‍

The past two years have a proliferation of claims by various products in the hospitality tech space that they are using Artificial Intelligence (AI). But the tagline “Now with AI” is often just marketing-speak for “we use a computer program to decide what to do, exactly the way a human would do it.” This might technically meet some definitions of AI, but it is not what most people expect from AI today. For that, the machine should do something that even a human with unlimited time could not practically achieve, or it should do something that seems to require cognitive capabilities, not just raw computing power.

AI is hot, so everyone claims to be using it. And some truly are, but many are not, and skepticism is warranted.

While many applications of AI have been around for decades, there has been a veritable explosion of so-called AI products in the two and a half years since OpenAI launched the first large language model (LLMs). And to be sure, this has opened up the field of potential AI applications in ways we could hardly imagine previously. Most products using LLMs still have a long road of future improvements ahead of them. Nevertheless, there are AI applications (LLM-based and others) that make sense for hotels to consider using today, and I want to explore some of them.

This will be a tour of hotel departments, calling out some of the AI capabilities that I think are “ready for prime time” for each one (at least for certain applications and some types of hotels). “Ready” is not a black-or-white judgment, however, so I will qualify my assessments by noting the use cases where I might turn AI loose on its own, ones where I would use it but with guardrails, and ones where it’s just not ready. In some cases, I have seen the technology in other industries and believe it can be easily adapted to hotels, even if I have not seen it yet.

In most cases my perspective is based on capabilities I have seen in product demos; in a few, based on feedback from hotels who are using them; and in others, based on applications I have seen in other industries. Some are based on the sheer conjecture that if no one is doing it yet, it is an obvious opportunity that someone will do soon. The landscape is evolving rapidly, and I know there will be products hotels are using that I have missed. I would love to hear from you if you know of any. Send me an email or post your thoughts on LinkedIn (use the link at the end of the article to find my profile; I will post a link to this article soon after it has been published).

Let us begin the tour.

Marketing and Ecommerce

Hotels publish a lot of descriptive material on their websites, in mobile apps, in brochures, and to partners. It is time-consuming to write and there is an almost unending list of things to be written or rewritten. Most hotels have no staff member trained in marketing copywriting and rarely anyone who can do it efficiently. LLMs, however, do an excellent job of quickly drafting descriptions for a hotel, room types, restaurants, hotel facilities, rates, and packages. What you can get will be well written, well targeted (if you tell it what you want), grammatical, and you can do it in five languages almost as fast as one. It will not be perfect, but a quick human review can address any issues. This is one of the most common early uses of AI, and if you are not using it, you should ask yourself why. Even the free versions of LLMs can do this very well.

AI can also be used to optimize the order of product offerings on a booking website. This practice is not widespread in hospitality, but it is extensively deployed in retail ecommerce, and relatively simple and powerful. AI can screen through millions of historical interactions, classify customers by various factors (historical behavior, source market, size of party, length of stay, day of week, channel, type of device or browser, etc.) and suggest the best selection and ordering of products to offer in each micromarket, based on what similar customers bought when presented with certain options in a certain order. Typically, a first pass for each segment is developed based on historical data, and then further refined by live A/B testing. With enough history, you can personalize the offer down to a single customer (think Amazon).

Leading ecommerce sites do this very well, but I have seen only one hotel implementation, and none with OTAs. And I am so tired of scrolling through dozens of different room, rate, and policy combinations, usually presented in a completely meaningless order, to find the best one for me. A few checkboxes on the website (like refundable or not, prepayment or not, accessibility requirements, bed type preference) can be used to further refine the selection and ordering.

For any given set of booker characteristics, you can optimize the number of options to put on the first page of the response (usually in the range of 5-10), and the order. Done well, almost all visitors will find what they want on that first page, and the ordering can subtly influence them to book a higher rate. The few who want something different can use a “more options” button. This type of approach has dramatic impacts on both conversion and average spend, both in other industries and in the few hotels where it has been applied. It baffles me that for all the money the big hotel brands and OTAs spend on booking engines, no one yet seems to have gotten this right – the most I have seen is the ability to do a bit of filtering.

AI can also be used to monitor, analyze, and respond to guest reviews on platforms like Google, Tripadvisor, and Booking.com. The plain-language analytics can scan through thousands of reviews and analyze the content for sentiment, specific likes/dislikes, complaints, and compliments. I have seen at least one product that can even prepare a monthly plain-language management summary with suggested actions for improvement.

AI can also auto-generate responses to reviews, posting them automatically for the simple ones (if the hotel so desires) or drafting responses for human review and approval before posting. This enables staff to respond to reviews much more quickly. It ensures that they maximize the customer’s perception of caring and responsiveness, identify bad experiences faster to facilitate service recovery, and improve the response-time metrics that many booking platforms use to manage the display positioning of the hotel in searches.

