The Architecture of Tourism Is Being Rewritten And Control Is Shifting
Artificial intelligence is not improving the system: it is redefining where decisions are made, how they are executed, and who ultimately controls them.
By Christian Delom, President, Global Resilience Network
A transformation misread at the surface
Artificial intelligence in tourism continues to be interpreted through what it renders most immediately perceptible: the generation of content, the refinement of conversational interfaces, and the incremental optimization of customer interactions. The rapid diffusion of generative tools across platforms such as Booking.com or Expedia Group illustrates this tendency, as itinerary builders, automated descriptions and conversational assistants become visible markers of transformation.
These developments are neither trivial nor negligible. McKinsey & Company estimates that generative AI could unlock between $60 and $110 billion annually across travel, transport and logistics, largely through productivity gains in marketing, customer service and operations. These visible applications matter not only because they improve productivity at the edge of the system, but because they make artificial intelligence tangible to organizations and thereby accelerate its adoption.
Yet they remain confined to the surface of the system and, in doing so, contribute to a misreading of what is taking place. For what is unfolding does not primarily concern the production of outputs, nor even the automation of discrete tasks. It concerns a displacement at a far more structural level: the relocation of where decisions are formed, structured and executed.
This shift is already measurable. In hospitality, dynamic pricing systems, now used by more than 70% of major hotel chains according to Boston Consulting Group, integrate real-time signals such as booking pace, competitor pricing, event calendars and transport flows. In aviation, yield management systems process millions of pricing adjustments daily. More importantly, in leading environments, these adjustments are no longer confined to pricing alone but increasingly linked to inventory allocation, distribution visibility and operational prioritization. What is emerging is not simply optimization, but continuous decision recalibration at system scale. Artificial intelligence is no longer embedded within an existing system to assist its actors; it is progressively redefining the architecture through which decisions are generated, sequenced and executed. The transformation, therefore, does not lie in what the system produces, but in how it decides
From decision systems to operating systems
To apprehend this transformation, it is insufficient to distinguish between types of models or applications. The relevant distinction lies in degrees of integration. At one level, artificial intelligence augments human activity. At another, it optimizes functional domains. At a further level, it executes decisions across operational processes. Revenue management systems used by major hotel groups, or airline yield management platforms historically developed by actors such as Amadeus, already reflect these stages.
Emerging AI-first operational models in hospitality, combining pricing, distribution and demand forecasting within increasingly synchronized loops, already indicate how previously separate decision domains are being drawn into a shared operating environment. In several Chinese deployments, these loops are extended further to include identity recognition, payment systems, service orchestration and facility management, creating closed decision cycles that operate with minimal human intervention.
Yet these layers do not exhaust the transformation. Beyond them emerges a configuration that can no longer be adequately described as a set of tools or even as isolated decision systems. What is taking shape is closer to an operating system for the industry itself: an integrated architecture in which data ingestion, model-based reasoning, decision arbitration and execution loops are synchronized in real time across the entire value chain.
In such a system, pricing, distribution, operations, mobility, infrastructure and customer interaction cease to function as distinct domains. They become interdependent variables within a single environment of continuous arbitration. This is not a technological upgrade, nor even a functional optimization. It is a reconfiguration of the system’s organising logic, in which the coherence of the whole replaces the optimization of the parts.
The economic consequence: coordination becomes the locus of value
Tourism has long been structured around fragmentation. Airlines optimise yield, hotels calibrate occupancy, platforms intermediate demand, destinations cultivate attractiveness, and infrastructure absorbs and redistributes flows. Each actor controls a segment of the value chain, and coordination emerges imperfectly from their interaction. Each operated within a bounded horizon, based on partial information and within temporal cycles that allowed for intermittent rather than continuous decision-making. This configuration has endured because coordination costs historically exceeded the value of integration.
That equilibrium is now eroding. As OECD analyses have shown, the cost of coordinating complex systems has fallen dramatically due to advances in real-time data processing and interoperability. The constraint no longer lies in the availability of data, but in the absence of an environment capable of arbitrating heterogeneous variables simultaneously. When coordination becomes cheaper than fragmentation, the basis of the system shifts.
Value migrates accordingly. It no longer resides primarily in the ownership of assets, nor in the control of intermediary positions, but in the ability to coordinate decisions across the chain in real time. What emerges is not simply a new source of efficiency, but a new centre of gravity for value creation. This is the moment at which control begins to relocate, following not ownership, but orchestration.
Control and the reorganisation of the value chain
The emergence of operating systems therefore implies more than efficiency gains. It entails a redistribution of control across the value chain. Historically, control has been distributed. Airlines controlled pricing, hotels controlled inventory, platforms controlled demand aggregation, and destinations influenced flows. No single actor had full visibility or authority over the system.
Operating systems alter this distribution fundamentally. By integrating data, modelling, decision-making and execution within a unified architecture, they enable a form of end-to-end coordination that transcends traditional boundaries. The actor, or set of actors, that controls this architecture effectively controls the system itself.
