Airline Intelligence Brief — ITB Berlin 2026

Audience: Airline Commercial Directors, Network Planners, Distribution Heads, Aviation Investors Classification: Strategic Intelligence Source: 135 ITB Berlin 2026 session transcripts, cross-track synthesis, hypothesis testing Prepared: 2026-03-11

Executive Summary

ITB Berlin 2026 delivered a clear verdict on airline distribution: the window to establish direct agentic presence is open, it is measured in months, and the consequences of missing it will compound for a decade. Stuart Barwood (Moi AI), already live with American Airlines on 100% of US business, estimated the agentic channel shift could be "meaningful in less than 12 months, possibly less than 6." NDC took 15 years to reach ~35% of air bookings through major aggregators because no passenger was asking for it. Customers are demanding conversational AI search today. The speed differential is not incremental — it is structural.

The distribution threat is asymmetric in its downside. Peter Marriott (Data Error), who launched the industry's first multi-merchant agentic payment settlement with Mastercard approximately six weeks before the panel, framed the risk precisely: "If the AI learns to default to OTA bookings, that behavior will self-reinforce." Every day that airlines are absent from agentic channels is a day the LLMs are being trained on OTA booking patterns. LLM training is not reversible on demand. Ryanair's 95% direct share — the cost and data foundation of its entire business model — is a structural asset being put at existential risk by a distribution architecture the airline did not create and cannot control from the outside.

The K-shaped demand environment is simultaneously the best and worst backdrop for this transition. Premium cabin revenue continues to surge — Delta premium revenue rose 9% year-over-year, United is launching caviar amuse-bouche in new business class suites, and hotels commanding $1,000+/night have tripled since 2019 — but these gains mask structural fragility. Revenue growth concentrated in the top spending decile creates vulnerability to luxury-segment downturns, and the same K-shaped dynamic is reshaping the payment and distribution infrastructure airlines depend on. Booking Holdings processed $130 billion in travel transactions in 2025 with 63% year-over-year growth; that platform is cementing itself as a payments orchestration layer, not just a booking intermediary.

European rail's competitive threat to short-haul aviation is real but slower-moving than advocates project. The evidence from ITB is that distribution — not infrastructure — is rail's primary barrier: 80% of non-European visitors do not consider taking a train in Europe. That distribution gap is a competitive buffer for airlines today, but it is precisely the kind of problem that a single competent aggregator or AI-native booking layer can solve in a product cycle. ÖBB night train sales grew 20% year-over-year despite concerns about a fading sustainability narrative. The demand for an alternative exists. What has not yet materialized is the frictionless booking experience that would make it the default choice. Airlines should not interpret the current distribution barrier as a permanent feature of the competitive landscape.

The 5 Things You Need to Know Right Now

1. The agentic booking window is closing — not opening. Andrew Boch (Moi AI) and Sanjay Vakil (Direct Booker) were unambiguous: the 6–12 month window to establish direct-to-airline agentic booking patterns is now. LLMs learn from observed behavior. If OTAs are where bookings happen today, models will recommend OTAs tomorrow. Agentic AI currently accounts for "high single digits" at Ryanair and "low single digits" at Air France-KLM — those figures are the early-warning signal, not the peak.

2. AI agentic settlement infrastructure is live, not experimental. Peter Marriott (Data Error) demonstrated multi-merchant agentic payment settlement with Mastercard — a full trip (flight + hotel + experience) within a single WhatsApp conversation, single checkout, multi-PSP settlement — six weeks before the ITB panel. This is not a roadmap item. His prediction: up to 25% of airline bookings fully agentic direct-to-supplier within 3–5 years.

3. Loyalty profiles may not belong to airlines for long. Marriott predicted 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 rather than their airline account. If Mastercard-tokenized preference profiles become the default identity layer, airlines lose the first-party data advantage that underpins their entire CRM and ancillary strategy.

