Europe's First Agentic Purchase Was a Trip
Among the first live merchants of the European Visa rollout was lastminute.com. That agents buy travel first is no accident: travel is complex, advisory and context-driven. There, the quality of product data and context decides whom the agent recommends.
The Merchant List and What It Signals
When Visa Intelligent Commerce launched in Europe in July 2026 with more than 30 issuer partners, the connected merchant list included lastminute.com, Frasers, Cleverbridge, and BrickDepot. [Q1] The presence of lastminute.com in that initial cohort is not incidental.
Visa's partners are not passive recipients of a payment protocol. They are organizations that made an active decision to build agent-compatible infrastructure ahead of the market. That lastminute.com was ready at launch, while many larger travel operators were not, signals something about where the travel vertical sits in the agentic commerce transition.
Why Travel Is First
Travel is complex, advisory, and context-dependent in ways that make it the natural early territory for AI agents. A travel purchase is not a single attribute decision. It involves dates, destinations, routing options, accommodation quality signals, transfer logistics, price-to-value tradeoffs, and cancellation conditions, many of which interact with each other.
A keyword search engine returns documents ranked by approximate relevance. An AI agent parses the actual constraints of a specific trip, evaluates the options against the user's stated and inferred preferences, and produces a recommendation that integrates multiple dimensions simultaneously. For a complex, advisory product category, the agent performs a qualitatively different task than a search engine.
This is why travel did not wait for agentic commerce to mature before engaging with it. The value proposition for the end user is immediately apparent in travel in a way that it is not for a commodity product purchase.
What the Agent Evaluates That a Search Engine Does Not
A search engine returns a set of results. The user evaluates the results. The evaluation is entirely on the user's side. An AI agent performs the evaluation and returns a recommendation. The evaluative criteria that the agent applies, and the data it draws on to apply them, determine whose offering appears in the recommendation and on what terms.
In travel, the agent-relevant data layer includes structured availability and pricing feeds, policy machine-readability (cancellation, change, rebooking), review data that can be semantically processed rather than only aggregated, and context matching between the trip parameters and the offering attributes. A travel product without a machine-readable, semantically rich data layer is structurally disadvantaged in an agent-evaluated market, regardless of its ranking in human-facing search results.
Adoption Trajectory
AI-influenced traffic converts at 42 percent above baseline, with revenue per visit 37 percent higher than non-AI-influenced traffic. [Q11] The directional implication for distribution is clear: traffic arriving from AI-evaluated recommendations is disproportionately commercial.
At the market level, projections estimate up to USD 1 trillion in US B2C revenue flowing through agentic channels by 2030, with the top decile of providers capturing over 85 percent of the economic profit in that segment. [Q12] The concentration dynamic is as relevant for travel as for any other sector: the window for establishing the data and infrastructure position that agents will evaluate is a 2026 to 2027 decision, not a 2029 decision.
Top decile providers are projected to capture over 85% of economic profit in agentic commerce by 2030.
Source: McKinsey/ICSC (Q12)Product Data and Context Are the New Distribution Layer
In a human-facing distribution model, SEO, paid search, and metasearch placement determined discovery. In an agent-mediated distribution model, the quality and machine-readability of product data determines whether the agent can evaluate the offering at all, and how accurately it represents it when making a recommendation.
A travel operator whose product data is available only in formats designed for human browsers, whose cancellation policy is expressed only in natural language rather than structured attributes, and whose availability feed requires a session-based authentication flow will not be evaluated by agents that operate without a human in the loop.
The infrastructure investment for travel is therefore a data and API investment, not primarily a consumer interface investment. The consumers who matter most in this transition are the agents acting on their behalf.
The broader control layer in which travel distribution sits is analyzed in The Control Layer: Why Agentic Commerce Is Not Decided at Checkout. The market timing argument is in The Loud Summer of 2026.
Sources:
- Q1: Visa Intelligent Commerce press release, July 2026. 30+ European issuers including lastminute.com, Frasers, Cleverbridge, BrickDepot.
- Q11: Adobe Analytics report. AI traffic converts 42% better, revenue per visit +37%.
- Q12: McKinsey/ICSC. Up to USD 1 trillion US B2C revenue via agentic commerce by 2030; top decile takes over 85% of economic profit.
Advisory on Product Data and Context Strategy in Travel Distribution
Travel operators who invest in machine-readable product data, structured availability feeds, and agent-compatible policy formats now will be evaluated by agents. Those who do not will not appear in the recommendation set at all.
Request an advisory conversation