AI: From Artificial Intelligence to Automated Impact – aviation enters the age of the agent

AI is rapidly moving beyond the era of experimentation. For aviation and travel, the question is no longer whether artificial intelligence can produce impressive answers, generate content or analyse vast quantities of data. The more important question is what happens when AI starts to act - influencing decisions, automating processes, personalising offers, anticipating customer needs and increasingly executing tasks on behalf of travellers and businesses.

The headlines may be on the practical deployment of AI: from airline retailing, distribution and revenue management to customer service, disruption management, operations and the emerging world of agentic commerce. But this analysis questions where the technology is genuinely delivering value, where the industry is still navigating hype and experimentation, and what changes when the customer interacting with an airline may itself be an AI agent rather than a human.

That question is already becoming commercially relevant as technology providers explore AI agents that can discover, compare, book and service travel, potentially reshaping the relationship between airlines, intermediaries and customers.

The underlying argument is that AI's significance for aviation will ultimately be measured less by its intelligence than by its impact. The competitive advantage will increasingly belong not to organisations that simply possess AI capabilities, but to those capable of embedding them into the commercial and operational fabric of the business - turning better data and better predictions into faster decisions, lower friction, new revenues and more responsive customer experiences.

Drawing on discussions at T2RLEngage in London in Sep-2026, this analysis examines the transition from artificial intelligence to automated impact. It explores the emerging role of agentic AI in travel discovery and booking, the implications for airline distribution and customer ownership, and the opportunities to improve retailing, revenue management, testing and operational efficiency. It also considers the obstacles: legacy technology, fragmented data, complex integration, governance and the difficulty of validating decisions in an industry where errors can have consequences well beyond a digital transaction.

It is no longer about whether AI can demonstrate impressive capabilities. It is whether airlines can translate those capabilities into reliable, measurable business value. Competitive advantage will depend less on who adopts AI first than on who can connect it to real decisions and workflows - while retaining control of the customer relationship, the underlying data and the authority to act.

Summary

• AI in aviation is shifting from experimental tools to agentic systems that execute tasks and influence real decisions across commercial and operational workflows.
• Agentic AI could reshape travel discovery and booking as software agents search, compare, recommend, and transact on behalf of travellers.
• Airlines must make offers machine-readable and executable, or risk being excluded from AI-agent consideration sets despite strong consumer-facing marketing.
• Customer ownership and distribution power may shift toward whoever controls the AI interface, potentially reducing airlines to interchangeable suppliers.
• AI value will depend on embedding it into end-to-end workflows with measurable outcomes (faster decisions, lower friction, new revenue), not on impressive demos.
• Key blockers remain legacy systems, fragmented data, integration complexity, and the need for rigorous validation and governance in high-consequence operations.

The AI story has changed - and the pace is accelerating

Two years ago, artificial intelligence occupied a very different place in the airline technology conversation. Today, the focus is shifting rapidly from what AI can say to what it can do.

Richard Clarke, CEO of T2RL, captured the change succinctly at the 2026 edition of T2RLEngage. The "AI story has shifted". "AI was almost nowhere two years ago," he observed, but the focus on executing transactions has now become massive, with progress moving at an incredible pace.

That is more than a technological observation. It signals a change in how aviation should think about AI's commercial significance.

The first wave of adoption has been dominated by experimentation: generating text, summarising information, assisting employees and making large volumes of data easier to interpret. These applications can be useful, but they do not necessarily change how an airline operates or competes. They can improve an existing process without fundamentally altering it.

The next stage is more consequential. AI is increasingly being developed to interpret an objective, determine the steps required to achieve it, interact with different systems and execute a task. This is the territory of agentic AI: technology that moves beyond responding to instructions towards managing workflows and taking actions within defined parameters.

This is a huge step. An AI assistant that explains a passenger's options during a disruption is useful. An AI agent that identifies suitable alternatives, checks eligibility, changes the booking and communicates the outcome could remove much of the friction from the process.

