The Impact of Agentic AI on the Shopping and Purchasing Habits of Airline Travellers
Summary
This panel discussion at ITB Berlin 2026 examines how agentic AI is reshaping airline distribution, shopping behavior, and the broader travel value chain. Moderated by Dr. Felix Dannegger, the session featured Marcel Pouchain Meyer (Ryanair), Bas 't Hooft (Air France-KLM), Stuart Barwood (Moi AI), and Peter Marriott (Data Error).
Marcel Pouchain Meyer opened with a keynote-style presentation on Ryanair's perspective. He outlined three macro shifts: AI moving consumers from comparison-mode (20 browser tabs on Skyscanner) to receiving a single concrete decision; AI enabling real-time price and disruption monitoring; and natural language search replacing structured queries ('warm beach holiday in May, €120'). He highlighted that 95% of Ryanair's sales come through direct channels, with 126 million app visits per month and 1.7 million per day. Ryanair uses an internal AI tool called the 'Panini Tracker,' which predicts onboard sandwich sales per flight by feeding historical order data, passenger load, and route demographics (e.g., Irish passengers on Dublin–Gran Canaria being predicted to consume more beer). Ancillaries represent one third of Ryanair's total revenue per passenger, making structured data access for AI agents critical. Marcel noted that agentic AI weakens brand identity at the point of purchase because AI ranks results by price, time, and reliability — a trend he described as net positive for low-cost carriers like Ryanair, since it deprioritizes brand loyalty in favor of best-value selection.
Bas 't Hooft (Air France-KLM) noted that while agentic AI is not yet a major booking channel (low single digits in terms of traffic share), it has measurably and significantly changed the inspiration phase of travel planning. He also cited strong impact in disruption handling, where agentic rebooking capabilities are already reducing service workload. He described 20–30 years of industry behavior training customers to find the lowest fare, and sees agentic AI as an opportunity to shift that paradigm toward value-based conversations rather than purely transactional fares.
Stuart Barwood (Moi AI) described being live with American Airlines on 100% of their US business with conversational agentic commerce and live with Wyndham Hotels on Claude. He pushed back on the NDC comparison, noting that 'no passenger was asking for NDC' whereas customers are actively demanding conversational search today. He argued agentic AI is evolution, not replacement, and that airlines can experiment immediately with low barriers to entry if they have an NDC or API connection — no PSS migration required. He warned that Booking.com and OTAs are already present in AI channels, and if the AI learns to default to OTA bookings, that behavior will self-reinforce.
Peter Marriott (Data Error) announced a world-first with Mastercard approximately six weeks prior: an end-to-end agentic payment within WhatsApp, enabling a connected trip (flight + hotel + experience) with a single checkout across multiple PSPs — no form fill. He predicted that within four weeks of the panel, true multi-merchant settlement to PSPs via agent pay would be live. He made the controversial claim that customer data will shift from airline-held profiles to customer-owned agent profiles, tokenized via Mastercard provisioning and portable via single sign-on into ChatGPT or WhatsApp. He predicted that within three to five years, up to 25% of airline bookings will be fully agentic direct-to-supplier, and that in five to ten years, apps will be obsolete as consumers access travel through WhatsApp, ChatGPT, and similar ambient channels.
A recurring tension in the discussion was the direct-versus-indirect classification challenge: if an AI agent books a flight on behalf of a customer, is it a direct or indirect channel? Both airline representatives acknowledged uncertainty, with Marcel noting it 'feels more indirect' for now. The panel consensus was that culturally, airlines move slowly, but the foundational backend work (orchestration layers, MCP rails, NDC/API connectivity, agentic payments) is largely done and the transition could be faster than most anticipate — potentially less than six months to meaningful channel impact.
Key Takeaways
- •Ryanair's 95% direct sales share makes agentic AI a perceived threat to its distribution model, but Marcel Pouchain Meyer acknowledged Ryanair is simultaneously onboarding OTAs through its approved OTA program and accepting that multi-channel presence is unavoidable.
- •Ryanair reports 'high single digits' of bookings now attributable to AI-driven channels; Air France-KLM reports 'low single digits' — both agree search and booking via agentic AI will grow but take time, while inspiration and disruption handling are already materially impacted.
- •The Panini Tracker example illustrates real production AI at Ryanair: flight-level predictive models ingesting passenger count, route profile, and historical order data to forecast onboard item sales with high precision — a practical demonstration of AI structured data use cases beyond chat.
- •Ancillaries are existential to the AI readiness argument: with ancillaries representing one-third of Ryanair revenue per passenger, Marcel emphasized airlines must structure backend ancillary data so AI agents can surface and sell these items dynamically within any channel.
- •Peter Marriott (Data Error) announced a world-first agentic payment with Mastercard launched approximately six weeks before the panel: complete travel bookings — flight, hotel, experience — executed within WhatsApp with a single checkout, agent pay settling to multiple PSPs simultaneously, no checkout form required.
- •Peter Marriott predicted that within four weeks of the panel (i.e., imminently), true multi-merchant PSP settlement would be live — framing the timeline as 'four weeks, not years.'
- •Stuart Barwood framed the NDC comparison as fundamentally different: 'NDC took 15 years because no passenger was asking for NDC. Customers want to transact this way already.' He predicted less than 12 months — possibly less than 6 months — before agentic AI becomes a viable airline distribution channel with high double-digit growth potential.
- •Peter Marriott made the controversial prediction that customer data will no longer reside within airline loyalty profiles but will be portable, tokenized, and owned by the passenger — associated with their AI agent, with Mastercard credentials and preferences specified once and carried across channels via single sign-on.
Transcript
Glad to be here. Glad to be talking on behalf of Rainaire today uh on the topic of AI to you guys. Thanks for your presence. Thank you for for joining the stage today with us. I'm looking forward to have the discussions later with my colleagues over there. So um let's start you know I want to start with a very generic perspective. Meanwhile I'm turning the screen to me. Um so what we are seeing is basically the shaping industry trend that people are now delegating the task to AI instead of searc...
