3 Use Cases of Generative AI in Travel

The travel and hospitality industry is buzzing with excitement about generative artificial intelligence (AI). It is a hot topic at conferences and industry publications and rides the Peak of Inflated Expectations on the Gartner? Hype Cycle for Emerging Technologies 1, 2023, projected to reach transformational benefit concerning mainstream adoption within two to five years.

In this article, we provide valuable insights for travel and hospitality market leaders venturing into AWS for the first time, as well as existing AWS clients seeking to harness the power of generative AI within their organizations. Furthermore, we will explore important considerations when building generative AI applications.

The Current State of AI’s Impact on the Travel Industry

AI technologies have driven the travel, hospitality, and leisure sectors forward, opening doors to new business models , customer touchpoints, and value opportunities.

AWS T&H Solutions Team has curated a comprehensive array of typical AI/ML use cases accessible via the AI Use Case Explorer , boasting over 20 cases that underscore the integration of machine learning and AI in travel and hospitality. These include personalization, AI-enabled contact centers, chatbots, and virtual assistants, and more.

To dive deeper into generative AI, PhocusWire's "AI Insights" Series features interviews that delve into current applications in leading enterprises. The discussions spotlight use cases pertinent to business travel , vacation rentals , tour operators , and travel distribution firms.

Additionally, Skift’s generative AI report in April 2023 and their innovation session in July underscore the exciting possibilities in the travel industry and generative AI.

Prominent Use Cases of Generative AI in the Travel Sector

Travel and hospitality companies are using tailored generative AI solutions for various internal and external processes. The most prominent use cases are the following:

• Customer Support: Fortified with agent desktop tools for smart customer centers.
• Personalized Offerings: To tailor services and offerings that resonate with individual preferences.
• Operational Efficiency: To streamline processes such as information extraction and unified helpdesk solutions.
• Content Customization: To personalize content and descriptions to diverse advertising goals.
• Reputation Management: To monitor sentiment and manage online reputation.
• Developer Empowerment: For boosting developer productivity and efficient creation of requirements.
• Predictive Maintenance: To prevent technical debt by enabling early issue detection and resolution.

Initiatives by DataArt's AI/ML Labs

DataArt AI/ML Labs has been driving initiatives to harness technology for data extraction, platform design, and travel-centric applications.

Concurrently, since 2021, we have actively hosted webinars highlighting the use of AI/ML for personalization, offer creation, entity extraction, and analysis:

• June 2021: The webinar “Unleash Travel Data: Breaking Down Silos and Unlocking Opportunities with The Power of AI and ML” addressed the challenges and opportunities within the travel industry, particularly in the context of AI. Real-world examples were shared, and creative use cases were brainstormed.
• October 2021: “AI, ML and Big Data for Travel and Hospitality ” webinar delved into the AI domain. As the competition for customer engagement intensifies, the webinar emphasized the need to expedite and enhance personalized offerings by leveraging existing data to its fullest potential.
• May 2022: On the webinar “How to Use AI & Automation to Better Understand Your Traveler ,” a panel of industry experts discussed the challenges of determining traveler profiles and the opportunities new AI technologies offer in supporting businesses while increasing customer satisfaction.
• June 2023: The session “Beyond the Prompt: Transforming Travel with Generative AI ” explored real-world proof of concept examples in document processing and support automation for travel businesses.

In our latest webinar, guest speaker Mark McSpadden, VP of Product at AMEX GBT, shared insights about his team's exploration of around 40 travel counselor use cases. He emphasized the urgency of embracing generative AI as quickly as possible because “the interactions of your customers with the AI model are going to be very specific to their needs and your area of expertise. What you do to get those learnings early on can make a big difference in where you spend your time and attention.”

Prototyping with Generative AI

Since December 2022, DataArt’s AI/ML Labs have focused on equipping customers to navigate the ever-evolving sphere of generative AI, facilitating hands-on experimentation with emerging technologies. In 2023, DataArt's AI/ML Labs successfully built 20 prototypes, including three tailored for the travel and hospitality industry.

Let us explore these three prototypes to illustrate the possibilities:

#1 Travel Agency Contact Center

Imagine a scenario where travel agencies and management companies handle a high volume of emails. Efficiently categorizing, prioritizing, and extracting essential traveler information from these emails demands significant time and effort, leading to slower response times and reduced customer service quality.

