AI In Travel, The Translation versus Transformation Gap

Previous articles in this series have explored what agent workflows and agentic travel technology requires: operational excellence as a competitive moat, architecture shifting from broadcast optimisation to precision verification and the critical need for coordination layers that enable agent discovery and trust.

Examining what should happen is only half the story. The technology exists today to build comprehensive agent infrastructure - verified data, transparent pricing, guaranteed outcomes, measurable reliability - but the path forward is obstructed.

This article explores why market forces make full Business Model Transformation unlikely, even though the technology exists and the benefits are obvious. Understanding why transformation probably will not happen through natural market evolution helps in making better strategic decisions about where to invest, what to build and which future to prepare for.

The Critical Distinction

The travel industry is building one type of infrastructure whilst agents actually need something fundamentally different. Understanding this gap is the key to making sense of where the market is heading.

Translation Infrastructure makes existing systems agent-accessible. This includes better APIs, LLM-friendly content and conversational interfaces. It is about helping agents query what exists without changing what those systems can actually guarantee.

Transformation Infrastructure rebuilds the foundation. This includes verified data, transparent total pricing, guaranteed outcomes and measurable reliability. It is about creating systems where agents can make autonomous decisions with confidence because the underlying infrastructure supports that level of trust.

Translation infrastructure teaches agents to speak the language of current systems. Transformation infrastructure rebuilds those systems to speak the language agents need. The entire market opportunity lives in that gap.

Why Translation Infrastructure Hits a Ceiling

To understand why this matters, consider a concrete example of what translation infrastructure achieves and where it reaches its limits.

A hotel website today might say, "Oceanfront rooms from $199, subject to availability."

With modern optimisation tools such as Natural Language Web formatting, Schema.org markup, embeddings optimisation, agents can parse this data more easily. All of that is genuinely useful; it makes the agent's job easier by surfacing relevant options.

However, here is what has not changed. The agent still cannot verify whether "oceanfront" means direct water view from the balcony versus water visible only if leaning far out the window. It cannot confirm whether the $199 rate is actually available for the traveller's dates, or what additional, unavoidable fees will appear at checkout/payment (drip pricing). It cannot guarantee the room shown in photos matches what will be provided, nor can it access historical data on whether this property consistently delivers on its promises.

Standard website booking workflows handle availability and price guarantee only at shopping cart checkout or payment. The gap is in meaningful comparison before booking. Without standardised definitions, quantified loyalty values and verified amenities, agents cannot optimise for total traveller value; they can only execute technically available bookings.

Translation tools help agents work with existing data, making it more discoverable and better formatted for LLM consumption. What they do not do is make the data verifiable, accurate or complete. Agents get better access, but they still cannot guarantee outcomes.

The Cognitive Load Question

This limitation prevents the closure of the 35x trust gap (70 percent of travellers express interest, but only 2 percent grant full booking autonomy).

Translation infrastructure serves the traveller who wants the agent to curate a shortlist of options, reducing research time whilst maintaining final decision control. The traveller still evaluates and chooses because the agent cannot definitively answer "which delivers best value to your overall needs."

For travellers willing to grant full autonomy - the core of the transformative vision - translation infrastructure hits a ceiling. Without verifiable, consistent outcomes, travellers maintains cognitive load rather than delegating fully.

The transformative vision of truly autonomous agents optimising comprehensively on behalf of travellers requires infrastructure that makes optimisation calculable rather than subjective. As the current market reveals, the path of least resistance is not building that infrastructure.

What Is Actually Being Built (The Evidence)

A walk through recent market announcements reveals a clear pattern. The industry is overwhelmingly committing to translation infrastructure.

The Pattern of Incremental Adaptation

Google, through its AI initiatives, has announced agentic booking partnerships with major incumbents including Booking, Expedia, Marriott and Wyndham. Similarly, OpenAI launched its app ecosystem featuring major travel platforms such as Booking, Expedia and TripAdvisor. OTA Platforms make natural language search available and enter into social media channels to capture more traffic at the discovery opportunity. Payment providers are building agentic ready payment protocols and interfaces.

