Ascott's AI-Ready Infrastructure Bet: Rebuilding Inventory and Core Systems So AI Agents Can Transact

Agent readiness starts in the core systems

Ascott advances AI-ready strategy to support agentic commerce

AI in hospitality drives Ascott strategy with Accenture, Amadeus and EHL, enabling agentic commerce and AI-led travel booking.

www.traveldailynews.asia

On April 23, 2026, Singapore-headquartered Ascott Limited announced an AI-ready infrastructure investment framed explicitly around agentic commerce. The lodging business unit of CapitaLand Investment operates more than 1,000 properties across 14 brands in over 230 cities and more than 40 countries, making it one of the world's largest serviced residence operators. The announcement rests on three partnerships, with Accenture, Amadeus, and EHL Hospitality Business School, rebuilding digital architecture, distribution, and people capabilities in parallel.

Agentic commerce refers to the model in which AI agents carry out discovery, comparison, booking, and purchase on a customer's behalf. Kevin Goh, CEO of Ascott, put the posture plainly: "Instead of waiting to see how agentic AI plays out in travel, we are building the infrastructure to shape how it does." The declaration draws attention, but the substance is more telling. The money is going not into a visible chat surface but into the core systems that connect reservations, inventory, and membership.

This article reads the announcement not as a hotel-industry item but as a design problem every merchant now shares: how do you build a product foundation that AI agents can transact with. Swap hotel rooms for products, availability for real-time stock, and the loyalty program for customer data, and the questions line up almost exactly with those facing retail and e-commerce teams.

Attribute-based inventory as the machine-readable shelf

Traditional hotel reservation systems express inventory as fixed combinations of room types and rate plans. For a portfolio as varied as Ascott's, spanning serviced residences, hotels, resorts, and social living properties across every length and purpose of stay, that thin vocabulary translates directly into lost selling opportunities. As AI agents become a primary interface for travel discovery and planning, the question is whether inventory can be exposed in a form agents can evaluate correctly.

Ascott's answer is to implement the Amadeus Central Reservations System (ACRS). Built API-first, the system lets Ascott define and distribute inventory by richer property attributes in addition to room categories. According to the press release, it will activate properties and promotions faster, propagate rate logic more consistently, and match guests, and the AI agents acting on their behalf, to stays based on what actually matters to them. Paul Wilson, who leads Amadeus's hospitality business in Asia Pacific, framed it the same way: "Attribute-based shopping is where distribution is heading."

Nor is this an Ascott-only experiment. Marriott agreed to deploy ACRS in 2021, and Accor has announced its own migration. Attribute-based inventory is spreading as a standard for lodging distribution, and Ascott has layered a clear purpose onto that shift: becoming a supplier that agents can transact with.

Pulled into retail and e-commerce terms, this is precisely the problem of product data readiness. Beyond names and prices, can you hold material, use case, compatibility, and delivery conditions as structured decision inputs? The sharpest lesson of this case is that attribute-based product data is a design question for the systems that define inventory, not for the product page that displays it. Improving the presentation layer alone does not produce a shelf that agents can read.

Cubby's agent turn is a concept; the protocols are pilots

Accenture carries the architecture layer. Together the companies are designing a foundation layer that lets information flow across Ascott's core systems for reservations, property management, CRM, and loyalty. Since 2023 Ascott has operated Cubby, its in-house digital concierge, which has supported more than 900,000 guest enquiries, autonomously handling most routine interactions and contributing to booking outcomes. On that track record, Accenture's vision is for Cubby to evolve into a personal travel agent that compares options, plans itineraries, and completes bookings on a guest's behalf.

Here the facts and the ambitions need to be kept apart. The press release's wording is conditional, saying Cubby "could evolve," and no autonomously booking agent has launched. The same applies to connectivity standards. Frameworks based on MCP (Model Context Protocol), the specification for connecting AI to external systems, along with LLM-enabled applications and early-stage unified commerce concepts, are all explicitly described as capabilities Ascott "will be pilot testing." Not conflating shipped features with technology under evaluation is essential when reading announcements of this kind. For the broader landscape, see our overview of the major protocols.

