The Next-Generation Commercial Engine

Why travel and hospitality brands need to rethink how they manage inventory, pricing, and selling, not just replace a legacy system with improved functionality.

May 2026

The following article has been developed with insights from Tim Davis MBE, Founder of PACE Dimensions, drawing on his experience leading commercial and technology transformation across the global travel and hospitality industry.

When to replace a CRS? The Wrong Question Has Been Asked for Too Long

There is a question that gets asked repeatedly across boardrooms and technology committees in travel and hospitality: when do we replace the CRS? It is usually followed by a risk assessment, a cost estimate and a decision to defer. The system is old. The interfaces are fragile. The project would be enormous. But it works, more or less, and there are more pressing priorities.

The problem is not that the question is wrong. The problem is that it is far too narrow. The central reservation system is not simply a legacy system that needs replacing with a better version of itself. It is the embodiment of a commercial logic that was designed for a world of stable room types, fixed rate plans, limited distribution channels and reservations as the dominant unit of commercial record. That world no longer exists. The question that needs to be asked is not when to replace the CRS. It is how to rethink, entirely, the way travel and hospitality companies manage what they sell, how they price it, and how they make it visible and accessible to the consumers and systems that are now doing the buying.

This is a question with urgency attached to it. As explored in our analyses Strategic Imperatives and Winning Ways for 2026¹ and Who Owns the Guest in the Age of AI?², the commercial environment in travel is being reorganised around AI-mediated discovery, real-time offer assembly and machine-readable distribution. As Skift reported in April 2026³, Amazon, Meta and Google are each taking a different approach to buildingtheir AI travel planning systems. Achieving visibility and integration on one platform is not transferable to another. Each new AI channel has its own architecture, its own data requirements and its own standards. Hotel and travel companies that have not reengineered their commercial foundations will find they cannot serve any of these channels well, let alone all of them simultaneously.

This article is a practical guide to what that rethinking looks like. It sets out why the CRS is structurally insufficient for the world it now operates in, what a next-generation commercial engine actually consists of, what the architectural and organisational implications are, and how companies can move from where they are to where they need to be without the whole transformation becoming another deferred, risk-laden programme that never delivers.

Understanding What the CRS Actually Does and Why That Is the Problem

To understand what needs to change, it helps to understand precisely what a CRS does and, more importantly, what it does not do.

A central reservation system homogenises inventory. Every hotel in a brand or distribution network is different. One property has corner rooms with exceptional views. Another has rooms near the lift that some guests prefer and others avoid. Room dimensions, bed configurations, floor positions, proximity to amenities, views, specific furnishings: each physical room has dozens of attributes that matter to different guests in different ways. The CRS cannot hold all of that. Instead, it collapses the infinite variety of individual rooms into a simplified set of room types — Deluxe, Standard, Junior Suite — defined by the lowest common denominator of what all rooms in that category share. The consequence is that when a consumer queries the system, what they get back is not a picture of the specific room they might occupy. It is a generic description of a category, illustrated by a generic photograph, priced according to hard-coded rate rules. If they do not like the options returned, there is only one thing they can do: start again.

“Every CRS in the industry gives you the same thing: a room type, a bed type and a rate. That is the entirety of the commercial decision it enables. What the room is actually like, whether the price reflects your situation, what you would genuinely pay for — the system cannot see any of that. And increasingly, neither can the AI agents querying it on behalf of travellers.” — Tim Davis MBE, PACE Dimensions

Pricing in the CRS is equally constrained. A rate is not a simple number. It is a bundle of conditions: how far in advance the booking is made, what deposit is required, what the cancellation conditions are, whether breakfast is included, whether executive lounge access is part of the deal, whether the rate applies to a minimum stay. There are dozens of legitimate pricing variants for any given room on any given night, each reflecting a different combination of guest commitment, revenue management intent and product inclusion. The CRS manages these through hard-coded rate plans. Adding a new rate variant, adjusting a condition or responding to a shift in market dynamics requires a rate plan change, which takes time to implement and often requires touching multiple dependent systems simultaneously.

Availability adds a further layer. Revenue management controls what the CRS surfaces in response to any given query. The booking window, the day of arrival, the length of stay, the source of the booking: each of these variables can alter what the system shows. A consumer searching for a room two weeks out on a Friday arrival might see entirely different availability to one searching three months out on a Tuesday. This logic is valuable, it is how hotels protect their peak periods and manage demand, but it is determined by the Revenue Management System (RMS), which in turn operates in ways that are opaque to other systems and entirely invisible to AI agents trying to query inventory in real time.

