Tech insight
AI in Hospitality: where it is working in Asia and where it is just marketing
April 2026
Every hotel owner in Asia Pacific has, by now, sat through a pitch that promised artificial intelligence would transform their property or their asset portfolio. The pitch is often well-produced, the demo is often impressive, and the price tag is almost always substantial. What is rarely clear, however, is what problem the technology solves for this asset – at this stage, under this ownership — and what it will take to realize that promise.
The two things happening at once
This article is a sober look at where AI is doing real work in Asian hotels today, where the claims outpace the results, and what the distinction means for capital decisions.
The question is not whether AI is good in the abstract. The question is whether the choice of technology and application is the right decision for the asset in question – at this stage, under this ownership.
There are two AI stories running in parallel in hospitality today, and they are easy to confuse. The first is the genuine, measurable progress AI has made in a handful of specific applications — most of them operational, and very few guest-facing in the way marketing suggests. The second is the surge of AI-branded products built on large language models (LLMs) that look impressive in a demo but underperform in production.
It is important for owners and operators to be able to tell them apart. The cost of conflating the two goes beyond wasted CAPEX. It results in integration debt, staff disengagement, board disappointment and a technology story that looks worse at the next refinancing or sale.
Where AI is working
Three categories stand out in our advisory work across Asia:
- Revenue management and demand forecasting. This is the most mature application of machine learning in hospitality and has been for a decade. Modern RMS platforms ingest booking curves, pace, competitor rates, and external demand signals, and produce pricing recommendations that consistently outperform manual approaches in mid-to-high complexity properties. The gains are usually incremental rather than transformational; typically in the low-to-mid single digits of year-on-year RevPAR uplift. However, they are real, repeatable, and do compound over time.
- Back-of-house automation. Invoice processing, night audit reconciliation, payroll exception handling, and procurement workflow automation are all now credibly AI-assisted in a way they were not three years ago. The savings are modest per transaction but meaningful at scale — a finance function that used to need 3 full-time equivalents (FTE) for a ~300-key city hotel can now run on 2 with better exception handling. The change is not glamorous, but the labour arithmetic is.
- Operational analytics. Housekeeping routing, staff rostering, energy management, kitchen production forecasting, and predictive maintenance have all seen genuine gains where properties have clean operational data to feed the models.
The commonality across these 3 categories is that they are internally facing, data-rich, and run in domains where “good enough” is acceptable because a human is still in the loop. They are also, in almost every case, embedded in platforms the property already uses — not procured as standalone AI products, but with AI-enhanced capabilities.
Where Results Need More Validation
The guest-facing applications of AI in hospitality are, for the most part, still nascent, and in some cases, cosmetic.
- Concierge and guest-service chatbots. The technology is impressive and sexy. The guest experience in most deployments, however, is not. The handful of properties that have made it work have invested heavily in knowledge base development, escalation process design, and staff training. The backend work is critical for deployment to be successful, and cannot be underestimated.
- AI-generated guest personalisation. The premise — that AI will analyse guest preferences and tailor the stay experience — runs into a structural challenge familiar to many: most hotels do not own the cohesive guest data they need and do not have a unified guest profile across the core systems, namely the Property Management Systems (PMS), Central Reservation Systems (CRS), Customer Relationship Management System (CRM) and Points-Of-Sale. In addition, due to data privacy regulations, properties often cannot legally use what they have across jurisdictions. Personalisation at scale is, at its root, a data problem dressed up as an AI problem.
- Generative AI for marketing and content. This category is more nuanced. The productivity gains for marketing teams producing copy, translations, and image variants are real and worth capturing. However, output quality is inconsistent, brand-safety risks are real, and the teams capable of using the tools well are the ones who did not need them as badly to begin with. The honest assessment is that GenAI is a useful tool for marketing teams, but is not a replacement for marketing strategy.
Why the gap exists
If the working applications of AI are so different from the marketing hype, why does the narrative persist?
From our observations, 3 forces keep this gap alive:
- Commercial
The vendors with the most attention-grabbing AI stories are generally the ones selling new, top-of-mind platforms, while vendors whose AI applications are already working are embedded in existing RMS, PMS, and back-of-house systems. That said, most established technology providers are adding AI capabilities on top of their existing products and tech design. Younger start-ups on the other hand are building AI-first but do not yet have the track record, scale and support systems to establish trust and secure engagements. - Organisational
Boards and investment committees are demanding AI strategies, and operators are expected to have plans. However, saying “we are improving our data foundations” does not impress the same way as “we are deploying a generative AI concierge platform across the portfolio.” The pressure to sound ambitious drives decisions that may not be the right call at the right time. - Structural
The region’s hotel market is dominated by international brands operating under management agreements with local or regional owners. In this ecosystem, operators typically mandate the core technology stack, owners fund it, and accountability over the performance of technology is unclear. AI claims are easy to make and hard to verify when the accountability for outcomes sits in different places.