Notice what is happening here:

1. Power Consolidation. Existing inventory relationships are being extended into the agent world, not disrupted by it.

2. Focus on Curation. Initiatives like transforming social posts into itineraries focus on the 'better search and curation' segment - improving the front-end experience - not the guaranteed outcomes required for full delegation.

3. Payment Protocols.  Stripe/OpenAI (Agentic Commerce Protocol - ACP) and Google (Agent Payments Protocol - AP2) are prioritising open standards for frictionless transaction completion in chat. These protocols solve the payment initiation problem but do not mandate the data integrity or guaranteed outcome systems required for transformation.

The operational layer, too, confirms this pattern. Firms like DerbySoft are building highly valuable AI solutions, such as their AI Voice Agent to automate manual phone and email confirmations for corporate bookings. They are solving genuine pain points - manual handling, long commission reconciliation cycles - but these efforts are categorised as Process Transformation. They automate existing, fragmented workflows rather than rebuilding for a foundational, agent-ready verification layer.

The consistent pattern is the industry is optimising for agent visibility, process efficiency and transaction friction reduction within the current system, not rebuilding for agent requirements (verified, guaranteed outcomes).

An Emerging Counter-Signal

A notable exception to the pattern is the move by fintech giants. Revolut's acquisition of Swifty, an AI travel agent incubated at Lufthansa Innovation Hub, signals a vertical integration attempt by a player whose core competency is payments and financial integrity. As Revolut does not carry the burden of OTA steering revenue, it is better positioned to attempt a transformation play that links reliable payments/FX with agent-driven booking. This move is significant, but currently remains a counter-signal to the dominant trend.

The Strategic Conclusion from the Evidence

If agents capture 15-20 percent of routine bookings whilst working with current infrastructure, delivering profitable convenience services without requiring transformation investment, does anyone actually have an economic incentive to invest billions in transformation infrastructure?

The market finds equilibrium where translation infrastructure proves "good enough" for acceptable returns across all parties. The uncomfortable insight is that the transformation might not be necessary for the market to function profitably.

Why Transformation Infrastructure Will Not Be Built

Now to the heart of this post. Why does someone not just build the transformation infrastructure agents actually need? The technology exists, the requirements are clear and the benefits are obvious. Yet three interconnected market forces prevent transformation from happening through natural evolution.

A. Business Model Protection (The Rational Choice)

Current OTA platform business models depend on exactly the practices that transformation infrastructure would eliminate. The entire infrastructure is built around conversion optimisation and steering revenue, relying on practices like promoting preferred partners, dynamic pricing and drip pricing (fees revealed after engagement).

The rational response is to build just enough Translation Infrastructure (conversational interfaces, payment protocols like ACP/AP2) to appear agent-ready without threatening core revenue streams.

The move to agentic commerce exposes a core barrier: the operational jump from a deferred liability model to an instantaneous, automated compensation protocol.

The agentic world demands this clarification because autonomy requires algorithmic trust.  When a human traveller books, they tolerate deferred liability (the multi-day refund cycle) because their own human judgment approved the initial transaction and they can engage in the human-mediated process of negotiation and reconciliation when failure occurs. An autonomous agent, however, cannot wait. If an agent executes a booking based on data provided by a platform and that booking immediately fails or an issue comes to light after payment has exchanged, the agent's utility drops to zero. To maintain the agent's core function, which is to act continuously and autonomously in the traveller's best interest, the platform must provide instantaneous recourse. The platform must act as a Guarantor, ready to secure rebooking or issue compensation immediately. Without this algorithmic guarantee, the agent cannot be trusted to operate without human oversight and the entire transformative vision of full delegation collapses.

This leap requires instant liquidity (dedicated, liquid reserves), auditable governance (clear liability boundaries) and an evolution of the treasury operation. Building this liquid, auditable infrastructure takes substantial capital and exposes existing data quality failures, making incremental adaptation the safer, more profitable choice.

