How Virgin Atlantic is hitting new heights in generative AI with Databricks technology

Virgin Atlantic is implementing Databricks technology as part of its growing focus on data analytics and AI capabilities. The airline is using the platform to support generative AI applications aimed at improving operational efficiency and accelerating decision-making processes across multiple departments. This move comes as more companies in the travel sector look to data-driven solutions to address industry challenges following years of pandemic-related disruption.

Richard Masters, VP of Data and AI at Virgin Atlantic, explains at Databricks' London HQ how his organization is using the tech company's data platform to help the airline undertake pioneering explorations in Artificial Intelligence (AI).

Masters was Virgin's Data Science Manager between 2018 and 2021, which was when the company implemented Databricks. When he returned to the airline in late 2023 after a spell with consultant EY, the executive team was eager to exploit AI and data, which led Masters to dive into the Databricks platform.

Rather than starting from scratch, Masters recognized that Databricks' capabilities in handling metadata and scaling compute resources would allow Virgin Atlantic to maximize its existing data resources. He saw enormous potential for both generative AI (gen AI) and traditional Machine Learning (ML) applications if they could properly structure their data assets.

He explains the rise of gen AI from late 2022 demonstrated the potential to deploy interfaces that could answer customer and line-of-business queries in new ways. Executives were eager to see how these interfaces could be connected to Virgin's existing data sources – and that's where Databricks, particularly the Unity Catalog, plays a crucial role:

When there's something that needs to be combined and a decision is going to be made, the transformation goes through my team and, ultimately, through Databricks. Nearly everything goes through the Unity Catalog, and more and more data will go in there because the technology gives us a single pane of glass.

Databricks forms part of a larger digital stack. Virgin's back-office functions are based around Microsoft's enterprise applications, and the organization uses AWS for its retail and website environments. Masters comments the integrated nature of Databricks and its Unity Catalog meant it was an easy decision to use the data platform as Virgin scaled up its AI efforts, especially when presenting the business case to the C-suite executives.

Taking an orchestrated approach

Masters describes how Databricks helps Virgin roll out a range of gen AI initiatives, particularly for summarization and categorization. He gives the example of how the Health and Safety team uses the technology, including via Databricks' gen AI interface, Genie:

When you're in operations, at the airport or on a plane, you have to report on incidents. However, that culture drives a volume of messages. We're using AI to help the Health and Safety team focus on the most important incidents versus having to go through a whole list every day.

This proactive approach is used as a model to sift and prioritize questions in other areas of the business. According to Masters, the airline is applying similar techniques to customer care messages, allowing them to process them more efficiently. The solution identifies the core issue in each message, suggests potential solutions, and determines whether customer interaction is needed. Crucially, this capability can be reused across different departments, eliminating the need to build separate summarization mechanisms for each business function.

Masters asserts that one of the other great things about this orchestrated approach is that staff have the flexibility to run a range of AI models on the data platform. Depending on the complexity, the business requirements and costs, employees select first-party models or finely tuned, open-source models:

You don't have to commit. You can pick the right models for the use case. People come to me and explain why they want to use a different model. We can sign that decision off, and it's governed through the orchestrating platform. So, Databricks helps us govern, monitor, and pick the right models.

Looking for new opportunities

While Masters was unable to quantify the impact of AI initiatives on the business, he did point to some benefits. He gave the example of Fetcherr, an AI-enabled tool that helps Virgin with dynamic pricing and is fed by integrated data sources in the Databricks platform. The partnership is delivering a "net-positive impact on revenue". More generally, he suggests the smarter, AI-enabled way of working across the business produces other benefits:

We're making decisions and optimizations quicker. So, some of our Finance processes for reconciliation might have taken weeks to months before. Now, with the data platform, we've had people making decisions in a couple of days or even a few hours. We can manage our cash flow better and our relationships with vendors.

Masters warns one of the most significant cultural challenges associated with a shift to AI is ensuring people can make the most of the insight. He says Databricks' attempts to refine its user interface will be important in helping the business reduce some of the noise:

All the tables are great for analysts, but the managers don't see all that data. They need to ask a question and get a clear answer. So, getting that balance right is a challenge, but it's much easier than when we had all these different ways of presenting information, whether it was via Power BI, a tablet or an Excel spreadsheet.

In the longer term, Masters envisions Databricks as the enterprise platform for further work on data, AI and agents. His ambition is to leverage the technology to move into new areas beyond data-enabled analysis and decision-making, with particular emphasis on real-time interactions for both customers and crew. He believes that utilizing Databricks' metadata capabilities and Catalog for these efforts would significantly accelerate the development of fast-moving use cases.

Masters advises other business and digital leaders who are looking at Databricks and AI to focus on business objectives:

The platforms are becoming easier to connect, so that becomes less of a focus for you as a data leader. Have conversations with the business and spend a good chunk of time articulating those priorities back to get the whole team curious about the work.