Language models based on generative AI are revolutionizing our everyday life at a rapid pace. The language model "ChatGPT" (Chat Generative Pre-trained Transformer) is booming worldwide. The release of subsequent versions of the model has not only improved the quality enormously, but also generated an infinite number of new possibilities for products and services. The Lufthansa Group is also experimenting with generative AI in various areas, for instance:
Apply Natural Language Processing for replacement, maintenance or repair of the aircraft
Our data analytics experts have developed a NLP prototype for the aircraft maintenance manual for Lufthansa Technik to solve the problems caused by manual references in the Aircraft Maintenance Manual (AMM) and defect reports. Automated manual referencing between AMM tasks makes engine maintenance faster and more efficient.
Eurowings Holidays AI-powered travel advisor
In February 2024, Eurowings Holidays, the tour operator brand of Eurowings, put a travel advisor with artificial intelligence (AI) called "Holly" into operation. It is the first online travel advisor that not only provides travel ideas tailored towards individual preferences within a very short time, but also enables the immediate booking of complete package holidays via the Eurowings Holidays website in the same step – a first in German tourism. The AI tool takes personal preferences and special requests such as food preferences, sporting activities, excursion ideas or physical limitations into account right from the start and compares them with billions of flight and hotel combinations. The Eurowings Holidays AI travel advisor also takes, for example, hotel reviews into account, and immediately checks whether suitable adventure excursions or car hire offers are available. The AI tool analyses billions of flight and hotel combinations to find exactly the travel package that best suits the preferences entered by the customer.
Personlized, intelligent packing list generator
Our Lufthansa Group Digital Hangar team has developed a ChatGPT-powered packing list generator prototype that creates personalized packing list suggestions based on a specific flight, enriching the Lufthansa Group travel experience with personalized suggestions and reducing the feeling of stress during the trip. It also improves the relevance of content based on variables such as destination, season and planned travel activities and promotes conscious decision making as it helps our travelers not to overpack or underpack, reducing weight and thus CO2 emissions and one-way purchases.
NOTAMIQ (Notice to Air Missions)
Our Lufthansa Systems team is currently programming a MVP to optimize and standardize our flight planning data such as NOTAMs (NOTICES TO AIRMISSION = real-time aeronautical messages) using AI. There is an increasing number of NOTAMs per year, and NOTAMs are being processed mostly manually by NOTAM officers. This consumes lots of resources and is a repetitive task. During the ChatGPT Hackathon challenge it was identified that using a Large Language Model (LLM) and potentially Supervised Learning to read NOTAM text and translate it into a machine-readable formats will make the entire process way more efficient.
Smart Documentation
A team at Lufthansa Cargo and Lufthansa Industry Solutions is currently implementing a ChatGPT bot to support our Lufthansa Cargo service, booking and sales service center in processing complex bookings (such as pharmaceuticals, healthcare, perishables or animals) by quickly finding the right information in the countless manuals. As a next step the solution will be to roll out for our B2C service centers.
As Innovation & Data Management in the Lufthansa Group, we want to facilitate the use of generative AI such as ChatGPT and promote Group-wide exchange and observe the new possibilities and use cases very closely - as well as the risks. Therefore, we have created guidelines with all stakeholders on how ChatGPT can be safely integrated into daily work, have established a ChatGPT Team, and regularly conduct exchange formats.
Santiago Erroz Ferrer
Data Science Technical Lead, Swiss International Air Lines
Future Quick Reference Handbook for Pilots
As of today, information concerning flight procedures, aircraft status and limitations as well as further technical background is accessed via multiple sources and tools (e.g. digitized manuals, electronic flight bag applications). The project “Future Quick Reference Handbook” assists pilots in quickly getting relevant, context-related support with the help of Artificial Intelligence (AI) based on a Large Language Model (LLM). The LLM is built upon the existing GPT model and enhanced via a context layer containing specific, relevant Lufthansa Group and Original Equipment Manufacture data. The prototype is accessible via voice and chat interfaces and enables further AI-based use cases and the development of products that can be used in training and flight operations.
Marketing Texts
A team at SWISS asked ChatGPT to create marketing texts to introduce new destinations to passengers and whet their appetite for travel, which took about 5 minutes. At the same time, they produced their own content in about 120 minutes and measured time savings and text quality. Both productions were sent to end customers via the SWISS newsletter and it turned out that the ChatGPT newsletter had a slightly higher click-through rate (+0.5%). Finding: ChatGPT can minimize the time spent on content creation in particular, but it still lags behind on brand tonality which was reworked manually in the test, as well as checks for text quality (e.g. word repetition). It is likely though that these problems can be trained out in subsequent tries, as the Lufthansa Group trains models on own data and incorporates brand identity documents into future trials.
Chatbot Elisa
Lufthansa Systems and zeroG support the Digital Hangar team in evaluating how ChatGPT could support the chatbot Elisa. The chatbot can answer customer queries according to templates written by humans. Through ChatGPT, it could deliver personalised offers.