The same distinction applies commercially. Analysing customer behaviour is one capability; using that insight to assemble an offer, evaluate its value and complete a transaction is another.

Neither transition will happen uniformly across aviation. Some applications are already moving towards practical deployment, while others remain constrained by legacy systems, fragmented data and questions about reliability. But the direction is becoming clearer: the value of AI will increasingly be measured by what it enables an organisation to accomplish, rather than the sophistication of the technology itself.

From artificial intelligence to automated impact

The terminology surrounding AI can obscure important differences between applications. Traditional AI has long supported analytical tasks such as forecasting, classification and optimisation. Generative AI adds the ability to create content, interpret natural language and make technology more accessible through conversational interfaces. Agentic AI takes the next step by connecting reasoning to action, potentially coordinating several tasks across multiple systems to achieve an objective.

These categories overlap. An airline does not need to replace its existing technology with a single autonomous system to benefit from the progression. It can combine predictive models, generative interfaces and automated workflows, introducing greater autonomy where the business case and risk controls justify it. The practical opportunity is to move from isolated assistance towards end-to-end execution.

An airline employee might currently use several applications to investigate a problem, interpret the available information, identify an appropriate response and initiate a transaction. An AI-enabled workflow could bring these steps together, presenting the employee with a recommended action or completing routine tasks automatically within agreed boundaries.

The same principle applies to customer service, revenue management, software testing and disruption recovery. In each case, the potential benefit comes not simply from a more intelligent system, but from reducing the time, effort and friction between identifying a need and resolving it.

Yet automation should not be confused with autonomy without limits. Aviation is a complex, interconnected industry in which a seemingly straightforward decision can affect passengers, aircraft, crew, airport operations, commercial commitments and regulatory obligations. The strategic objective is therefore not to remove human involvement indiscriminately. It is to determine where decisions can safely be automated, where human approval remains necessary and how the system should respond when circumstances fall outside its operating parameters.

This will separate practical deployments from demonstrations that look impressive but struggle in the real world.

The customer may no longer be doing the shopping

Perhaps the most consequential development discussed at T2RLEngage concerns the emergence of AI agents as intermediaries between travellers and airlines.

Surain Adyanthaya of PROS highlighted a fundamental change: AI agents are increasingly becoming the shopper, yet many airline offer platforms were not designed to be read and acted upon by machines.

The traditional customer journey is familiar: search, compare, choose and buy. The traveller visits websites, assesses alternatives and makes the decision. An agent-led journey could look very different. A traveller states an intention - perhaps a trip with specific arrival times, baggage requirements, loyalty benefits, flexibility and budget constraints. An AI agent then evaluates alternatives, weighs the trade-offs and recommends an itinerary, potentially completing the booking as well.

Here, the traveller remains the customer, but the immediate decision-maker may be software. That changes the competitive environment. Airline websites have traditionally been designed to communicate with people, using visual presentation, brand messaging, merchandising and a sequence of choices to influence purchasing behaviour. AI agents need something different: structured information they can retrieve, interpret, compare and act upon reliably.

A compelling offer that is difficult for a machine to understand may never make it into the consideration set. Conversely, an offer that is clear, comparable and executable may be more likely to be selected - even if another airline has invested more heavily in conventional digital marketing.

The challenge is therefore not merely to make airline content available to AI systems. Airlines must ensure their products, prices, conditions, ancillary services and booking options can be understood accurately and translated into a valid transaction.

This is a distribution issue as much as a technology issue. AI-driven discovery could become another important route through which demand reaches airlines, alongside direct channels, global distribution systems and New Distribution Capability content. But its commercial characteristics may differ from those of established channels, particularly when the agent controls the interface through which options are evaluated.

Airlines will need to understand where their offers are being discovered, how they are being represented and what happens between a recommendation and a completed booking. Without that visibility, they risk losing influence over the purchase journey before the transaction has even begun.