The cumulative effect of all of this is a system that was built to manage a limited number of distribution channels and a simplified view of inventory. It was not built to be read by machines. It was not built to support attribute-based selling. It was not built to assemble a personalised offer for a specific consumer in a specific channel at a specific moment. And it was not built to capture the behavioural signals that reveal how a consumer actually shops, what they value, and when they are most likely to buy in the future.

“The CRS was designed for a world where you defined your options, published them to a limited set of channels, and waited for a booking to arrive. In a world where AI agents are assembling offers on behalf of consumers in real time, that model does not just underperform. It becomes invisible.” 

What the Commercial Engine Is and What It Does Instead

The successor to the CRS is a commercial engine: a modular, layered set of technology services that separates commercial decision-making from operational execution, stores data at the most granular level available, and exposes what it knows to every channel, every interface and every AI agent through governed APIs. The difference from the system it replaces is not incremental. It is architectural.

Understanding what this engine consists of in practical terms requires unpacking five core capabilities that work together as a system.

A product and inventory model that describes what is actually being sold

The first capability is a product model that represents inventory at the most granular level that is commercially relevant. Not room type. Room. Specific room, on a specific floor, with a specific view, with a specific bed configuration, with specific furnishings, facilities and attributes. Price is treated with the same granularity: not a rate plan, but a set of conditions, each of which can be varied independently. Deposit requirement, cancellation policy, inclusions, minimum stay, loyalty eligibility, channel constraints: each is an attribute of the price, not a property of a bundled rate code.

This matters for two reasons. The first is commercial: when a guest can see exactly what they are getting and select precisely the combination of attributes they value, they pay more willingly and with greater confidence. The second is structural: AI agents can only evaluate and present what they can read. An agent querying inventory to assemble a response to a natural language request from a traveller needs to be able to understand the full range of what is on offer, in terms the agent can process and compare. A room type with a rate attached is not sufficient. An attribute-rich offer model is.

An offer engine that manufactures rather than publishes

The second capability is the most significant conceptual shift. The legacy CRS publishes availability and prices. The commercial engine manufactures offers. These are fundamentally different activities.

Publishing means making a fixed set of options visible to channels: here are the rates, here is the availability, take it or leave it. Manufacturing means assembling, in real time, the specific proposition that is most appropriate for this consumer, in this channel, at this moment, given what the system knows about their identity, their preferences, their loyalty status, their history and the current commercial priorities of the business. Revenue management informs this layer, providing forecasting, displacement logic and optimisation outputs. But revenue management answers what to optimise for. The offer engine answers what to present, to whom, in which channel, with which conditions, right now. The gap between those two questions is where most of the commercial opportunity in hospitality currently sits unrealised.

“Most revenue management systems are optimising for occupancy and rate. That is the right question for a CRS world. The offer engine asks a different question entirely: what should we show this person, right now, in this channel, given everything we know about them? That shift is where the commercial opportunity is and the industry has barely started to exploit it.” 

A real-time availability and commitment layer

The third capability is an availability layer that operates in real time across every channel simultaneously. Legacy systems handled availability control reasonably well when the channel set was stable and the booking pace was predictable. Neither condition holds today. Direct channels, OTAs, metasearch, loyalty redemption flows, group bookings, corporate negotiated rates, partner bundles, conversational AI interfaces and agentic booking pipelines all need to query and commit against the same inventory pool without latency or inconsistency. The batch process — update availability once a day, push it to channels — cannot support this environment. Availability must be event-driven, instantly updated and able to maintain commercial integrity across a continuously expanding set of access points.

Order management, not just reservations

The fourth capability addresses what may be the most consequential conceptual gap in the current commercial stack. A reservation is a record of a booking. An order is a record of a commercial commitment across its entire lifecycle: the initial search, the selection, the booking, every subsequent modification or cancellation, every ancillary added or removed, the fulfilment obligations, the payment, the servicing and the final settlement. A reservation record captures what was booked. An order captures what was promised, what changed, what was consumed and what still needs to be delivered.

This distinction matters because personalisation, upselling, loyalty treatment and service recovery at scale all depend on a system that understands the full commercial history of a guest relationship, not just the most recent transaction. When a guest modifies a booking and then cancels a room upgrade, that event sequence is commercially meaningful. The current industry infrastructure throws it away. A commercial engine captures and retains it as part of the order record.

“When a guest changes a booking, upgrades a room, then cancels the upgrade, that sequence is commercially meaningful. It tells you something about what they value, what they were willing to pay for and where they hesitated. The current infrastructure throws all of that away. We have been running on transaction records when what we need is a complete commercial history.” 