B. The Coordination Trap (Why Competition Does Not Fix This)

The second force preventing transformation is the classic chicken-and-egg problem. Agents need comprehensive, verified inventory; platforms have no incentive to rebuild infrastructure until agents drive meaningful volume. Neither side moves first.

Travel's acute market fragmentation (thousands of suppliers, multiple channels) prevents collective action on standards. This problem is compounded because most suppliers cannot yet deliver the Transformation Infrastructure requirements (real-time PMS integration, structured data compliance, TRUV-ready pricing).

Crucially, the ecosystem lacks a sufficiently capitalised, neutral entity with the incentive to build and monetise the foundational verification layer itself. In other fragmented sectors, a large, non-conflicted intermediary often internalises the cost of building this required utility layer. The travel market structure currently prevents this emergence, as any such effort immediately faces insurmountable challenges in establishing credibility, network scale and a sustainable cost model against incumbents motivated to protect existing steering revenues.

Executives look at the cost of supporting agents while operating with dramatically smaller inventory (Tier 1 verified suppliers only) and rationally choose to sweat existing assets through Translation Infrastructure instead.

C. Power Structures Reasserting (Why Incumbents Win)

Traditional competitive advantages are not eroding in the agent era they are compounding.

• Network Effects are Strengthening. AI platforms prioritise partnering with platforms (OTAs, Shopify, etc) that immediately solve the supply-side cold-start problem. High uptake on a marketplace validates and locks in the initial OTA partners. Power consolidates at these critical coordination points.

• Scale Economies are More Pronounced. Building B2A (Business-to-Agent) integration is expensive. Only large players can afford to integrate simultaneously.

• Supplier Switching Costs Persist.  Suppliers are locked into incumbents by years of integrated data and operational processes.

• Incumbents Block the New Cornered Resource. The most valuable asset, market-wide semantic, auditable performance data and history is held by incumbents, who maintain advantage by keeping this data unverified and proprietary.

The strategic question is whether B2A distribution becomes another concentration mechanism that reinforces existing power structures rather than disrupting them. Current evidence strongly suggests this is the outcome.

The Narrow Paths to Transformation

Three possible forcing functions could override market dynamics and drive transformation despite rational business model protection.

Scenario A: Regulatory Intervention

Systemic, high-profile agent booking failures damage consumer trust activating existing consumer protection laws. This would force industry-wide data accuracy requirements with real liability. It is unlikely unless agents reach sufficient scale to cause mass consumer harm.

Scenario B: New Entrant with Capital and Counter-Positioning

A well-capitalised start-up builds agent-native infrastructure from scratch with verified inventory, demonstrating measurably superior outcomes. This could work because fragmentation prevents coordinated defence, but it faces the classic chicken-and-egg problem with extreme capital burn.

Scenario C: AI Platform Competition Drives Arms Race

AI platforms (Google, OpenAI) discover that booking failure rates severely damage their core product reputation, forcing them to mandate verification standards from suppliers. This is unlikely as AI platforms are currently partnering with incumbent OTAs and prioritising payment protocol alignment (ACP, AP2) rather than forcing costly data transformation.

The Honest Probability Assessment

• 70 percent probability - Incremental Adaptation. The market settles into "agents as better interface" powered by translation infrastructure.

• 20 percent probability - Selective Transformation. Specific, high-margin segments (luxury, corporate travel) transform where premium pricing supports infrastructure investment (e.g., the potential target for a Revolut-like integrated solution).

• 10 percent probability - Full Transformation. Requires external pressure powerful enough to overcome business model protection and coordination failure.

5. Market Signals and Strategic Positioning

The key is not to predict the future, but to build the adaptive capacity to respond as the market reveals its answer. Focus on signals that would actually shift the 70/20/10 probability distribution.