Personalisation at scale is, at its root, a data problem dressed up as an AI problem.
What this means for capital decisions
For owners, asset managers, and senior leadership teams looking at AI-related technology decisions over the next 12-24 months, 4 key questions need to be considered.
1. What is the underlying problem, and is AI the best tool for it?
The most common pattern we observe in AI-related proposals is a solution looking for a problem. A proposal arrives for an AI-powered guest messaging platform; the underlying problem, on closer examination, is that the property has three different messaging channels that do not integrate. Fixing the integration would solve the problem; while the AI layer adds cost, complexity, and a vendor dependency without addressing the root cause. These nuances and implications typically surface quickly during our Lean Inception workshops with clients.
Consider this diagnostic question: if AI solutions did not exist, what would we do about this problem? If the answer is “nothing, because it is not actually a priority,” the proposal is not worth pursuing. If the answer is “we would fix the data, the integration, or the process,” then that is worth investing in.
2. What does the technology replace, and what does it add?
Real AI deployments replace something, such as a manual and/or repetitive process, an underperforming rules-based system, or a gap in analytical capability. “Nice-to-have” AI adds a layer without removing anything. The operational cost of adding layers compounds over time. This leads to more integrations to maintain, more vendors to manage, more training overhead, and more ways for the overall system to break.
A useful test: if this platform is deployed and it works exactly as promised, what existing system, process, or headcount becomes unnecessary? If the honest answer is “nothing,” the business case is unlikely to hold.
3. Is the data foundation ready?
AI performs only as well as the data it runs on. In the properties we assess, data readiness is almost always the binding constraint. Examples include guest profiles fragmented across systems, booking data locked in vendor platforms, operational data logged inconsistently across shifts and departments. Deploying AI on top of that foundation is a recipe for disaster.
Data consolidation, integration, and governance are the unglamorous investments that need to underpin AI investments. Owners who fund the foundation first reap real returns from the AI layer subsequently. Owners who fund the AI layer first often find themselves funding the foundation eventually, but at higher cost, more pain and under schedule pressure.
4. Who is accountable for the outcome?
In most hotel technology investments, the accountability for the outcome is diffuse by default. The operator recommends, the owner funds, the vendor delivers, the property runs it; but when the results disappoint, no one is clearly on the hook. AI investments amplify this problem because the success metrics are often soft, the baseline is unclear, and the time horizon often extends beyond the tenure of decision makers.
Before committing to an AI-related investment, there needs to be a clear and aligned understanding of the desired outcomes, timelines and accountability.
A pragmatic approach for the next 24 months
Our advice to owners and senior leaders regarding AI in hospitality is neither bullish nor sceptical. Instead, it is pragmatic and it prioritises three actions.
Firstly, invest in your data foundation now. Regardless of AI ambitions, guest data unification, integration hygiene, and operational data quality are worth doing on their own merits. They also happen to be the binding constraint on everything AI-related that will follow. The work is unglamorous, unevenly funded, and almost always underestimated; but it is the investment that makes every subsequent technology investment perform better.
Secondly, capture the mature applications. Modern RMS, back-of-house automation, and operational analytics are available now at reasonable cost, with measurable returns. These are not the applications that make conference keynotes, but they are the applications that show up in your GOP performance a year later. Owners who are not capturing the mature applications are leaving basis points of margin on the table.
And lastly, dissect the problem instead of being lured by available solutions. When we work with clients to break down problem statements, we work through existing processes to better understand the issues before we decide if it is a fix, build or buy decision. Without this critical diagnostic step, it is easy to fall into the trap of Maslow’s Hammer – “If the only tool you have is a hammer, you tend to see every problem as a nail.”
The owner’s question is never really about AI. It is about whether this decision, on this asset, at this stage, under this ownership, makes the asset more valuable or less. AI is a tool and not a strategy.
How we can help
The Horwath HTL Technology Practice advises hotel owners, operators, and investors on the technology decisions that shape asset value. We are independent of vendors, and our recommendations are accountable only to the client’s investment case. To better equip hotel owners, operators and investors, we have developed various workshops to help teams break down problems, align teams and prioritize requirements so that solutions can be right-sized and fit-for-purpose. We have also developed a learning platform to introduce Generative AI to hospitality leaders, so that they can make more informed decisions around technology and AI applications.
If you are evaluating an AI-related investment, reviewing a proposal, negotiating technology provisions in a management agreement, or thinking about the technology narrative buyers will scrutinise at sale, we would be glad to have a conversation.
Ho Shyn Yee, Director and Technology Practice Lead
| Ho Shyn Yee serves as a Director at Horwath HTL Singapore, with more than 20 years of experience in general management, commercial strategy, data analytics, technology and consulting in the travel, hospitality and real estate sectors. |  |