A distribution layer that machines can read and act on

The fifth capability is where the strategic implications of AI become concrete. The Model Context Protocol, introduced in late 2024 and now adopted across the major AI platforms, is establishing itself as the standard through which AI agents access, query and transact against supplier systems in real time⁴. Booking.com and Expedia were among the first commercial partners when ChatGPT introduced MCP-based apps in late 2025, an early signal of how OTAs are positioning themselves as the AI-ready interface for suppliers who are not⁵. As analysed in The Battle for the Consumer, the commercial logic of every major intermediary is to insert itself between hotel brands and the next generation of consumer technology — AI distribution is simply the latest and most consequential front on which that battle is being fought. For hotel and travel brands, the implication is direct: if the commercial engine cannot expose products, pricing logic, availability, policies and order actions through machine-readable, governed APIs, it will not appear in the AI-mediated channels that are growing fastest. IDC predicts that by 2030, 30 per cent of travel bookings will be executed by AI agents, and that brands with incomplete or fragmented data will effectively disappear from agent decision sets⁶.

“The commercial engine does not replace the CRS with a better version of the same thing. It replaces the logic of the CRS: from publish-and-wait to assemble-and-respond, from static rates to manufactured offers, from opaque availability to machine-readable inventory.” 

What This Demands of Technology Architecture

Building a commercial engine of this kind requires a technology architecture that most hospitality businesses do not currently have. The fragmented system environment described in the previous section is the starting point, not the solved problem. As examined in Who Owns the Guest in the Age of AI?, the architecture challenge is not simply a matter of upgrading individual systems. Three foundational shifts are required that the individual systems themselves cannot provide. 

The architectural shift required has three essential characteristics. 

The first is a single data architecture. Not a data warehouse that stitches together outputs from different silos after the fact, but a canonical data model that defines what a product is, what an offer is and what an order is, at the most granular level, in a way that every system in the organisation can align behind. Every other capability depends on this foundation. Without it, the offer engine assembles offers from inconsistent inputs. The availability layer queries conflicting inventories. The order record contains gaps. The AI agent receives fragmented responses and either guesses or omits.

The second characteristic is a layered service architecture. The commercial engine builds capabilities at the most atomic level — services that do one specific thing precisely and expose that capability through a governed API. These atomic services are then combined into composite layers that solve more complex commercial problems, such as assembling an offer or managing a modification flow. The benefit is modularity: any layer can be changed or upgraded without cascading changes through the entire stack. This is the opposite of the monolithic CRS, where a change to one element requires testing and validating its impact across every dependent component.

The third characteristic is event capture at every touchpoint. A booking is a transaction. A check-in is a transaction. But the search that preceded it, the filters applied and removed, the rate options examined and rejected, the point at which the guest abandoned the booking flow: these are behaviour. So too are service requests made during a stay, complaints lodged, and responses to surveys. Each of these signals something meaningful about who the consumer is and what they value. Behaviour is what the commercial engine uses to build a progressively richer picture of each consumer and when they are most likely to buy in the future. Capturing every event and streaming it to a unified customer intelligence layer is what transforms a commercial platform from a transaction processor into a learning system.

“A platform that only captures transactions is blind to behaviour. And behaviour is the only thing that tells you who the customer actually is, what they genuinely value, and when they are most likely to buy. Without it, personalisation is just guesswork dressed up as intelligence.” 

What It Means for the Systems That Run the Business Today

A transformation of this kind does not require trashing every operational system in the business. It requires separating control from execution. The commercial engine becomes the control plane: the single authoritative source of commercial decisions about what is for sale, on what terms, to whom and through which channel. The operational systems remain, but their role changes. They execute. They do not decide.

The practical approach is what might be called “surround and erode”. Rather than replacing hotel operating systems in a single high-risk cutover, the commercial engine is built above and around them. Channels and customers are abstracted from direct system access. When a consumer makes a query, whether through a website, a mobile app, a call centre agent or an AI interface, the query flows through the commercial engine, which determines which underlying systems to consult, assembles the response and returns it through a single consistent interface. Hotel staff also use access systems through an abstracted layer. Over time, as the commercial engine grows in capability and confidence, the functionality of the underlying hotel systems is progressively eroded. Gradually and deliberately, the dependency on legacy system logic diminishes. There is no midnight switch from old to new. There is a managed transition in which risk is distributed across time rather than concentrated in a single cutover event.

The PMS continues to handle what it does well: operational execution, room assignment, check-in and check-out, folio management and housekeeping. It is no longer the commercial hub. Commercial decisions are no longer deferred to whatever the PMS happens to expose.