4. Ancillaries are at risk before seat revenue — and that's the larger number. Ryanair derives one-third of total per-passenger revenue from ancillaries. For AI agents to surface and sell ancillaries dynamically, those products must exist as clean, queryable, machine-readable data structures. Airlines without API-first ancillary architectures will have their most profitable per-passenger revenue stream invisible to the fastest-growing shopping channel.

5. The late-market pricing conflict is the next yield management crisis. Three CCOs (TUI Fly, EasyJet Holidays, Condor) confirmed that tour-operator-led capacity planning is already dead — but the structural tension it leaves behind is not resolved. As airlines move to dynamic yield management, package prices and seat-only prices diverge in ways that damage both tour operator relationships and consumer trust. Condor's Christian Lesjak described dynamic pricing as "the only chance for us to get hold of some margin." That is correct — and it requires getting the data architecture right before the agentic channel amplifies mismatches at scale.

Strategic Threats

NDC Servicing Gap as Agentic Amplifier NDC penetration reached approximately 35% of bookings through major aggregators as of ITB 2026, with Traveloft claiming 49–50 direct airline connections. But the well-documented NDC servicing gap — no self-service change and cancel in many implementations — becomes catastrophically worse in an agentic environment. An AI agent attempting to modify an NDC booking on behalf of a traveler, hitting a servicing dead end, and defaulting to an OTA for the next booking is a compound failure with training consequences. Airlines that have not resolved NDC servicing completeness are exposing their entire agentic strategy to a structural back-office flaw.

The OTA Default Training Loop The clearest articulation of the existential risk came from Sanjay Vakil (Direct Booker): "If we seed that territory to the OTAs, we will have another 20 years of dominance by those players." The mechanism is training data, not negotiation. Airlines that are absent from agentic channels in the next 6–12 months are not merely losing current bookings — they are shaping the default behavior of systems that will process hundreds of millions of queries. This is path-dependent in a way that traditional OTA negotiation is not.

GDS Architectural Incompatibility The GDS infrastructure that still carries the majority of managed travel bookings has a structural incompatibility with AI-era personalization: 20-character field limits make it architecturally impossible to pass the contextual data that AI conversations elicit. Hypoallergenic preferences, seat configuration requests, connecting room requirements — all of this is truncated at the GDS layer before it reaches the airline. For airlines dependent on GDS for corporate distribution, this is not a configuration problem; it is an investment decision about which layer of infrastructure gets replaced.

Demand Concentration Risk in the K-Shaped Market Delta's 9% premium revenue growth and United's caviar investment are rational responses to today's demand reality. The top 10% of US earners now account for nearly half of all consumer spending, and the gap is "even more pronounced in travel" (Kopit, Skift). But revenue concentration creates fragility. A luxury-segment correction — triggered by equity market decline, geopolitical instability, or consumer confidence deterioration — lands disproportionately on carriers that have optimized their commercial model around the top decile. Budget and mid-market route economics have been deprioritized at exactly the moment when that demand pool is most price-constrained.

Agentic Liability: No Framework Exists When an AI agent assembles a flight + hotel + experience package and the flight is delayed, no current legal framework determines which party owes the rebooking cost. This was raised at ITB and left entirely unresolved. As agentic booking volumes grow from single digits toward Marriott's 25% prediction, the absence of a liability framework is a regulatory timebomb. Airlines that move aggressively into agentic distribution without establishing clear terms of service for AI-assembled itineraries will find themselves in the worst possible position when the first high-profile case breaks.

Strategic Opportunities

Premium Architecture: The Demand Is Structural, Not Cyclical The K-shaped economy is not a post-COVID anomaly. Moody's data (cited by Kopit) shows the top decile's consumer spending share is at a record high. Hotels commanding $1,000+/night have tripled in the US and Europe since 2019. Condor's 94-seat premium cabin on 18 new Airbus A330-900 Neo widebodies is "reportedly selling well" in German leisure outbound. The premium leisure segment — distinct from corporate business class — is an underserved middle ground between full-service network carriers and ultra-low-cost operators. Carriers that can credibly own this positioning (comfortable hard product, flexible fares, no corporate fare complexity) have a structural demand tailwind.

