樓毚排

Sabre has a long history of developing data-driven solutions and has already introduced, in partnership with Google, Sabre Travel AI, a technology that learns continuously from consumer behavior. This article continues that tradition as we explore how recently released innovations in Large Language Models can be targeted toward applications and use cases in travel.

At a recent Generative AI hackathon organized by 樓毚排 and Google Cloud, a group of Google, Infosys and Sabre developers and engineers came together at the Google offices in Addison, Texas, to explore the potential of artificial intelligence in solving real-life business problems. Over the course of two days, the team worked on three distinct use cases, each presenting unique challenges and opportunities for innovation. The hackathon resulted in the development of proofs of concept (POC) that targeted improving customer service, increasing revenue, and overall efficiency.

In this article, we will delve into each of the three use cases tackled during the hackathon, highlighting the key objectives, methodologies, and outcomes of each project. From improving customer care through natural language search to automating travel request processing and generating structured data from unstructured text, the teams showcased the power of generative AI in driving transformative solutions.

Use case 1: improving self-service in customer care and support 

The first use case focused on enhancing Sabre’s internal call center’s natural language search capabilities. Sabre Travel Solutions Care organization provides support to customers worldwide, handling hundreds of thousands of cases annually in multiple languages and assisting travel agents and airline employees in navigating Sabre’s travel reservations system and other products.

The teams self-service knowledgebase of how-to articles, known as Finder, plays a crucial role in addressing customer queries and issues. However, the existing keyword-based search system sometimes misses the most relevant knowledge article, leading to extra clicks and frustration as customers look for the right content. Additionally, the articles returned are not always ranked based on the best relevance, further hindering the resolution process.

During the hackathon, the team focused on improving case deflection by training a Large Language Model (LLM) on the Finder knowledge base. The goal was to provide a natural language interface for customers to interact more efficiently, enhance search results by prioritizing the most relevant articles, and highlight specific areas within articles to reduce the need for extensive review. The resulting solution has the potential to help customers to find answers to their questions more efficiently with fewer escalations to support agents, improving customer satisfaction by reducing the effort for the customer to get the right answer, and delivering cost savings for Sabre.

Use case 2: delivering trip proposals via email

The second use case focused on automating the processing of travel requests received via email. With an increasing number of contact center transactions occurring through text-based channels, such as email and messaging, there is a need for efficient interpretation and response to these requests. The team aimed to leverage Google’s Vertex AI and Large Language Models to convert unstructured travel intent into specific travel API requests to 樓毚排 and generate valid offers for air, hotel, and other travel components.

The process involved using a service that encapsulates Vertex AI and subsequent travel API calls.  The email text is processed based on the Large Language Model’s understanding of traveler intent to make calls to Sabres shopping APIs and generate valid trips as requested in the email. The solution generates legitimate offers for air and hotel, and potentially other elements like rental cars and activities. By providing an intuitive display of the offers, and responding via automated emails, the team aimed to increase agent efficiency and streamline the review process for both agents and travelers.

Sabre provided travel shopping APIs such as for air travel shopping and for hotel shopping. The solution developed during the hackathon utilizes generative AI to transform unstructured email content into structured trip bookings, simplifying the workflow for agents and improving response time for travelers.

Use case 3: simplifying automated refunds

The challenge in the third use case revolved around enabling automatic refunds for airline passengers who want to change or cancel their flights. Airlines have specific refund rules for these voluntary changes, including ATPCO Fare Categories 31 changes and 33 refunds (CAT-31/33) and Category 16 penalties (CAT-16). CAT-31/33 follow structured rule formats to provide for automated reissuing, while CAT-16 involves unstructured text descriptions, which vary across different airlines. As a result, processing refunds based on CAT-16 rules often requires manual intervention.

The project aimed to enable automatic refunds by transforming unstructured text (CAT-16) into structured refund information (like CAT-31/33). The objective was to provide accurate interpretations of rules, interpreted data, and an automatic refund/penalty calculation engine. The solution (which was partially successful) involved developing a system that could understand and interpret the diverse rules described differently by different airlines. By leveraging generative AI, the team made progress towards a simplified and automated refund process, which would reduce manual efforts and enhance efficiency.

Key take-aways

The Sabre/Google hackathon demonstrated the power of generative AI in addressing business challenges and showcased the technologys potential for enabling transformative solutions. Through the collaborative efforts of talented engineers and experts, three innovative prototypes were developed, addressing critical areas of customer care, travel bookings, and structured data generation. These solutions have the potential to enhance customer satisfaction, increase revenue, and drive operational efficiency.

With the success of the hackathon, Sabre is looking into how to transition these prototypes into production, enabling the implementation of tailored, efficient solutions for our customers. Sabre recognizes the value of partnership and will continue to explore opportunities to collaborate with Google and other industry leaders in harnessing the potential of generative AI techniques, artificial intelligence and machine learning.

About the Authors

Richard Ratliff, Executive Scientist at Sabre Labs, has held roles of increasing responsibility in IT at Sabre for over 30 years with a focus on travel retailing and revenue management.

Peter Melcher is part of Sabre’s global communications team.