Through AI lifecycle automation, drift, and bias mitigation, Lufthansa increased customer experience and airline efficiency.

Combined talents help the airline raise efficiency

In the airline industry, timing and synchronization are everything when it comes to the customer experience. Mitigating unforeseen circumstances against customer expectations and good old supply and demand are all issues well within the wheelhouse of AI’s predictive capabilities.

It’s no wonder that Deutsche Lufthansa AG, Germany’s largest airline, recognized early on that with the right data and AI strategy, it could enhance the customer experience and better empower its employees while achieving operational excellence.

In less than two years, the airline has quickly moved from AI proof-of-concepts to scaling data science projects further into the organization, moving past constraints, such as how much test data they could include in their models. They did it thanks to a partnership with IBM that brought forth deep expertise and solutions inherent in IBM’s prescriptive method the AI ladder – together with Lufthansa’s migration of AI services to the IBM Cloud.

The rules and regulations of an airline that operates all over the world are infinitely complex—from baggage allowances for specific routes and status levels—to visa requirements for passport holders from one country traveling to any other. No agent can know all the answers.

Since early 2019, an IBM  has been collaborating daily with Lufthansa employees – quickly testing and piloting new AI-based business ideas and services. The Lufthansa AI Studio’s first project integrated IBM Watson products, including Watson Assistant and Watson Explorer in the Service Help Centre. Previously disparate data sources are now searchable in natural language and aviation terms to more easily address close to 100,000 customer queries annually. Watson manages, searches, analyzes and interprets the various relevant and connected data sources, such as Microsoft SharePoint and internal ticket systems.

The rise of a modern data science platform on IBM Cloud

Once the AI Studio’s muscles started to build, the conversation at Lufthansa turned to modernizing the company’s data science platform to bring all the disparate projects under one virtual roof – boosting the cache and effectiveness of its data science group and tying their activities closer to the needs of the business.

Data scientists and data engineers often struggle spending too much time maintaining their projects and not enough time on proving their business value. At Lufthansa, all of the above was true, and it was also compounded by limited scalability, lack of access to public software updates, plus a need for security improvements. What they needed was a tool inside the data science pipeline to monitor, build and scale models. IBM joined the Lufthansa team in a two-day Design Thinking Workshop to build out a data science platform that would offer a single environment where data scientists could experiment with new techniques, and quickly roll out models with monitoring and modeling already in place.

Over a 10-week engagement, IBM set up a new operational workflow to support the development of new data science projects using Watson Studio and Watson Machine Learning to create an open platform on a public cloud using PaaS and SaaS. This gave Lufthansa scalability and flexibility to handle mission critical workloads and accelerate the deployment of those projects in production. Lufthansa Data Scientists worked with IBM to prototype three use cases to help the airline run smarter and more efficiently – helping avoid delays, better predict boarding time and avoiding long queues at check-in counters.

The Lufthansa data science team can now develop new use cases in Watson Studio, while making improvements to the old ones. Aside from the three use cases, Lufthansa data scientists can now push out other projects – mostly to further increase passenger experience or to support operational or strategic decisions from employees.

The data science platform allows data scientists to work with new data sources. Or, by virtue of being open source, they can work more collaboratively or in their preferred language – or take advantage of other data science capabilities in IBM Watson Studio such as AutoAI and Watson Machine Learning for model development and deployment.