Since its launch in 1984, Virgin Atlantic has been synonymous with innovation and delivering a great customer experience. So, while at Adobe Summit in April, I was excited to listen to Neil Letchford, VP of Digital Engineering at Virgin Atlantic, speak about their latest digital initiatives and how they were staying ahead of competitors in what is an increasingly fast and ever-changing landscape.

Virgin Atlantic’s AI Concierge Runs on a Multi-Agent Architecture

During Letchford’s session, he highlighted how Virgin Atlantic are delivering personalization and CX orchestration at scale as well as how generative AI is actively changing customer behavior and altering the traditional marketing funnel.

He went on to explain that, historically, Virgin Atlantic relied heavily on Search Engine Optimization (SEO) to bring users to the top of their funnel, guiding them through a journey of inspiration and discovery on their own website before moving them on to flight selection and checkout. However, Letchford said that as customers increasingly turn to generative AI assistants like ChatGPT and Anthropic’s Claude for high-level, top-of-funnel research, the shape of the funnel is fundamentally changing.

As a result, travelers are arriving at Virgin’s digital properties lower down the funnel, having already completed a significant amount of their upfront research, thinking and planning.

Responding to this shift in consumer behavior, Virgin Atlantic have recently launched an “AI Concierge” designed to support customers wherever they enter the digital journey. To do this, they have built a sophisticated multi-agent bot architecture. This agentic platform features a series of specialized, distinct agents working in the background, including a flight agent, a holidays agent, an FAQ agent and a Flying Club agent.

Now, that’s all very well and good, you might say, and add that many other brands are doing very similar things.

And you’d be right in saying that.

What Multi-Agent Architecture Powers Virgin Atlantic’s AI Concierge?

Virgin Atlantic’s AI Concierge deploys four specialized agents — flight, holidays, FAQ and Flying Club — to meet travelers who now arrive further down a funnel shortened by generative AI research tools like ChatGPT and Claude.

Inside Virgin Atlantic’s ‘Minimum Viable Knowledge’ Approach

But what stood out to me in Letchford’s session was the mention of a tactic called “Minimum Viable Knowledge” and how they have used it to help their AI Concierge perform more effectively.

Intrigued by this idea, I recently had the opportunity to speak with Letchford to learn more.

When I asked him about the “Minimum Viable Knowledge” idea, he told me that their data showed that their highest basket values for holidays came not from their website or contact center, but from their physical retail stores, where their reps were able to sit down and talk through their customers’ plans.

The problem is that, on the surface, booking travel or a holiday can seem like a simple task. You select some dates and a destination, then book flights and accommodation that fit. Simple. In reality, however, it’s far more complex, and reaching a decision can often involve a complex web of personal preferences, budgets, physical constraints, risk tolerances and real-time logistics.

Moreover, capturing all of that complexity in a digital format can be hard.

With that in mind, and to better understand how they could replicate the success of their human reps in their retail stores and codify their approach for their holiday agent, Letchford’s team sat with retail staff to observe how they spoke to customers and guided them through their options.

They found that staff typically uncovered seven key contextual preferences, such as family size and weather preferences, through casual conversation, and these formed the basis of their “Minimum Viable Knowledge.”

Armed with this knowledge, Letchford then explained that they “Gave their holiday agent the task of finding out these bits of information, these seven bits of minimum viable knowledge, just through conversation. So we didn’t program if-then-else. We literally said (to the agent), you have to discover this information through conversation, and that’s what it tries to do.”

The conversational discovery method has proven to be remarkably successful, and it has surprised the team how much users enjoy the back-and-forth interaction and how that is resulting in customers actively booking holidays directly through the concierge. In fact, selling holiday packages is where they have seen the greatest success with their AI Concierge.

What Are the Seven Contextual Preferences Behind “Minimum Viable Knowledge”?

Virgin Atlantic’s digital team observed retail staff to identify seven contextual preferences — including family size and weather tolerance — that reps typically uncover through casual conversation, then trained the holidays agent to surface the same information conversationally rather than through scripted if-then logic.

3 Things to Know About Virgin Atlantic’s AI Concierge Rollout

The following table highlights the most important lessons, actions and strategic considerations emerging from Virgin Atlantic’s AI Concierge rollout and its “Minimum Viable Knowledge” approach.

What’s Next: Personalization, Voice and an AI-First App Experience

Currently, the AI Concierge operates without knowing a user’s personal context. This is a deliberate design choice to ensure a safe and responsible initial rollout.

However, Letchford told me that their next major feature will allow customers to log in using their Flying Club accounts, which will enrich the system’s context graph with zero-party data, enabling their concierge to offer a highly  based on a customer’s individual profile.

In addition, Letchford told me that Virgin Atlantic is also exploring the embrace of voice technology after seeing surprisingly strong engagement, particularly among younger demographics who naturally prefer speaking over typing.

As a result, the airline is building a native version of the concierge directly into its mobile app. With the support of rapidly improving real-time voice models from OpenAI, Virgin Atlantic is heavily focusing its energy on creating a deeply integrated “AI-first app experience”.

Overall, Virgin Atlantic’s story is a wonderful example of how capturing real-world contextual intelligence can influence the design and effectiveness of digital and AI products, and an approach that I hope more brands replicate, given that research shows many customers remain frustrated with the experience they receive from digital and AI products and that they still “feel that human agents better understand their needs.”