Revenue management evolves from a stand-alone pricing system into a decision science layer within the commercial engine. It operates in real time rather than in nightly batch cycles. It manages attributes rather than room types. It feeds pricing intelligence, displacement logic and optimisation outputs into offer creation rather than hard-coding rate plans into a CRS.

Content management becomes structured and linked directly to the product and offer model. Every piece of content, text, image, video, review, is tagged to the inventory attributes it describes, kept current, and made accessible through APIs to every channel and AI agent that needs it. The distinction between content that is machine-readable and content that is not is, in a world of AI-mediated discovery, the distinction between commercial visibility and invisibility.

The result, over time, is a single source of commercial truth: one authoritative view of what the business has to sell, what it is worth, who can buy it and on what terms, available to every channel and every system through a consistent, governed interface.

The Pathway: Building Progressively, Not All at Once

Every executive who has lived through a large-scale technology programme in this industry has a story about the one that ran over budget, over time, and delivered its value so late that the business case was obsolete by the time the system went live. The pathway to a commercial engine must be designed to avoid that outcome. The key principle is that each phase must deliver stand-alone commercial return, and that return should compound with each subsequent phase. Consider the following key phases:

1. Architectural design before any code is written or any contract signed. Define the layered target architecture, establish the canonical product-offer-order data model, and set the API and event standards that every subsequent component will conform to. This is the design work that prevents expensive rework. Without it, every tactical investment risks becoming another silo.
2. Abstract channels and customers from existing hotel systems. Build the layer through which all commercial interactions flow, even if all that layer initially does is pass queries through to the existing CRS. The value is not in what the layer does on day one. It is in the fact that it is now in place, ready to be made progressively more capable without disrupting what already works.
3. Establish event capture and API discipline. Every commercial event — shopping, searching, booking, making a complaint, responding to a survey — is captured and streamed to a unified data layer. Every service is exposed through a governed API. This is the infrastructure that makes machine readability, real-time personalisation and AI-driven offer assembly possible. It also begins generating the behavioural data that will make every subsequent phase of the commercial engine more effective.
4. Redesign the operating model in parallel with the technology build. Commercial, revenue management, distribution and property technology teams must move from operating around separate systems with separate metrics to working from a shared commercial architecture with shared definitions of success. Technology without this organisational alignment will underdeliver. The operating model change is not the last thing to address. It is something that must be designed and managed from the beginning.
5. Migrate pragmatically. Wrap legacy systems where necessary. Extract high-value capabilities first, starting with the components that generate the most immediate return or that are most exposed to competitive risk from AI-driven distribution changes. Each phase should deliver a specific commercial benefit that can be measured: higher conversion from a richer product model, lower cost of sale from better offer targeting, improved AI visibility from machine-readable content, faster response to revenue management signals from real-time availability control.

The compounding logic is important. Phase one’s benefit is standalone. Phase two builds on phase one and amplifies it. By phase three, the foundations are producing halo effects across everything that was built earlier. The sequencing is not just about managing risk. It is about accelerating returns.

“The goal is not a programme that delivers value at the end. It is a series of investments, each of which pays back on its own terms, and each of which makes the one before it more powerful.” 

The Commercial Case: What Is Gained and What Is Lost

The commercial upside of a well-built commercial engine is substantial and multi-dimensional. Attribute-based selling captures willingness to pay more precisely, improving revenue per available room without requiring rate increases. A real-time offer engine improves conversion by presenting the right proposition to the right consumer at the right moment. Machine-readable distribution expands access to the AI-mediated channels that are growing fastest. A unified order and customer intelligence layer deepens the data advantage that enables increasingly precise marketing, personalisation and retention.

The competitive dimension is equally important. The OTAs and major technology platforms understand what is happening. Each is racing to position itself as the integration layer between hotel inventory and the new wave of AI-driven discovery channels. The brands with the richest, most granular, most accessible commercial data will have the structural advantage. Those without it become dependent on intermediaries who have built it on their behalf, on terms they do not control.

The cost of inaction is not standing still. The commercial environment is moving regardless. Brands that do not build machine-readable inventory will be less visible to AI agents. Brands that do not manufacture offers will be outcompeted by those that do. Brands that do not capture behavioural events will make slower and less accurate commercial decisions than those that do. The gap between the leaders and the laggards in this transition will not be small and will compound over time, for exactly the reason set out in The Global Hotel Industry in 2026: Discipline, Data and Differentiation: in flat growth environments, data transforms from supporting analytics into a competitive weapon. The brands with the richest, most current commercial data will compound advantage faster than any competitor can match simply by spending more.