Signals That Actually Matter (The Diagnostic Framework)

Signal Type	Incremental Adaptation Signals (Confirming 70 percent Trajectory)	Transformation Signals (Would Indicate a Shift)
Partnership	OTAs launching proprietary agent tools without neutral verification. Open standards focused on payments (ACP, AP2) without data mandates.	AI platforms acquiring travel infrastructure start-ups rather than just partnering. Fintech/Payment leaders (like Revolut) vertically integrating agent services and supply.
Infrastructure	Continued focus on interface improvements and using agents as a bridge to acquire new customers. Content optimisation (AEO) services thriving.	Independent verification services emerging as a separate infrastructure layer and raising significant capital.
Accountability	Steering revenue persisting under different mechanisms.	Regulatory enforcement actions with real financial liability for unverified outcomes. Agent-ready platforms publishing auditable trust and reliability rankings.

Strategic Positioning by Stakeholder

If Operating an OTA Platform

• The Hedge. Invest in incremental translation infrastructure (ACP/AP2 adoption, AEO tools).

• The Option. Maintain a small, transformation infrastructure team.

• Watch. Capital flows into independent verification services and the success of vertically integrated players like Revolut.

If Operating as a Supplier (Hotel Chain, Wholesaler)

• The Investment.  Invest in operational excellence that is algorithmically verifiable regardless of which future emerges.

• The Experiment.  Experiment with agent-ready direct distribution using premium inventory.

• Avoid.  Expensive transformation infrastructure builds without validated agent demand.

If Building Agent Platforms or AI-Mediated Travel Services

• The Acceptance.  Accept that working with imperfect infrastructure for 3–5 years minimum is required.

• The Strategy.  Build architecture that could leverage transformation infrastructure if it emerges.

• The Lever.  Consider lobbying industry and regulatory frameworks mandating minimum data accuracy standards.

The Power of Knowing Why

The transformation might not happen. It probably will not. Market forces - business model protection, coordination failure and power consolidation - prevent the transformation infrastructure from being built through natural evolution.

The most likely outcome is incremental adaptation using agents as a better interface, delivering convenience value without genuinely optimised outcomes. The market functions profitably in this 70 percent equilibrium.

However, understanding why the gap exists is strategically essential. Knowing why transformation is unlikely makes effective operation in the reality that actually exists possible. The utopian vision serves as the necessary benchmark against which innovation can be measured.

The challenge is not predicting the future; it is building the capability to respond effectively. It is to thrive in the 70% world while maintaining the option to pivot if the 30% scenario (signalled by more moves like the Revolut acquisition) emerges.

---

The Builder's Dilemma

This analysis emerged from personal exploration, not academic distance. I have been working through a fundamental question, "Can transformation infrastructure be built despite the rational market forces favouring incremental adaptation?" The data suggests Translation wins with 70% probability. Business model protection, coordination failure and power consolidation create formidable barriers to Transformation Infrastructure being built through natural market evolution. Yet the opportunity keeps pulling me back. The 35x trust gap (70% of travellers want autonomous agents, only 2% trust them) represents potential category creation, not just market share gains. The first player who solves verified, instant-liability booking does not compete for existing distribution, they define a new category entirely. This creates the classic start-up tension: recognising massive barriers whilst believing the opportunity justifies attempting to overcome them.

Three questions guide my continued exploration:

1. Can a well-positioned player force transformation despite rational incumbent resistance? 2. What would minimum viable transformation infrastructure actually look like?

3. What would make transformation infrastructure compelling enough for early adoption?

I am actively exploring these questions. This analysis represents my current understanding of why transformation is extraordinarily difficult, not my conclusion that it should not be attempted. If you're working on transformation infrastructure, wrestling with similar questions or have perspectives on what I'm missing, I'm interested in your thinking. The most valuable feedback is not agreement with my analysis, it is insights that challenge my assumptions or reveal opportunities I haven't considered.

Connect with me on LinkedIn or reach out directly on my Contact Page to continue this conversation.