DataArt’s prototype streamlines this process, saving time and resources while ensuring quicker, more effective customer service.

The system extracts vital parameters from emails and categorizes them into various case types, including new bookings, modifications, cancellations, supplier communications (e.g., hotel, and airline confirmations), invoice-related queries, and trip support issues. Additionally, once the model detects the destination and the departure place, it automatically identifies the trip type based on its knowledge and provides a JSON representation of requested trip segments.

From the implementation perspective, it is straightforward:

1. Access and process email content.
2. Send email text and a prompt to a foundational model.
3. Process the model's response and pass it to the downstream systems.
4. Record original emails, parsed responses, and validated outcomes for future training.
5. Generate a draft of a response from Travel Advisor, either confirming receipt or requesting missing information to proceed with the request.

The system leverages Amazon Bedrock and the Anthropic Claude 2 model to do the processing.

#2 Travel Contract Loading

For suppliers and aggregators of hotel content, hotel contracting is crucial but complex.

Contracts typically outline details such as room types, seasonal price ranges, available discounts, minimum and maximum booking durations, allowable nights for booking, cancellation policies, and guest nationality considerations, among others.

Hotels and hotel chains use custom formats for those contracts, making standardized data extraction difficult and requiring a lot of manual labor. ?

Our prototype partially automates this process and is beginning to see positive results:

1. Leveraging Amazon Textract allows us to capture text from contracts in PDFs; it also helps to work with scans and images.
2. The extracted text is paired with a series of prompts for data extraction.
3. The resulting JSON files are then combined for further processing in the downstream systems.

We encountered several challenges with this prototype:

• Identifying the maximum number of children based on table and textual data.
• Extracting rate plans from multiple tables.
• Processing certain industry-specific terms.

We believe collecting additional data using the current process will help us fine-tune the system for more accuracy at a later stage.

#3 NDC Chatbot

New Distribution Capability (NDC) has led to complications in operations for travel agencies, travel management companies, online booking tools, and OTAs.

NDC results in content source diversification, triggering concerns across content completeness, content display, commission and markup management, mid/back-office settlement, servicing, and other areas of business.

As DataArt was preparing the NDC Adoption Guide based on our hands-on experience and research, we accumulated a lot of reference materials, which we wanted to make available to our consultants and technical delivery teams.

Leveraging Amazon Kendra and Amazon Bedrock, we built a chatbot that provides a simple question-and-answer interface to access the growing body of NDC-related research.

This is how the chatbot works:

• Our research on NDC is captured and enhanced regularly. Amazon Kendra indexes all that textual data.
• Users submit a request with a web application, and a language model then process this request to generate search queries for Amazon Kendra.
• The results from Amazon Kendra are combined with a user request and a prompt and processed by Claude 2.0 model using Amazon Bedrock.
• The response from the model is then shared with the user.
• If the user has follow-up questions, we address them within the same conversation.

This pattern is called Retrieval Augmented Generation, and it allows us to overcome a limitation of foundational models that did not have access to proprietary data when they were trained.

Learnings and Next Steps

Through the development of over 20 prototypes, we have witnessed the positive impact of generative AI on our clients' organizations.

Stakeholders gain a firsthand understanding of how this technology can transform their business while recognizing the challenges that must be addressed.

Early prototypes allow to explore fresh interaction models within user interfaces. While chatbots have been a prominent feature, we enthusiastically anticipate the emergence of other user experience patterns.

Moreover, as the landscape of libraries and solutions advances, managed options like Amazon Bedrock streamline the prototyping process, reducing timelines from months to days and weeks.

The travel and hospitality industry stands at a pivotal stage where learning, prototyping, and knowledge-sharing unlock immense business value. As an AWS Travel and Hospitality Competency Partner, we are eager to accelerate industry adoption of those technologies. Join us on this discovery and knowledge journey. Let’s schedule a session with the DataArt AI/ML Labs team and learn more about what we have built and the patterns we have detected!

1 - Gartner, Hype Cycle for Emerging Technologies, 2023, By Arun Chandrasekaran, Melissa Davis, Published 2 August 2023 GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.

Originally published here .