Non-Western Demand: The 9-Year Pipeline Google's global travel outlook presented a predictable demand pipeline: middle-class growth in emerging markets creates 2–3 years to first domestic travel, another 2–3 for regional, and up to 9 years for long-haul international. India, China, and the US are identified as the three demand superpowers. China outbound has not fully recovered to pre-COVID levels, which means a further catch-up wave is structurally overdue. India's outbound growth trajectory is earlier-stage and faster-growing. Airlines with network and product infrastructure positioned for non-Western demand — payment methods, language capability, fare family structures calibrated to different price sensitivities — are accumulating an advantage that will be visible in revenue in 3–5 years.

Agentic Direct: First-Mover Economics The flip side of the OTA training loop threat is the first-mover opportunity. Airlines that establish direct agentic booking capability in the next 6 months are not just capturing near-term bookings — they are shaping LLM training data. Stuart Barwood (Moi AI) is live with American Airlines on 100% of US business and with Wyndham Hotels on Claude. The infrastructure exists. The MCP deployment window described in the eTravel track — three months from decision to go-live for any company with clean product data — makes this a 90-day decision, not a 3-year program.

Payment Infrastructure as Margin Recovery Booking Holdings processed $130 billion in travel transactions in 2025, at 63% year-over-year growth, across 109 payment methods with approximately 1,000 methods tracked globally. The lastminute.com/Adyen case study demonstrates what a unified payment platform delivers: 70% reduction in go-to-market time for new payment methods, 60% total cost reduction inclusive of compliance, 2% card acceptance rate improvement. Airlines with fragmented PSP relationships, manual FX reconciliation, and virtual-card-dependent B2B settlement are leaving these margins on the table. The $1 fraud = $4 total cost multiplier (Repayd/Plummer) means the fraud dimension alone justifies the infrastructure investment.

Corporate MICE as Agentic Beachhead The MICE sector is the clearest early-win domain for agentic AI, and it is directly relevant to airline commercial teams with corporate and group sales functions. hivr.ai data: 70% of hotel RFP requests are declined or unanswered; Flemings Hotels went from 9% to 18% lead-to-sale conversion after automation; day-one proposals convert at 2x the rate of day-two proposals. Airlines with group fare infrastructure and corporate sales teams face the same RFP volume and response-time economics. Agentic qualification and proposal generation for group bookings is a near-term revenue opportunity that does not require solving the full B2C agentic booking problem.

Distribution: The Agentic Shift

The airline industry's distribution history has a consistent pattern: the channel that wins the discovery layer eventually wins the transaction. Travel agents won by owning the information asymmetry. GDS systems won by owning the inventory aggregation layer. OTAs won by owning the comparison interface. LLMs are now competing for the discovery and research layer — and the transition velocity is categorically different from NDC.

The evidence from ITB 2026 defines the current state precisely: - LLM-to-booking conversion rates rose from approximately 1% to 12% in 12 months (Wyndham/von Dietze) - 40% of consumers now use LLMs for travel search; 95% say they will use again - Agentic transaction share in travel is currently 5–8% (Fliggy/Dr. Alex Chen — the most credible operator-level data point in the conference) - AI-driven travel site visits in the US jumped 3,500% year-over-year (Adobe data, Kopit/Skift) - Fliggy booked 100,000+ airline tickets natively within AI chat during Chinese New Year with a 1-in-4 detail-page conversion rate

The distribution infrastructure race is already underway. Stripe spent all of 2025 building its Agentic Commerce Protocol (ACP) in collaboration with OpenAI. Amadeus has invested in "the MCP layer and the basics." Camino Network's AP2 protocol enables agents to share identity and execute payments across travel verticals. Google launched its Universal Commerce Protocol (UCP) in the US in January 2026. These are not research programs — they are production bets by companies with the scale to shape market structure.

