TripGenie: The Future Of AI-Driven Travel Itineraries

Given its widespread adoption, AI is taking over travel! From personalized recommendations to automated customer service, it is transforming the way we plan, book, and experience our journeys. One of the pioneers in this AI-driven transformation is Trip.com, with its advanced AI assistant, TripGenie, which is incorporated into their mobile app, leveraging on Large Language Models (LLMs).

This blog featuring the insights of Amy Wei, Senior Director of Product Management, Trip.com Group and Olivier Jolidon, Global Head of CRM, Insights & Analytics, AMAN discusses the impact of AI on user experience, business value, and the broader travel industry through their strategic use in itinerary building and personalised recommendations. Additionally, they address the importance of maintaining a human touch and the emerging risks of bias in AI technologies.

TripGenie: A Case Study

TripGenie is Trip.com's AI-powered itinerary generation feature. It personalizes travel plans based on user preferences, interests, and travel style.

Impact on User Experience:

The introduction of TripGenie's itinerary generation feature has significantly enhanced user engagement. Trip.com reported that users utilising this feature spend an average of 20 more minutes on the app compared to the general user base, indicating deeper engagement and satisfaction. This result highlights TripGenie's effectiveness in enhancing the user experience, while also providing marketers with valuable insights into user behaviour.

Business Value:

As Amy emphasizes, "TripGenie's primary goal is to provide users with high-quality itineraries that meet their needs."  Their winning strategy rests on user-centricity and providing personalized recommendations. As TripGenie is embedded as an AI-driven feature that sits on the Trip.com app, and as an OTA, they facilitate bookings to hotels, flights, and other services. Their commercial strategy is not driven by the monetisation of itinerary planning. Instead, those recommendations that have been generated are designed to serve users' best interests, ensuring a seamless and enjoyable travel experience.

Challenges and Opportunities

Implementing AI technologies presents several challenges, including integration issues, skills gaps, regulatory concerns or strategic focus. Olivier points out that human interaction remains indispensable, particularly in the ultra-luxury sector. Therefore, we need to ensure that internal teams work alongside AI as a collaborative tool in helping them gain information they need. In other words, the goal is find a balance - enhancing efficiency through AI whilst avoiding overreliance and losing essential human connections.

The Problem of Hallucination

At Trip.com, Amy explains that they have taken several measures to addressthe issue of hallucination in AI responses. The underlying principle is to ensure that the LLM refrains from answering questions it cannot handle accurately. In such cases, it will respond with something like 'I'm sorry' and seek further clarifications rather than providing misleading information.

Additionally, they have integrated Retrieval-Augmented Generation (RAG) technology, linking the LLM to internal knowledge bases. This helps ensure that responses are based on accurate, up-to-date data, significantly preventing the spread of falsified information.

At AMAN, Olivier highlights their focus on: Fairness, Privacy, Performance and Security. These underlying principles help them to define, reinforce control as well ensure high quality. This is critical when it comes ensuring that there is low to no risks in terms of hallucination. For example, fairness is used to gauge the quality of their answers and reinforce systems based on human-feedback. As all project, the risks need to be measured, controlled and addressed.

Quality of AI-Generated Content

Trip.com adheres to strict quality standards and a comprehensive review process for its AI-powered itinerary builders, evaluating itineraries based on three key criteria: Reasonableness (appropriateness for the season or other factors), Richness diversity of destinations, including dining, accommodation, and transportation), and Customization (tailoring to specific user preferences, such as family-friendly or culturally rich experiences).

Furthermore, they involve top-rated local travel guides from around the world in the loop during the review process to ensure that the itineraries meet high standards and are truly beneficial for their users.

Results unveiled that TripGenie's itinerary planning outperforms the rest in terms of richness and quality in recommendations as well as customisation. They achieved the highest average rating compared to OpenAI ChatGPT 4/3.5 and Google Bard (now Gemini), which were scored by industry experts in a blind test.

Our Contributors:

Amy Wei, Senior Director of Product Management, Trip.com Group

Olivier Jolidon, Global Head of CRM, Insights & Analytics, AMAN