With 100,000 customers carried every day on board more than 1,000 flights, Air France’s activity generates a considerable volume of data. Making optimum use of this resource to constantly improve performance is therefore a strategic challenge for the airline, combined with an absolute priority, which is to protect the company and its customers’ data.
This is not a new topic. In 1958, Air France set up an operational research department, in charge of promoting and structuring innovation within the company. The aim was to adapt it to the technological developments of a constantly evolving industry. This stream of information or “data” was rapidly identified as a valuable asset, enabling us to better understand customer behaviour. This groundwork notably led to the expansion of revenue management in the 90s, partially drawing on historical data and identification of trends.
In the early 2000s, this approach of using large volumes of data for scenario planning was applied to the maintenance business with the development of a so-called predictive maintenance solution, Prognos. With this tool, which is now used by over 80 airlines worldwide, Air France entered the era of predictive and prescriptive artificial intelligence.
Since then, artificial intelligence has occupied a natural place in all the Air France research and innovation programmes, and is currently being used in the different stages of the customer journey. With diverse uses including chatbots, tools for predicting the number of bags and meals on board, calculating the amount of water to take on board or eco-piloting systems for optimising in-flight trajectories to reduce fuel consumption, AI is used to optimise activity and resources, anticipate needs and make it easier for staff to access information relevant to their job. And all of these uses underpin a central goal – improving the customer experience.
2023, the turn of generative artificial intelligence
Since 2023, Air France has been engaged in a new revolution around generative AI. A new type of artificial intelligence capable of generating rich content autonomously. Unlike traditional artificial intelligence, which focuses on specific tasks such as data classification, generative AI actually generates new data, and relies to a large extent on machine-learning. In short, AI feeds on and learns from all interactions. Made popular by the ChatGPT tool, generative AI marks a real technological breakthrough and opens up new horizons for using data.
Today, more than 80 projects involving the use of generative artificial intelligence have been launched across all Air France business sectors. Some are at the stage of identifying the most appropriate solution (data management model, predictive AI, generative AI, etc.), while others have reached the Proof of Concept stage. Here are 4 examples illustrating Air France’s developments in the area of AI –
- TALIA: Air France’s internal ChatGPT, enabling employees to familiarize themselves with the functioning of this tool in a secured environment, without any information being passed on to third parties. TALIA is used daily by Air France employees to write e-mails, search for information in PDF documents, organise events or build to-do list.
- PAMELIA: a solution allowing Air France airport agents to obtain answers to customers’ queries directly on their iPad. Number of bags allowed, transport of live animals, formalities: PAMELIA searches for the answer in the airline’s reference manuals and procedure documentation and generates a written response ready to be shared with the customer, which can be instantly translated into 85 languages. Currently in the test phase, PAMELIA will be deployed at Paris-Charles de Gaulle in 2025.
- CHARLIE: a tool for the airline’s maintenance teams, enabling them to search for aircraft part numbers in the airline and manufacturers’ documentation. This tool saves precious time when repairing or replacing parts, thus contributing to flight punctuality.
- FOX: a tool for analysing customer feedback, to have a better understanding of their expectations and concerns. Using generative AI, FOX is capable of automatically analysing customer feedback and translating the gist of diverse and complex texts incorporating humour or irony. This tool can also be used to detect weak signals and share this feedback throughout the different company entities.
Air France’s approach to the use of artificial intelligence is based on several principles, which are regularly updated in line with technological advances:
A reasonable and controlled use of artificial intelligence, with priority given to the protection of customer and company data, respect for rules and ethical principles, and strict compliance with regulations. This means keeping the solutions used within a closed circuit, and setting up an AI Committee to ensure that all these principles are properly followed.
Promoting existing solutions and supporting initiatives within the company. The operational research team, which now has 150 employees, supports the different business sectors as they explore the possibilities offered by artificial intelligence. According to the needs expressed, the operational research department can guide teams towards a more or less complex solution, support them in the implementation of a Proof of Concept (POC), and participate in the implementation of the projects selected.
Using market solutions rather than designing models inside the company. Artificial intelligence, particularly generative AI, is still a relatively new field of expertise and the technologies are not yet mature. This can make them obsolete relatively quickly. Air France therefore relies on existing solutions and models rather than seeking to develop its own from scratch. This approach notably also helps us to control investment and retain our agility.