For airlines, the critical insight from the carrier track is the direct/indirect classification crisis. Marcel Pouchain Meyer (Ryanair) acknowledged that an AI agent booking a flight "feels more indirect." Bas 't Hooft (Air France-KLM) noted the commercial framework has not been established. This is not an academic question. Ryanair's entire cost model — and its ability to maintain a 95% direct share — depends on how agentic bookings are classified. If AI-mediated bookings are classified as indirect, Ryanair's structural cost advantage deteriorates overnight. If airlines collectively establish the standard that AI-direct bookings to airline APIs count as direct, the economics flip in their favor.

The practical NDC implication is this: NDC's incomplete servicing is the Achilles heel of the entire agentic strategy. An AI agent needs end-to-end transactional capability — book, modify, cancel, rebook, claim — to be a viable booking surface. Airlines with NDC implementations that cannot service post-booking requests are training AI systems to route around them. Fix the servicing gap first. Build the agentic layer second.

The Rail Question

The European rail threat to short-haul aviation deserves an honest assessment rather than dismissal or alarm. The ITB 2026 evidence supports a nuanced conclusion: rail's competitive pressure is real and growing, but the timeline is slower than advocates project, primarily because the barrier is distributional rather than infrastructural.

What the data shows: - European rail is a €70–100 billion annual industry, but cross-border rail represents only 7–8% of that total - 60% of European flights are cross-border — rail's cross-border share is structurally underdeveloped relative to aviation's - 80% of non-European visitors do not consider taking a train in Europe (Rail Europe data) - ÖBB night train sales grew 20% year-over-year, demonstrating durable demand growth that survives the fading of the sustainability narrative - Omio reported a 50% year-on-year increase in multimodal journey choices — the comparison-shopping layer is maturing

Why the barrier is durable but not permanent: Kurt Bauer (ÖBB) confirmed that rail pricing complexity is legally mandated — disability passes, child age brackets, Klimatickets, regional pricing obligations — creating genuine architecture constraints that cannot be simplified away. New entrant European Sleeper's managing director described incumbent operators as "reluctant, volatile, or hostile" toward distributing startup operators through national platforms. The hypothesis test (H11) concluded that the distribution barrier is real and legally entrenched — not merely inertia.

However, the hypothesis also concluded that the "<2 percentage point over 5 years" modal share change prediction is entirely unanchored to data. No conference session cited current rail modal share baselines. The EU is actively pursuing passenger rights reform that would force non-discriminatory data access for aggregators. Rail Europe is already selling European Sleeper tickets from March 2026. European Sleeper is launching Paris–Berlin via Hamburg now, and Brussels–Cologne–Zurich–Milan in September 2026.

The airline strategic read: The distribution gap is a competitive buffer today, not a competitive moat permanently. The most dangerous scenario for airlines is not rail building more infrastructure — it is a single well-funded aggregator or an AI-native booking layer solving the cross-border rail discovery and booking problem in one product cycle. At that point, the 200-to-800 kilometer routes where rail can genuinely compete on city-center-to-city-center door-to-door time become contested in a way they currently are not.

Airlines should monitor rail aggregator development (Omio, Rail Europe, any LLM-native integration) as the leading indicator of competitive pressure, not infrastructure investment announcements.

The Uncomfortable Truths

NDC is not finished — it is broken in the places that matter most. Approximately 35% of air bookings now flow through NDC on major aggregator platforms. This is progress. But the servicing layer — change, cancel, rebook, disrupt reaccommodation — remains incomplete in most implementations. In an agentic world, a booking system that cannot complete the full transaction lifecycle is not a distribution channel; it is a liability. Airlines have declared victory on NDC adoption while leaving the hard part undone.

Your loyalty program is more fragile than you believe. Airline loyalty programs are among the most valuable assets on balance sheets — some are valued above the airline itself. But the entire asset rests on the assumption that the airline controls the identity and preference data layer. Peter Marriott's prediction that customer preference profiles will be "portable, tokenized, and owned by the passenger" — attached to their AI agent rather than their FFP account — is not guaranteed to materialize, but it is not fanciful. Mastercard AgentPay is live. The tokenized preference profile is a real product direction. Airlines should be building defensible loyalty architectures now, not waiting to see how this develops.