This is not an abstract shift. It directly affects pricing authority, customer access, visibility within distribution channels and the prioritization of operations. In environments where these systems are already deployed, operators observe that decisions previously taken across days or weeks are now resolved in seconds, often before human intervention would even have been possible. Control no longer follows assets. It follows decision orchestration, and with it the capacity to structure how the entire system behaves.
Hard evidence: the execution layer is already shifting
The most telling developments are no longer experimental pilots, but production-level systems where artificial intelligence is already making and executing decisions across multiple layers without human mediation.
In one environment, a disruption occurs. Flights are delayed, connections missed, demand shifts within minutes. In a traditional system, resolution unfolds through escalation, coordination calls, and sequential human arbitration. In an integrated system, the sequence does not exist. Pricing adjusts, inventory reallocates, itineraries reconfigure, and customer communication is executed before any human intervention takes place. The difference is not technological. It is structural. One system reacts. The other decides.
In 2025, Alipay extended its “agentic commerce” capabilities through the deployment of an integrated AI concierge, capable not only of recommending itineraries but of orchestrating and executing entire travel journeys, transport, accommodation, payments, access and in-destination services, within a single conversational flow. What distinguishes this model is not the interface, but the collapse of boundaries between intention, decision and transaction. The system does not assist the user in navigating options; it resolves those options and executes them. In these environments, users no longer interact with multiple platforms. They remain within a single system across the entire journey, creating a form of behavioural lock-in that progressively reduces the relevance of external distribution channels.
In parallel, Trip.com Group has deployed AI systems capable of autonomously reconfiguring itineraries during disruptions—combining real-time flight data, pricing, availability and customer constraints—executing changes without requiring sequential validation. These systems do not assist agents; they replace the decision chain itself, compressing what was previously a multi-step process into a single continuous loop. These systems do not optimise existing processes. They collapse them, compressing multi-step decision chains into continuous execution loops.
At the infrastructure level, NVIDIA has publicly framed the transition toward “physical AI” and agentic systems capable of interacting with real-world environments. In travel and mobility, this translates into the integration of AI across logistics, robotics, predictive maintenance and flow optimisation extending decision systems beyond digital interfaces into physical operations and reinforcing the continuity between digital decision and physical execution.
In aviation, airlines such as Delta Air Lines have begun integrating AI into disruption management systems that automatically reassign passengers, reprice inventory and adjust capacity in real time across networks. These systems operate on a scale and speed that fundamentally alter how operational decisions are made, not by improving existing processes, but by bypassing them.
The first irreversible step: execution without human sequencing
What these developments have in common is not technological sophistication, but a structural shift that has already occurred
This is where the break becomes visible. Not when artificial intelligence assists decisions, but when it removes the need for sequencing them. In such environments, decision-making is no longer ordered. It is simultaneous.
The industry has begun to move from AI-assisted decisions to AI-executed decisions, and in the most advanced configurations, to AI-orchestrated journeys, where the system itself determines the sequence of actions across the entire travel lifecycle.
This distinction is critical.
In assisted environments, humans remain the point of convergence. In execution environments, systems become the point of convergence. In orchestrated environments, systems define both the sequence and the outcome.
Once this transition occurs, decision-making no longer follows a linear sequence. It becomes simultaneous, continuous and increasingly autonomous. At that stage, reverting to fragmented, human-mediated processes is no longer a neutral option. It implies slower execution, weaker coordination and structurally inferior performance.
For a decision-maker operating within fragmented systems, this shift is not immediately visible. Processes continue to function. Performance appears stable. But the divergence does not emerge in static conditions. It emerges under pressure—when speed, coordination and simultaneity determine outcomes. It is in these moments that the system reveals itself.
A divergence already measurable
The gap does not announce itself. It accumulates.
This shift is already producing measurable divergence between actors.
Organisations that have integrated execution and orchestration layers report faster response times in disruption scenarios, higher conversion through dynamic bundling, and improved yield through real-time coordination across pricing, availability and distribution. More importantly, they accumulate data not as isolated datasets, but as continuous feedback loops linking decision, execution and outcome.
By contrast, organisations still operating through fragmented systems continue to optimise functions independently, without access to system-wide arbitration. The result is not simply a performance gap. It is a divergence in learning speed, where some systems improve continuously while others remain dependent on episodic adjustment.
The gap is not only in performance. It is in learning velocity, some systems improve continuously, while others remain structurally intermittent.
From the outside, the difference appears incremental. From within the system, it becomes exponential.
China: scale as a structural accelerator
The divergence between regions must be understood in this context. China’s advantage lies not primarily in technological sophistication, but in its capacity to deploy systems at scale within integrated environments. Within ecosystems structured by Alibaba Group and Tencent, data flows continuously across payments, mobility, retail and tourism. These are not adjacent sectors, but interconnected layers within unified systems.
Trip.com Group processes vast volumes of transactions, continuously recalibrating pricing, inventory and recommendations across the travel journey. In cities such as Hangzhou and Shenzhen, these capabilities extend into urban systems, where transport capacity, site access and visitor flows are dynamically adjusted.