Reservations

One of the earliest applications of AI was website chatbots. And while the earlier ones were just text messaging platforms that enabled human-generated responses, many can now handle common questions autonomously, and escalate to human agents the ones they cannot. I have seen these used quite effectively to handle routine, simple queries, and a few can now even handle basic reservation requests via chat or voice response.

AI chatbots are not perfect, however, so you will probably want to give customers a simple and painless “offramp” to speak to a human at any time, and you need guardrails to pick out requests that might not be handled correctly and route them to a human. To be useful, an AI chatbot does not need to be as good as your best human agent, but it should be at least as good as an average one. Be sure to review message logs regularly to identify issues where it needs refinement.

Similar technology is now being applied to email reservation requests, which are common in some markets. AI agents can now parse emails, identify ambiguities, propose rates, confirm bookings, and process payments securely. Most hotels still have humans review the AI responses before sending them, but often find that the majority are just fine, and that the AI saves a lot of time. Some have reportedly let the chatbots respond autonomously in at least simpler use cases.

Contact center technologies (see earlier blog) are perhaps further along than any other in AI maturity. Many systems can now monitor, transcribe, analyze, and log every call, whether it is handled by AI, a human, or a combination. Transcription provides material to help evaluate and coach agents, and analysis can both help measure performance and even alert supervisors to a call that needs their intervention – even without the agent realizing it. Based on what the caller or agent says, many systems can display suggested language for the agent, and AI can score agents against checklists and even remind them if they skipped something on the list. This is leading to rapid evolution in the way contact centers hire, train, supervise, coach, and evaluate agents.

Additionally, some AI applications are now being deployed to mine information databases to aid contact center agents. For example, if a guest speaking with a central reservations office asks for a hotel’s restaurant hours, a list of that specific hotel’s restaurants and hours can be popped up on the agent’s screen without their even asking for it, much less having to hunt for it.

Revenue Management

Revenue management systems have been using AI for years to uncover actionable patterns in data that might be too subtle or time-consuming for an analyst to find manually. Based on what I am seeing with LLMs, we will soon have tools to allow revenue managers or even hotel executives with no revenue management training to ask plain-language questions to get guidance based on analytics that were never specifically programmed. For example, “Taylor Swift just scheduled a concert in our city for November 10, 2025. What revenue management actions should we take?”

Front Desk and Concierge

Room assignments can be time-consuming, especially in a large hotel or one with many different room types. Returning VIP guests or loyalty program elites may have specific preferences, and General Managers and Rooms Division Managers often handle special requests. Arrival and departure times may be known or, if not, then estimated, and room turn times can be affected by housekeeping staffing levels. This is an optimization problem where human input may be required (we cannot ignore the General Manager’s desire to give their friends a special room!) but too complex to optimize in most existing property management software. It is a great problem for AI, and I have seen several pieces of the puzzle addressed by existing products. It is only a matter of time before someone pulls them all together.

A second area where I have seen capable AI tools, but not yet baked into front desk systems, is upselling. Higher-end hotels and resorts want the front desk, concierge, and activities desks to sell room upgrades, late checkout, spa treatments, cabanas, food and beverage packages, and the like. There is a lot of science on how to decide what to offer, to whom, when, and in what order, however. AI has already been used to optimize this in the pre-arrival upsell phase (at or after booking), but again it has not been built into the property management system to use at the front desk.

AI can analyze the characteristics of the guest (size of party, ages and relationship of the guests, length of stay, day of week, etc.) and recommend the specific items to offer and the best order for presenting them, and generate a popup window to guide the hotel colleague. In similar environments, such tools have dramatically impacted upsell revenue. They enable even minimally trained front desk agents to present meaningful upsell ideas to guests.

At the concierge or activities desk, AI can generate personalized information on arranged activities, such as directions, transportation options, logistics, dress code, payment methods accepted, and the like. Much of the information can be prebuilt by the hotel using LLM-generated text for each activity, with human review. It can then be further personalized based on information to the specific guest or party, which may be extracted from the hotel systems or manually prompted by the concierge. For example, the directions or transportation options to a local restaurant might be impacted by a particular party’s accessibility requirements or by knowledge as to whether they have their own car, hotel arranged transportation, are willing to walk, or some other option.

Concierges are a font of knowledge, but even they do not know everything, and commercial LLMs can be quite good at finding specific options for guests with unusual requests. It may be a matter of the LLM doing research for the concierge, who then uses traditional means to further qualify the options, or the LLM could be used to simply provide the guest with a list of ideas. Hotels without concierges can use today’s LLM-powered chatbots to do this as well, particularly if the chatbot provides a way for the hotel to influence the results or to establish guardrails.