Key Takeaways

Step back from the technology detail and the big picture is straightforward, if uncomfortable. The travel and hospitality industry has spent decades building systems that manage inventory, pricing and distribution in a particular way. That way was designed for a world of limited channels, simplified room categories and static rate plans. The systems evolved incrementally, each adding more interfaces and more complexity, but none of them challenged the underlying commercial logic they were built to serve. That logic is now the problem.

What is required is not a better version of those systems. It is a rethinking of how inventory is defined, how it is priced and how it is sold. That is a different kind of change. It affects not just technology but the commercial operating model of the entire business: how revenue management, distribution, sales, property operations and digital teams define their roles and measure success. Companies that treat this as a technology refresh will arrive at the wrong answer. Companies that treat it as a commercial reinvention will arrive at the right one.

“This is not a decision that can be delegated to the IT department. It is a commercial decision that has to be owned by the business. Companies that understand that will invest at the right level and in the right sequence. Companies that treat it as a technology upgrade will underinvest and wonder why it did not work.” 

This is not a choice. The commercial environment is being reorganised around AI-mediated discovery, machine-readable inventory and real-time offer assembly. Brands that do not adapt will not decline gradually. They will become progressively invisible to the systems that increasingly mediate traveller decisions. The question is not whether to act. It is whether to act early enough to build advantage or late enough to be playing catch-up at the moment of greatest competitive pressure.

The advantage for those who move decisively will be disproportionate. This is not an industry where being a fast follower is a reliable strategy. The brands that rethink their commercial logic first will compound the advantage with every subsequent investment. Better product data makes offers more relevant. More relevant offers improve conversion. Better conversion generates richer behavioural signals. Sharper signals improve revenue management. The flywheel does not slow down for late movers. It accelerates away from them.

The change must join up functions that have long operated separately. The CRS and the systems around it did not just define technology. They defined organisational boundaries. Revenue management owned pricing. Distribution owned channels. Property operations owned the guest. The commercial engine does not respect those boundaries, because the commercial opportunity does not respect them either. Building it requires commercial, revenue, distribution, property and digital teams to work from a shared architecture with shared definitions of what success looks like. That is the organisational transformation that makes the technology transformation pay back.

The pathway rewards patience and penalises impatience equally. Attempting to replace everything at once is how transformation programmes fail in this industry. Building incrementally, with each phase delivering a measurable commercial return that funds the next, is how they succeed. The sequencing is not just about managing risk. It is about generating the internal credibility and commercial proof that sustains commitment through a multi-year transition. Each investment should make the previous one more powerful. If it does not, the sequencing is wrong.

About PACE Dimensions

PACE Dimensions is a research and consulting firm founded in 2010 with deep industry experience and a practitioner’s expertise in helping Travel and Hospitality companies excel through strategic clarity and operational excellence. The firm specialises in translating market insights and strategic imperatives into practical initiatives that deliver measurable performance improvement. Its consultants bring proven track records of success working with hotel groups of all sizes across upscale and luxury segments, combining rigorous analysis with pragmatic implementation approaches that drive sustainable results.

References

¹  PACE Dimensions (2026). Focus Areas in a Flat Growth Environment: Strategic Imperatives and Winning Ways for 2026.  https://pacedimensions.com/strategic-imperatives-winning-ways-2026/

²  PACE Dimensions (2026). Who Owns the Guest in the Age of AI?  https://pacedimensions.com/who-owns-the-guest-in-the-age-of-ai/

³  Skift (2026). Travel Is Facing a New Test: AI Fragmentation.  https://skift.com/2026/04/10/travel-is-facing-a-new-test-ai-fragmentation/

⁴  Skift (2025). MCP Explained: The AI Standard Reshaping Travel Tech.  https://skift.com/2025/12/23/mcp-explainer-travel-ai-agentic/

⁵  PhocusWire (2025). How MCP Could Reshape Travel.  https://www.phocuswire.com/how-mcp-could-reshape-travel

⁶  IDC (2026). FutureScape: Worldwide Hospitality, Dining, and Travel 2026 Predictions.  https://www.idc.com/resource-center/blog/agentic-ai-will-redefine-travel-and-hospitality-in-2026/

⁷  McKinsey (2025). Remapping Travel with Agentic AI.  https://www.mckinsey.com/industries/travel/our-insights/remapping-travel-with-agentic-ai

⁸  HTNG / AHLA (2025). Attribute-Based Selling and the OpenTravel 2.0 Object Model.  https://www.ahla.com/htng-completed-workgroups