How AI Travel Tools Transform Trip Planning
Based on an interview with Shie Gabbai
Most travelers enjoy dreaming about their next trip, but the planning process often creates frustration that can sometimes prevent the trip from ever happening. Roam Around founder Shie Gabbai saw this firsthand and began studying how people actually plan vacations. “In 2013, the average traveler used 38 websites to plan a single trip,” he says. Things have only become more complicated for travelers. “When I looked into it more recently, it was 243 pages,” Gabbai adds. This rising complexity became the starting point for Gabbai’s work on artificial intelligence (AI)-powered travel planning. His aim was to build a system that does more than provide background information, but acts as a true travel planning assistant.
Shie Gabbai is a product leader and entrepreneur focused on applying AI to real-world consumer problems. He founded Roam Around, one of the first GPT-based travel planners, where he developed tools that turned simple conversations into bookable itineraries. His work centers on improving accuracy, reducing friction in trip planning, and making AI more useful and accessible for everyday travelers.
Gabbai began by using ChatGPT to plan a trip to Barcelona. “ChatGPT told me about Barcelona’s architecture,” he says, “but I still had to figure out how to book tickets and plan logistics. It was informational, not actionable. I wanted a system that helped you actually accomplish the task.” At the same time, early general-purpose AI tools often returned suggestions that were incomplete or unreliable in a travel context. They could surface ideas, but they did not check opening hours, availability, or basic safety. That mix of high effort, partial information, and occasional inaccuracy is what pushed him toward creating Roam Around as a travel-specific AI planner.
Before starting Roam Around, Gabbai spent two years helping build the consumer-focused AI travel app Layla,1 where he refined many of the ideas that shaped his approach to conversational planning. He recently joined Simplicity2 to bring similar AI-driven planning capabilities to enterprise clients.
Turning Recommendations into Real Plans
Gabbai’s approach was to connect AI reasoning tools to real-time travel inventory.
“We made sure everything the AI recommended was bookable,” he says. “If it suggested a hotel, a guided tour, or a flight, it was in stock. Many sites send you to links that are sold out or not relevant. We filtered everything first.”
This structure allows travelers to describe the trip they want, then receive a complete plan they can book in a few clicks, without misleading information about unavailable travel options. “It cut the process from weeks to minutes,” he says.
A Simple Interface for Everyday Travelers
A major goal in this space is making AI accessible to average users. According to Gabbai, the interface matters as much as the technology. “One thing that helps is giving the AI a personality,” he says. “People open up more when it feels like they are talking to someone friendly. They share details they would never share with a human agent.”
Users often begin with personal context. As Gabbai found, “They say things like, ‘I am traveling with my wife. We have not traveled in 10 years. We want time to recharge.’ They do not do that with a generic system. They do it with something that feels approachable.” The Roam Around system responds with clear reasons for its recommendations. “We explain why a specific hotel or location fits what they told us. That consistency builds trust,” he says.
Designing Around Real Behavior
Building personalization into a conversation is not simple. Many companies try to do it through quizzes and forms, but Gabbai says that most people will not complete them. “Ninety percent of users will not fill out a quiz,” he says. “So, you have to gather context from the conversation itself.”
His method is to start with an example itinerary, then invite the traveler to refine it. “If you start with a blank screen, users leave. If you show a sample plan and offer to adjust it based on their needs, they stay. It shows value immediately.”
Improving Accuracy and Reducing Errors
Early in development, accuracy was a major issue. “When we launched in 2023, hallucinations were a serious problem,” Gabbai says. “The model once recommended swimming in a lake that was crocodile-infested.” Those early errors reinforced his view that a dedicated travel system needed to handle accuracy differently from a general chat tool.
Model quality has improved, but the more important change was limiting the AI to verified data. “Before the system plans anything, we ask partners like Booking.com, GetYourGuide, and Viator what is available for the selected dates,” he explains. “Then the AI builds the plan from that smaller set of trusted options. This reduces errors in a meaningful way.”
He notes that no travel system is perfect. “Even human agents make mistakes. The goal is to be reliable, consistent, and transparent.”
When AI Helps People Move Faster
Some users rely on the AI system for complex trips. One traveler from South Africa used the system to plan a 28-day, ten-city tour of Europe.
“He booked 28 nights of hotels,” Gabbai says. “The entire conversation took 12 minutes. A human agent would have needed weeks.”
Other use cases are more personal. Gabbai shares, “One traveler described wanting to reconnect with his wife after not traveling for ten years. The AI recommended several locations and explained the reasons for each.”
What stood out to Gabbai was not the booking itself but the shift it enabled. “The moment [they] had a capable, attentive assistant guiding [them] through the options, [they] finally booked a vacation,” he says. “That is the real value. It is not about extracting more spend. It is about helping people do something they would not have been able to do on their own.”
How Humans Still Add Value
Although many itineraries are generated without human involvement, there are cases where a human review improves the outcome.
“The travel industry is fragmented,” Gabbai says. “Sometimes a hotel and a booking site do not sync correctly. A quick human review can confirm that the traveler will get the room they expect, or even help request an upgrade. It adds confidence to the process.”
Interestingly, many human travel agents also use the AI tools. “Agents used it to research destinations they did not know well. The AI handled the heavy research, and they added the personal touch.”
The Future: A Travel Agent in Everyone’s Pocket
Gabbai believes the next phase of AI travel tools will feel more like a personal assistant than a one-time planner. “If you look at history, only wealthy people had chauffeurs,” he says. “Today, everyone effectively has one through Uber. The same pattern will happen with travel planning.”
His long-term vision is a persistent travel companion that learns user preferences over time. “If you interact with the same assistant for restaurant suggestions or weekend plans, it starts to understand you,” he says. “Then, when you ask about a big trip, it already knows your pace, your budget, your family situation, and what you enjoy.”
The key to this future is frequent, low-stakes interaction. “If you only use it twice a year, every trip is a cold start,” he says. “But if you use it for everyday decisions, planning a full trip becomes much easier.”
Conclusion
Gabbai sees travel as an ideal example of how AI can solve real problems for everyday scenarios. “We are moving from AI that impresses people to AI that helps them,” he says. “It is not about futuristic technology. It is about usefulness.”
For travelers, that usefulness means fewer tabs, fewer spreadsheets, and a clearer path from idea to action. “If AI removes the stress, people can focus on the parts of travel that matter,” Gabbai says. “That is the goal.”