Who owns the customer when an agent makes the decision?

The rise of agentic commerce raises a question that airlines have spent decades trying to answer through their distribution strategies: who controls the customer relationship? If an AI agent searches for a flight, compares alternatives, recommends an airline and completes the transaction, the airline may provide the transport while another organisation controls the decision-making interface.

That organisation could be an online travel agency, a technology platform, an independent AI service or an agent operating on behalf of the traveller or a corporate travel programme. The eventual balance of power remains uncertain, and the market is unlikely to settle on a single model immediately.

But the implications are significant. The organisation controlling the interface may influence which airlines are considered, how their products are compared and which attributes determine the final recommendation. Price may remain important, but the agent could also weigh schedule suitability, connection quality, baggage allowances, loyalty benefits, flexibility and travel-policy compliance.

This creates opportunities for airlines that can present differentiated offers in a form that AI systems can understand. It also creates risks for carriers whose products are reduced to a limited set of attributes, or whose commercial propositions cannot be represented accurately through an intermediary's interface.

Customer ownership is not simply a question of whose name appears on the booking. It concerns access to information about preferences, the ability to influence future purchases, the opportunity to sell additional products and the capacity to resolve problems directly. If an AI platform retains the context of the traveller's needs and preferences, the airline could find itself fulfilling a transaction without developing much knowledge of the person travelling. The agent might become the continuing relationship, while the airline becomes one supplier within its portfolio.

That outcome is not inevitable. Travellers may use agents supplied by airlines, travel management companies, technology platforms or other intermediaries. Some may prefer to delegate only the research, while retaining control of the final purchase. Others may permit more extensive automation. The strategic imperative is to prepare for these different relationships rather than assume that AI will simply become another interface directing traffic to existing airline channels.

Airlines should treat AI discovery as a distribution channel requiring explicit ownership, commercial policies, performance measurement and investment. They also need to establish how their offers will be represented, what information they can obtain about transactions and which actions an agent is authorised to perform.

The question is no longer only how to attract the traveller. It is how to remain commercially relevant when software increasingly mediates the traveller's choices.

Retailing becomes a continuous decision

AI's significance extends beyond the channel through which a passenger discovers a flight. It could also change how airlines construct and manage what they sell. Traditional airline revenue management has been built around inventory, fare classes and the optimisation of available seat capacity. Modern retailing seeks to integrate pricing, products, bundles and customer context into a more comprehensive offer.

The Offer Lab session led by Suzanne Grimes, Vice President, Strategic Consulting at PROS at T2RLEngage, explored this shift from optimising seats to optimising the offer itself. The underlying challenge is to connect pricing, product and capacity decisions rather than treat them as separate commercial activities.

AI can help airlines interpret demand signals, assess price sensitivity and determine which combinations of products are likely to create value for particular customers. Instead of relying exclusively on predefined fare structures and static bundles, an airline can work towards assembling offers that reflect changing demand and the wider commercial context.

Agentic systems could take this further by linking analysis to execution: identifying an opportunity, selecting an appropriate offer, applying the relevant commercial rules and presenting the result through the channel where the customer is shopping. However, personalisation should not be mistaken for simply offering every customer a different price. Its value depends on understanding the relationship between the product, the customer's needs, willingness to pay and the airline's own commercial objectives.

Nor can an AI model overcome a fragmented retailing architecture by itself. The necessary data, inventory, pricing logic, product definitions and transaction capabilities must work together. If these elements remain disconnected, automation may accelerate an inconsistent process rather than create a better one.

The commercial prize is a more continuous decision-making process: offers can be evaluated, adjusted and improved as market conditions and customer signals change. But achieving that outcome requires airlines to align their technology, data, revenue management and product teams around shared objectives.

Legacy systems remain a constraint - but AI changes the economics

The deployment of agentic AI confronts a familiar aviation technology problem: the industry's dependence on established systems and complex interfaces.