Concept stage or not, the design principle is worth studying. Ascott plans to deploy a standardised agentic layer across multiple guest interfaces: LLM chats, messaging apps, and its own direct booking platforms. Rather than standing up a separate bot per channel, it places one agent layer on the core-systems side, so every entrance connects to the same inventory, pricing, and membership logic. Emily Weiss of Accenture called agentic commerce the biggest shift in commerce in the last 20 years and argued that "brands who have systems that agents can trust enough to transact with have the advantage," adding that a technology overlay is not sufficient and platforms need to be re-architected while brands reimagine how they present themselves to machines.

Loyalty data and visibility inside algorithms

A comment from Tan Bee Leng, Ascott's Chief Commercial Officer, captures the character of the announcement. In an agent-led travel ecosystem, she said, properties must be visible where the real decisions are made, inside algorithms, and brand and property information must become machine-readable and optimised for generative engines. She also set a requirement that members of the Ascott Star Rewards loyalty programme be recognised at every touchpoint, whether they are searching on their own or through an agent.

Loyalty parity is an easily overlooked but heavy question. If member rates and benefits apply only on the brand's own site, agent-mediated comparisons may present the operator as more expensive than it effectively is. How member identity gets passed to an agent, and which party guarantees benefit and points calculations, requires design on both the technical and operational sides. Ascott is also strengthening its content ecosystem, cultivating reviews and digital advocacy to improve how its properties are found by AI-powered search and generative engines, and it counts "AI visibility" among its tracked outcomes alongside booking values, efficiency, and time-to-market.

The backdrop is an industry-wide shift in the environment. Marriott and Hilton have disclosed in their annual filings the risk that AI platforms divert bookings away from direct channels. Apps in ChatGPT, announced by OpenAI in October 2025, launched with Expedia and Booking.com as initial partners; Hilton has offered an AI planner on its own site since March 2026; and Marriott began the beta rollout of its conversational search Ask Bonvoy in June 2026.

Most of these moves sort into the choice between riding external AI channels and sharpening an owned conversational touchpoint, the open garden versus walled garden question. What sets Ascott's announcement apart is that it steps squarely into the redesign of inventory, core systems, and membership that sits upstream of that choice. Whichever strategy an operator picks, the machine-readable product foundation is required either way.

What retail and e-commerce operators should take away

First, treat product data attribute-isation as a core-systems investment, not a presentation-layer task. To widen the very definition of inventory from room types to attributes, Ascott accepted the heavy choice of replacing its reservation system. In e-commerce terms, that maps to reworking the product master, SKU granularity, and the update paths for stock and pricing until AI can interpret them. It is slower and less visible than a front-end refresh, but it is the precondition for being chosen by agents.

Second, redesign member and loyalty data so it travels across channels. Ascott's requirement to recognise members at every touchpoint gains weight as agent-mediated purchasing grows. Keep member benefits locked to owned channels and you lose agent-driven comparisons; open them unconditionally and the direct channel loses its point. Deciding which benefits can be carried outward, and how far, is a business-strategy call rather than a technical one.

Third, sequence concepts and implementations by their level of certainty. Read closely, the announcement mixes three confidence levels: ACRS adoption is the committed foundation work, Cubby's agent evolution is a concept, and MCP and unified commerce are under pilot testing. Because unresolved questions such as payment delegation and identity verification still sit between discovery support and a fully autonomous purchasing agent, building up from the dependable foundation first is an approach worth borrowing at any scale.

Conclusion

Ascott's announcement is a case study showing that the real battleground of agent readiness is not the conversational UI but the definition of inventory, the core systems, and the membership data. Rebuild inventory around attributes to create a machine-readable shelf, plan a standardised agent layer on top, and validate connectivity standards through pilots. Including the discipline of sequencing work by its level of certainty, it stands as a reference for building a product foundation for the AI agent era. Travel and physical retail differ in subject matter, but any business holding products, inventory, and member data faces the same questions. The operators who prepare a foundation agents can trust enough to transact with will be the ones who secure the next customer touchpoint.