The 95% direct share airlines are the most exposed, not the most protected. Counterintuitively, airlines with the highest direct share — Ryanair being the clearest case — face the greatest agentic disruption risk. Their entire commercial model depends on maintaining that direct relationship. Airlines with historically high indirect share through GDS have already priced that intermediation cost into their economics. For Ryanair, every percentage point of direct share that migrates to AI-intermediated booking is a structural change to cost and data assumptions simultaneously.

Premium growth is not a strategy — it is a bet on a single customer segment. Delta's 9% premium revenue growth is genuine and deserves credit. But the airline industry's uniform pivot toward premium is creating a crowded trade. When every full-service carrier is chasing the same top-decile traveler with the same caviar and flat beds, the differentiation collapses into a pricing war at the segment level. The carriers that will emerge from the premium arms race in the strongest position are those with genuine product and service differentiation — not those with the most recently installed seats.

The sustainability cost is coming and no airline has priced it. The conference's sustainability track and airline track operated in entirely separate conversations. Only 21% of Glasgow Declaration signatories are measuring emissions. The EU ETS is already collecting from aviation. The IMO global decarbonization framework decision is expected September 2026. SAF supply constraints mean carriers cannot buy their way out of this with fuel purchases alone. The carriers that have not built sustainability measurement infrastructure into their cost accounting are flying blind on a cost that regulators are about to make visible.

90-Day Action Plan

Days 1–30: Establish the agentic baseline and close the servicing gap

1. Audit current NDC implementation for servicing completeness — specifically: change, cancel, reissue, disruption reaccommodation via API. This is the prerequisite for everything else. Any agentic booking surface that hits a servicing wall will route future bookings around you.

1. Commission an agentic channel audit: map which LLMs currently surface your flights, what data they are using, and whether your direct booking API is accessible via MCP. If it is not, assign a technical owner to the MCP deployment track with a 90-day delivery mandate.

1. Conduct a direct/indirect classification analysis: quantify what percentage of agentic-influenced bookings are currently classified as indirect, what the revenue impact of reclassification would be, and what contractual changes with GDS partners would be required. This decision has multi-year P&L implications and needs executive sign-off, not a committee.

Days 31–60: Agentic infrastructure and data architecture

1. Inventory ancillary product data structures for machine readability. Every ancillary product — seat upgrades, extra baggage, lounge access, in-flight meal preferences, travel insurance, ground transport — must exist as a clean, queryable API endpoint for an AI agent to surface and sell it. Revenue management teams should treat this audit as a yield management project.

1. Initiate a payment infrastructure review against the Booking.com/Adyen benchmark: 70% go-to-market time reduction, 60% total cost reduction, 2% card acceptance rate lift. Identify the 3–5 payment methods in your fastest-growing non-Western markets that you currently cannot accept without a 4–6 week integration. Those are your Q3–Q4 payment priorities.

1. Brief the loyalty program team on the portable preference profile threat. Commission a scenario analysis: if customer preference data migrates from FFP profiles to passenger-owned AI agent profiles within 3–5 years, what are the implications for partnership economics, ancillary targeting, and tier benefit valuation? Do not wait for Mastercard AgentPay to scale before thinking through the response.

Days 61–90: Market positioning and competitive intelligence

1. Assign competitive monitoring responsibility for European rail aggregator development. Specifically: Omio's multimodal journey product, Rail Europe's new entrant partnerships, and any LLM-native rail booking integration. Set a trigger: if any aggregator or LLM achieves frictionless cross-border rail booking for US/India/China traveler profiles, escalate to the network planning team within 30 days.