Airlines do not operate on clean technology foundations. They must connect reservations, inventory, departure control, customer records, distribution, operational systems and an extensive ecosystem of third-party providers. Data may be duplicated, processes may cross organisational boundaries and older systems may not expose the information or functions required for more autonomous workflows.

The T2RLEngage discussions highlighted these integration challenges, particularly in the context of agentic AI and modern airline retailing. Legacy Passenger Service Systems can constrain the introduction of new capabilities, while data silos and compliance requirements make it harder to create consistent experiences across channels.

AI is not a substitute for resolving these architectural problems. An agent cannot reliably execute a task if it cannot establish the current state of a booking, determine which rules apply or confirm that a transaction has succeeded. But AI may also change the economics of addressing them.

T2RL's own market analysis has identified the growing significance of functionality developed outside traditional PSS boundaries. More efficient software development could make it easier for airlines and technology partners to build components that were previously too costly or complex to develop independently.

This could gradually loosen the relationship between airline innovation and the capabilities bundled into incumbent platforms. Larger airlines may have more resources to pursue this approach, while smaller carriers could remain more dependent on established suppliers and packaged solutions.

The implication is not that airlines should abandon their core systems or develop everything themselves. Building and maintaining airline technology remains demanding, and a proliferation of disconnected applications could create new costs and operational risks.cInstead, AI strengthens the case for modular development, clearer interfaces and selective partnerships.

Airlines need to decide which capabilities are strategically differentiating, which are better sourced from specialist providers and how new tools can be introduced without compromising the integrity of existing operations. This also makes data access and contractual rights more important. Airlines need sufficient control over their data and technology interfaces to experiment, validate new capabilities and retain the ability to change providers where necessary.

Automation is only as good as its validation

The ability of AI to generate software, coordinate workflows or recommend actions does not establish that the resulting output is reliable. For aviation, validation is central to deployment. An apparently plausible answer can still be wrong, a recommended transaction may breach a commercial rule and an automated action may be inappropriate in the operational circumstances.

Discussions at T2RLEngage highlighted the difficulty of validating AI agents in real-world airline environments, where variability makes it challenging to establish that decisions will remain accurate and authorised across a wide range of circumstances.

One practical application is software testing. Modern airline retailing and modular offer-and-order systems can involve thousands of possible interactions and end-to-end scenarios. AI-assisted generation and execution of test cases can help expand the coverage of testing, reduce repetitive effort and make complex integration work more manageable. AI ultimately does not need to make final commercial or operational decisions to create meaningful value. It can improve the processes used to build, test and maintain the systems on which those decisions depend.

Agentic applications need similarly rigorous boundaries. Airlines should determine which actions an agent may perform independently, which require human approval, what evidence must be recorded and how the system should behave when data is incomplete or contradictory. The appropriate controls will differ by task. A low-risk customer enquiry is not equivalent to a booking change, a financial transaction or a decision that affects the safe and orderly operation of a flight.

Human expertise will therefore remain important, but its role may evolve. Rather than handling every routine step, employees may increasingly supervise automated processes, resolve exceptions and apply judgement where the system cannot confidently determine the correct response.

The objective is not maximum automation at any cost. It is dependable automation that can be measured, audited and improved.

The value is in the workflow, not the demonstration

The airline industry has no shortage of potential AI applications. Customer service, disruption recovery, revenue management, retailing, fraud detection, software development and internal decision support all offer opportunities to reduce friction or improve performance. The difficulty is deciding where to start.

A compelling demonstration can conceal the work required to integrate a new capability into day-to-day operations. A chatbot may answer questions impressively but still fail to complete a transaction. An AI agent may produce a sensible itinerary but be unable to verify availability, apply the correct conditions or manage a subsequent change.