The FlyZoo Hotel illustrates this logic at the operational level, where identity recognition, room allocation, service delivery and payment are integrated into a single decision loop.
In such environments, innovation does not precede scale; it results from it. The more systems are deployed, the more data they generate. The more data they generate, the faster they learn. The faster they learn, the more they improve. This creates cumulative advantages that are not merely competitive, but structurally self-reinforcing. What appears, from the outside, as technological advancement is, in practice, the result of sustained, large-scale execution.
The United States: infrastructure dominance and systemic convergence
The United States retains a decisive advantage at the level of technological infrastructure. Google, Amazon and Microsoft underpin a large share of global AI capabilities. This dominance is foundational and shapes the evolution of the global ecosystem.
Yet it does not automatically translate into control. The travel ecosystem remains fragmented across platforms, suppliers and intermediaries. The transformation underway is therefore less a question of technological capability than of alignment across layers that have historically evolved separately. The central issue is therefore one of convergence: whether these layers can be aligned into integrated decision architectures before alternative models consolidate elsewhere. The outcome is not predetermined, but the timeframe within which it must occur is increasingly constrained.
Europe: components without system
Europe presents a more complex configuration that cannot be reduced to delay or absence. The continent already possesses many of the structural components required to build operating systems. Distribution and data infrastructures are anchored by actors such as Amadeus. Mobility platforms capable of orchestrating flows exist through organisations such as SNCF Connect & Tech. Cloud and infrastructure capabilities are emerging with players like OVHcloud and Deutsche Telekom, while enterprise-scale software integration is structured by actors such as SAP. More recently, initiatives such as Mistral AI signal that Europe is also re-entering the foundational layers of model development.
Taken individually, these elements are not marginal. Together, they form the outline of a potential architecture.
Yet they do not operate as a system.
Europe is therefore not merely behind in execution; it operates within a framework that limits the very form of system it seeks to enable. Through the European Commission, Europe exerts a powerful influence on data governance, competition and artificial intelligence. This normative capacity shapes global markets. But the same frameworks that regulate concentration and protect competition also constrain the alignment, integration and data aggregation required to build operating systems at scale.
This is not a temporary lag. It is a structural contradiction between the logic of regulation and the logic of system formation.
Europe does not lack assets.
A European airline may optimise pricing. A destination may refine its attractiveness strategy. A platform may improve conversion. But if these decisions are not synchronised within a shared system, they remain structurally isolated. Meanwhile, elsewhere, these variables are already arbitrated simultaneously not as separate decisions, but as interdependent components of a single system. The difference is not effort. It is architecture.
It lacks systemic convergence.
Integration without control: Europe’s central risk
The central risk for Europe is therefore not exclusion, but displacement. European actors will continue to operate within global AI ecosystems. But without a European-controlled operating layer, the logic that structures pricing, visibility, customer interaction and operational coordination will increasingly be defined externally.
Actors remain present within the value chain. But the architecture that organises that value chain escapes them. They operate assets, while others operate the system.
This distinction is not semantic.
It defines where control ultimately resides.
A transformation beyond tourism
The construction of such operating systems extends beyond tourism itself. It requires coordination across data infrastructure, mobility systems, cloud environments, enterprise software and public coordination mechanisms. The architecture of the future industry will therefore not be built by tourism actors alone, but at the intersection of sectors that have until now been analysed separately.
A transformation affecting all stakeholders
The implications of this shift extend across the entire ecosystem. For companies, the issue is no longer simply to adopt artificial intelligence, but to determine whether they will operate within, or contribute to shaping, the decision layers that organise the system. In environments where these systems are already operational, the difference is measurable in revenue capture, cost structure and speed of execution.
For destinations, the challenge lies in the ability to manage flows, optimise infrastructure usage and regulate environmental and social impacts in real time.
For public authorities, the transformation raises questions that go beyond regulation. It concerns industrial policy, data governance and the organisation of coordination at scale. It also raises the issue of sovereignty, not as an abstract concept, but as the capacity to influence the infrastructures through which decisions are made.
A closing window: the irreversibility of system formation
This transformation is not open-ended. As operating systems take shape, learning effects intensify, coordination advantages compound, and the relative position of actors becomes increasingly difficult to alter.
What emerges is not simply competitive advantage.
It is structural lock-in.
The next five to seven years will be decisive. By the end of the decade, the question will no longer be how to adopt artificial intelligence, but whether one retains any meaningful capacity to influence systems whose architecture has already been established.
Conclusion
The dividing line is no longer between those who understand artificial intelligence and those who do not.
It lies between those who have grasped that tourism is becoming an operating system, and those who continue to approach it as a fragmented set of optimisations.
As these systems consolidate, adaptation ceases to be the central issue.
What remains is a question of control.
And once control has shifted, it rarely returns.
The question is no longer who will adopt these systems, but who will still be able to shape them.
Why Partner With Us
GRN unites leaders across public and private sectors to deliver practical solutions for resilient, competitive, and future-ready tourism. Our programs turn insight into action, creating measurable impact worldwide.
