How i built an AI-Powered travel app that makes research and planning fast and fun

After using Google's MyMaps to plan all my trip for years, i was very frustrated with its limitations. So i wanted to build my own app that lifts the limitations and adds AI-Powered research capabilities.

Along with the traditional user interviews I started building the app itself, beginning with a brain-dumped concept of what an MVP needs. Its first versions tested the riskiest assumptions: how precisely free services place pins, how good the research results are and how good the image sources are. I wrote down what I learned. Later I combined the interview insights, those learnings and the initial concept into a new concept. The clickable prototype grew out of it, and the app now catches up with what the prototype added.

Testing the feasibility of the concept live

In the first version of the app I tested the riskiest assumptions: how precisely free services place pins, how good the quality of the built-in research could be and where to source the best images.

Research & Discovery

I interviewed travelers aged 40 to 56 who plan their own trips, and I treated my own interviews as one more participant. They covered how people find places, how they decide, and where their tools get in the way. The results hold for experienced, self-organized planners only. In parallel I built a test app. Its main job was to investigate technical limits: how good built-in travel research is, whether every pin can get exact coordinates, and whether good photos can be found.

The insight

Planning is a funnel from many finds to a shortlist to day-sized routes, and each step lives in a different tool.

What was made and why

I built Vayo along that funnel: find, shortlist, plan the days.

Finding: free questions instead of fixed categories

I started with a batch tool that searches fixed categories by region. On paper it made sense. In the test app I saw that choosing from fixed categories can make you miss good spots, and in the interviews nobody described searching from a fixed list. So I built the research as free questions to the AI. The results land on the map right away, and I decide on each one.

Shortlisting: keep or skip first, ratings later

In the interviews, one traveler said he wanted to go through spots quickly, like swiping: like or dislike. So every find gets a fast keep or skip first, and ratings come later, once a shortlist exists. The alternative was rating every place the moment it appeared. Rating pins at all is a big step up from MyMaps, which has nothing like it, and it turns the kept finds into a shortlist I can build days from.

Planning the days: from a list of stops to a day that adds up

In MyMaps, a route was a list of stops. In Vayo, drive times between places show instantly and add up per day and for the whole trip. I added a start time for each day, so I can see whether a day can be done while the sun is still up. For that I needed the time spent at each place. Google's API has it, but it costs extra. Then I had the idea to ask the AI that was already connected to the app. It finds the time at no extra cost, and I can still change it.

Outcome

Planning a trip like this used to take me 2 to 3 months. With Vayo it took 2 to 3 weeks, and I planned a full 4-week trip with it. So far only I have used it, so I can't yet say how it works for anyone else.

If I did it again, I would start discovery from what I already knew. I had years of my own trips and years of conversations with friends about how they plan, and none of that went into the research. I treated the topic as unknown, so the discovery phase repeated things I could have started from.

Related projects