Virgin Atlantic kept its AI infrastructure simple by building on two core partnerships: OpenAI and Databricks. Every team shared the same data, context and skills. Richard Masters, its former VP Data & AI, explained how at aiblLIVE London 2026.
The same platform powers internal agents and the customer AI Concierge in the Virgin Atlantic app. It also lets business teams build and run their own agents, under central controls. Here is how it worked.
All figures as stated on stage by Richard Masters.
It consolidated on OpenAI and Databricks. ChatGPT Enterprise, Workspace Agents and the Codex app all connect to Databricks, the core data platform, through shared skills and plugins. Digital, tech and knowledge teams work on one infrastructure. Skills are built centrally, synced to the identity platform and distributed across the organisation.
Its mistakes pointed to bad source content. The AI Concierge got answers wrong at first. That exposed website content that had been wrong for ages, including aircraft the airline did not fly. Business users also improved metadata on Databricks, such as what a passenger, customer or guest means.
It was highlighting content on our website that had been wrong for ages.
Richard Masters, Former VP Data & AI, Virgin Atlantic
Only the people who need it. Of about 10,000 staff, 1,500 were given AI tools. Pilots, engineers and other highly regulated roles got specific tools with more evaluation, such as a health and safety triage agent. A human still makes each decision and stays responsible for the action taken.
The working capital team used Workspace Agents to clear an invoice backlog that would have taken a month. It then ran the agent weekly and monitored it. Because the platform was already in place, central data, AI and technology teams only gave L3 or L4 support. Credits and observability kept costs visible.
Don’t give everyone AI. It sounds simple, but people give everyone AI and they shouldn’t.
Richard Masters, Former VP Data & AI, Virgin Atlantic
Richard Masters is the former VP Data & AI at Virgin Atlantic. He holds a PhD in astrophysics and has spent most of his career in data. He spoke on the AI Infrastructure track at aiblLIVE London on 20 October 2026.
If your AI estate is sprawling, this is a simpler model to borrow. The full PDF goes deeper than this page, in the speaker’s own words.
For CTOs, CDOs, CIOs and data or AI leads planning their AI infrastructure. Six pages, about an eight-minute read.