The most credible deployments begin with a defined business problem rather than a general ambition to adopt AI. It is becoming clearer that airlines should identify a workflow in which delays, manual effort, fragmented information or inconsistent decisions create measurable costs. They can then determine whether AI can improve the outcome, what systems it must access and how success will be evaluated.

Modular deployment can help. Rather than attempting to automate an entire customer journey at once, an airline can introduce capabilities in stages, validate each one and expand when the evidence supports doing so. Specialist partners can provide expertise and reduce the burden of building every component internally, although airlines must retain sufficient oversight of how those components interact.

This is also a management challenge. AI initiatives can become isolated experiments owned by technology teams, disconnected from the commercial or operational processes they are meant to improve. Without clear accountability, a successful pilot may never progress into routine use.

The test should be straightforward: does the application reduce the time required to complete a task, improve the quality of a decision, lower the cost of serving a customer, create incremental revenue or make the operation more responsive? The measures will vary, and not every benefit will be immediately financial. But airlines need a credible method of distinguishing measurable improvement from technological novelty.

As Mr Clarke's observation in his T2RLEngage keynote suggests, the pace of development is accelerating. That makes disciplined prioritisation more important, not less. The danger is that airlines spend their time pursuing the latest capabilities without building the foundations needed to turn them into repeatable results.

The next competitive divide

AI will not create the same advantage for every airline. The opportunities available to a large network carrier with extensive data, specialist technology teams and substantial investment capacity will differ from those available to a smaller airline operating a simpler business model. Some carriers will build selected capabilities in-house; others will rely more heavily on technology providers. Some will focus on customer-facing agents and modern retailing, while others may achieve earlier returns from software testing, employee productivity or operational workflows.

There is no single adoption path. But there is a common requirement: AI must be connected to the business, not treated as a separate layer of innovation. That means investing in accessible and reliable data, modern interfaces, clearly defined commercial and operational rules, appropriate controls and the skills to manage systems that increasingly act rather than merely advise. It also means maintaining a clear understanding of who owns the customer relationship and who has authority over each transaction.

For airlines, the strategic question is not whether to pursue complete autonomy. It is where greater automation creates a defensible advantage, where human judgement remains essential and how to move between the two without introducing unacceptable risk.

The industry should also resist assuming that the organisation with the most sophisticated AI will necessarily win. If competing airlines can access similar underlying models, differentiation will depend on the quality of their data, the design of their workflows, the relevance of their offers and their ability to execute reliably.

AI capabilities will continue to evolve. The harder task is organisational: aligning technology, commercial strategy and operations so that those capabilities produce meaningful results.

Intelligence matters when it changes the outcome

Aviation has always adopted technology to improve prediction, efficiency and decision-making. What is changing is the potential to connect those capabilities directly to action.

The transition from artificial intelligence to automated impact is not simply another technology upgrade. It could change how airline products are sold, how passengers interact with carriers, how employees manage exceptions and how technology itself is developed. Agentic AI introduces a further commercial complication: the person buying the flight may no longer be the entity doing the searching, comparing or transacting.

That creates both opportunity and risk. Airlines that make their offers accessible to AI agents and connect intelligence to executable workflows could reduce friction, improve responsiveness and find new ways to create value. Those that fail to adapt may find that other organisations increasingly control the interfaces through which customers make decisions.

But the industry should not confuse speed of development with readiness for deployment. Legacy architecture, fragmented data, integration costs and the need for reliable validation remain substantial obstacles. The challenge is to introduce automation without losing control of the transaction, the customer relationship or the consequences of a decision.

The decisive advantage will therefore not belong automatically to the airline with the largest AI programme or the most impressive demonstration. It will belong to the airline that can identify where automation matters, integrate it into the fabric of the business and demonstrate that it improves the outcome.

As Richard Clarke observed, the AI story has shifted - and the pace is accelerating. For aviation leaders, the response must be more than experimentation. It must be a clear plan for turning intelligence into action, and action into measurable value.