1. Pilot an agentic MICE and group fare tool with the corporate sales team. hivr.ai's Flemings Hotels result (9% to 18% lead-to-sale conversion) was achieved in a high-complexity, high-volume, previously manual process — exactly the profile of airline group bookings. This is a revenue recovery opportunity with measurable ROI in a 90-day pilot, not a multi-year technology bet.

1. Engage legal and regulatory counsel on agentic booking liability frameworks. Do not wait for a high-profile AI-assembled itinerary failure to establish your position. The company that tables the first industry-standard terms of service for AI-mediated bookings will shape the regulatory environment rather than react to it.

Key Quotes

> "NDC took 15 years because no passenger was asking for NDC. Customers want to transact this way already." > — *Stuart Barwood, Moi AI* (live with American Airlines on 100% of US business)

> "If the AI learns to default to OTA bookings, that behavior will self-reinforce." > — *Stuart Barwood, Moi AI*

> "If we seed that territory to the OTAs, we will have another 20 years of dominance by those players." > — *Sanjay Vakil, Direct Booker*

> "Memory is the most undervalued asset in the travel industry. Our industry has amnesia at every handoff." > — *Tino Klähne, Principal Futurist, Lufthansa Innovation Hub*

> "Capacity planning should never be tour-operator-led anymore, even if it's TUI." > — *Paul Bixby, CCO, EasyJet Holidays*

> "The only chance for us to get hold of some margin." > — *Christian Lesjak, Commercial Director Sales, Condor* (on dynamic pricing and the elimination of fixed allotments)

> "A huge upstroke at the top and a flat to declining line everywhere else." > — *Sarah Kopit, Skift* (on the K-shaped demand economy)

Data Points That Matter

| # | Data Point | Source | |---|---|---| | 1 | LLM-to-booking conversion: ~1% → 12% in 12 months | Uta von Dietze, Wyndham (eTravel Track) | | 2 | Agentic transaction share in travel: 5–8% currently, "on the brink of a breakthrough" | Dr. Alex Chen, Fliggy (primary source) | | 3 | Ryanair direct channel share: 95% of sales; 126 million app visits/month; ancillaries = one-third of revenue per passenger | Marcel Pouchain Meyer, Ryanair (Carrier Track) | | 4 | NDC bookings via major aggregator: ~35% of all air bookings processed; Traveloft: 49–50 direct NDC airline connections | Carrier Track sessions | | 5 | Agentic AI booking prediction: up to 25% of airline bookings fully agentic direct-to-supplier within 3–5 years | Peter Marriott, Data Error | | 6 | Fliggy Chinese New Year: 100,000+ airline tickets booked natively in AI chat; 1-in-4 detail-page conversion rate | Dr. Alex Chen, Fliggy | | 7 | Delta premium revenue: +9% YoY (Q3 2025); hotels at $1,000+/night tripled in US and Europe since 2019 | Sarah Kopit, Skift | | 8 | Booking Holdings transaction volume: $130B in 2025, 63% YoY growth, 109 payment methods, 130+ provider relationships | Daniel Marovitz, Booking Holdings (eTravel Track) | | 9 | 80% of non-European visitors do not consider taking a train in Europe; cross-border rail = only 7–8% of the €70–100B European rail market | Rail Europe / Carrier Track | | 10 | ÖBB night train sales: +20% YoY despite fading sustainability narrative; Omio multimodal journey choices: +50% YoY | Carrier Track (Rail session) | | 11 | GDS field limit: 20-character note fields structurally incompatible with AI contextual data requirements | Sanjay Vakil, Direct Booker (eTravel Track) | | 12 | Fraud cost multiplier: $1 fraud = $4 total operational cost; 100% of travel revenue flows through the payments layer | Repayd / Travelsoft Pay (eTravel Track) |

*Generated from ITB Berlin 2026 research corpus: 135 session transcripts, 17 track research memos, cross-track synthesis, and structured hypothesis testing across protocol consolidation (H02), K-shaped pricing dynamics (H08), European rail modal share (H11), and payment infrastructure consolidation (H15).*