At Sabre, as we evaluate new opportunities for leveraging Artificial Intelligence (AI) and Machine Learning (ML) algorithms, we realize that a fundamental challenge that needs to be addressed is how to scale this specialized knowledge across the organization. Central to realizing scale across the organization is to raise awareness of the role of AI, communicate where it can be used, encourage teams to be curious, learn and leverage our extensive data sources and bring forward new travel-related value propositions and use-cases that add value to our customers.
There are two key challenges with leveraging AI/ML in any enterprise. First is the skill to identify use-cases and areas where AI inspired algorithms can be leveraged to create a new solution or enhance an existing solution. Second, is the ability to scale the adoption of AI/ML across the organization. We are currently addressing these two challenges with the following initiatives.
AI Special Interest Group (AISIG)
We established an Artificial Intelligence Special Interest Group (AISIG) in January 2018. This AISIG today has over 400 members globally across Sabre, who have expressed an interest in AI / ML, from business units and technology with representation from product marketing, product management, product development, data analytics and research. The AISIG community uses group discussions and periodic notifications to stay informed about the latest developments in AI, Sabre priorities and use cases. In the AISIG microsite, innovations in ML/AI algorithms, industry innovations and academic research papers are reviewed and debated.
AISIG Town Hall Meetings and Hackathons
We also established a quarterly cadence for town hall meetings to discuss industry trends, Sabre use cases in action and provide transparency into prioritized initiatives across the company, as well as invite external speakers to talk about advances in commercial tools and state-of-the-art ML research. Besides the Town Halls we also established and supported hackathons that explored particular use-cases, and applicability of various ML techniques to such real travel-industry use-cases.
Over 550 employees have gone through our training class titled “Introduction to AI/ML at Sabre”. The introductory class provides a business-oriented overview of artificial intelligence (AI) and machine learning (ML) methods with examples drawn from the Sabre travel ecosystem. It is intended for individuals who are not familiar with AI/ML methods but are interested in learning the key concepts and the application of these technologies. The objective of this class is to help instill an intuitive understanding of what types of applications can (or cannot) benefit from AI.
Technical content covered in this course included:
Supervised Learning: Classification (Support Vector Machines, Discriminant Analysis, Naïve Bayes, Nearest Neighbor) and Regression (Linear, Generalized Linear Model (GLM), Decision Trees, Ensemble Methods, Neural Networks).
Reinforcement learning: Genetic Algorithms, Multi-armed Bandit test and learn, Approximate Dynamic Programming, Markov Decision Processes.
Unsupervised Learning: Clustering (k Means, k Medians, Fuzzy cMeans, Hierarchical), Gaussian Mixture, Hidden Markov Model.
Deep Learning: A brief introduction into Convolutional Neural Networks (CNN) and Recurrent Neural Networks (CNN) with sample use cases.
We have found that such introductory training is critical toward stimulating interest across a broad class of developers in our company. Developers that take this course have consistently rated such training at 4.5 out of a maximum score of 5.
Many vendors provide “black-box” toolkits for AI/ML. We evaluated toolkits, including cloud-based solutions, proprietary solutions, and open source solutions (in-house and cloud) to enable data science and development teams with AI tools and capabilities. The exercise was to provide consulting support to teams who are developing AI/ML applications and have a preferred toolkit. The evaluation of AI toolkits used the following criteria: data ingestion and connectors, data wrangling and visualization, flexible modeling capabilities and deployment capabilities that support automation. Further, given the mix of employees at Sabre we had to factor in citizen data scientists, who prefer drag and drop capabilities, versus technical data scientists who prefer to write code and build models using Python, R, Scala or other programming languages.
The key to Sabre’s AI Initiative is to raise awareness and imbed a “sense and respond” mindset among employees to enable rapid adoption of AI Technology. In travel, it is difficult to predict which companies will capitalize on the AI-enabled landscape and be successful. A critical success factor will be the willingness of managers and individual contributors to be nimble, establish success criteria, run experiments, evaluate results quickly and move an application to the next stage with a fail fast attitude. We anticipate that these small fundamental steps accelerate awareness and urgency for adoption of AI across the organization for competitive advantage in the marketplace. At Sabre, this journey has begun and is continually evolving.