and red risk levels. A green booking matched the traveler’s requirements and needed a straightforward review and confirmation. A yellow booking called attention to a specific difference, like a new supplier or a smaller budget overrun. A red booking involved a more serious issue, such as high extra cost or a non-refundable option, and directed the traveler to review alternatives or contact a human travel expert. Each level had a different amount of explanation and review before payment.
Designing those screens revealed a tension I had not expected. Requiring approval for every booking sounds reassuring, but people may stop reading if every screen looks the same. On the other hand, an agent that moves too quickly could leave the traveler feeling that they had agreed to something they did not understand. The risk levels were our attempt to put more attention on decisions with greater consequences. They are prototype choices that would need extensive testing in a big sample instead of just a proof that we have solved approval fatigue.
I also included a confirmation and feedback stage. Once the traveler approved a simulated booking, the prototype showed that it had been confirmed, added to the trip card, and sent by email. The feedback questions asked whether the traveler felt sufficiently informed about anything that differed from their requirements. I wanted the experience to continue past the payment button because, from my hotel experience, trust depends partly on what happens after a reservation is made. People need a clear record and a way to get help if something goes wrong.
Making the prototype in Canva was useful precisely because it had interaction limits. I could demonstrate how the interaction might feel, but the screens did not make real reservations or prove that the agent could manage every supplier’s rules. The prototype helped us test whether the concept was understandable; it did not prove that the full operation would work.
Collaboration helped me move away from my first idea while still utilizing its core concepts. My interest in the social side of travel kept me interested in the human side of the booking experience – a traveler may want efficiency but they also want to feel understood and in control. Questions from teammates pushed us to explain what booking the whole trip really meant and where the traveler would make each decision. I learned that a project often improves when the team is willing to narrow an appealing idea into a problem it can explain and test.
I began with a vision of connecting travelers to one another and ended up designing how a traveler might work with an AI agent. Both ideas are about making travel feel easier but the second made me think much more carefully about business strategy. The opportunity is not only that AI can produce an itinerary quickly, but also about if a company can connect recommendations to reliable action while gaining enough trust for people to let it be in full control.