Guest Communications

The best LLM-powered chatbots on the market today are not perfect, but they are quite good and getting better at answering many of the questions a front desk colleague, PBX operator, or concierge might get: the location or operating hours of hotel facilities; how to log into the Wi-Fi; nearby restaurants or attractions; popular walking or jogging routes; and the like.

Should your hotel use one? I have found two litmus tests that you should apply to help answer this. The first is whether the chatbot (with whatever location-specific training it requires) can answer questions as well as your average front desk agent, or at least know when it cannot and escalate those questions to a human.

The second is whether any existing text messaging (non-AI) platform your hotel is already using is being managed (and measured) against appropriate response time standards. In my experience across a wide range of three-, four-, and even many five-star hotels, text messaging platforms that require staff responses (even approval of canned responses) are not operationally manageable to guest expectations. Guests generally expect responses to text messages in a matter of seconds or a few minutes, while actual response times are often many hours (my own average in the past year is over two hours). Your chatbot will not impress a guest who asks for restaurant ideas for dinner at 5pm and only gets a response after midnight.

Hotels should not simply assume their staff will manage this better than an AI-powered chatbot, but should gather the metrics. If you have a text messaging platform, review a week’s or a month’s worth of questions and determine how many were answered correctly vs. incorrectly or not at all; ask your front desk agents the same questions and see how they compare. And get the report from your platform on the average response time. These will give you the metrics you want an AI chatbot to beat. In my experience, for most hotels below the five-star level (and even for many of them), the bar will be far lower than you think.

If you like, you can supplement an AI chatbot with a voice response system in the guest room or mobile app. This is easy to do with today’s tools and many products offer the option. However, given the dictation-to-text capabilities that already exist on most mobile phones, it is questionable how much value this adds in a mobile app. Few hotels equip guest rooms with voice devices such as Alexa these days, but if yours does, this could be a useful addition as it would greatly expand the questions it could answer, without having to manually program each response.

Food and Beverage

AI-powered systems have emerged to both optimize menu planning and reduce food waste. These are interrelated: the right menu leads to less food left on plates, and AI analysis of food waste can lead to more cost-effective and less wasteful menu design.

On the menu planning side, AI can detect preference patterns that are hard for humans to see, such as times of day, days of the week, or characteristics of guests that correlate with specific dishes ordered. The resulting analytics can then be applied to actual bookings for a future period to more accurately predict what diners will order, leading to less food waste and fewer situations where desired menu items are sold out.

A few AI-powered solutions have come on the market to measure food waste, both in the kitchen and on diners’ plates. The better solutions capture and analyze video images and/or weights of food as it is discarded into garbage bins, identifying both the specific dish and amount of waste. The results show insights such as the vegetable or starch that most diners do not eat (and that can therefore be eliminated or replaced with something more desirable), and the number of portions of each dish prepared by the kitchen but thrown out because it did not sell.

Accounting and Finance

The biggest use I have seen of AI in hotel back-office operations is in the procurement and payables processes. The ability to scan, interpret, and encode paper documents is a simple but powerful labor-saver. There are multiple products on the market today that can scan packing slips, delivery receipts, and invoices, and make the necessary adjustments to inventory, flag discrepancies in deliveries, route for approval, or encode invoices into the right general ledger accounts.

Human Resources

LLMs can generate (or improve) job descriptions, ensuring that they include the right information and keywords to maximize exposure in online marketplaces. And many commercial products offer capabilities to scan resumes and score applicants, and to generate personalized responses to candidates with minimal human effort. I have also seen tools that use AI-based analytics to discover the key attributes for success from databases of existing staff and use them to help identify the best candidates for open positions.

Group and Corporate Sales

Some early products are using LLMs to generate responses to RFPs for group and corporate business. They can auto-respond to requests that a hotel cannot accommodate at all, eliminating the need for a human to do this while still maintaining a responsive relationship with the client and (where appropriate) referring them to sister properties. They can generate boilerplate responses with generic rates and availability and request key information or suggest alternate dates or properties. For promising inquiries, they can draft responses for human review. All of these can greatly reduce the time commitment from sales professionals, help them focus on the lower end of the funnel, and speed up the responsiveness to clients and, ultimately, conversion.

Engineering

Common in other industries but not yet widely used in hotels is the ability of AI to detect patterns in device performance that may suggest poor performance or imminent failure. More and more building components (water control devices, boilers, fan control units, freezers, ventilation hoods, and the like) report status continuously to centralized databases or dashboards. AI can analyze historical data to identify anomalies and prioritize maintenance tasks. It is far better for the hotel to discover and correct an imminent problem before something breaks, than to risk guest disservice from malfunctioning HVAC systems or from having to shut down the water for several hours to make an emergency repair on a peak-occupancy day.

Conclusion

The above examples are just scratching the surface of what AI can do, but all of them represent problems that AI is capable of solving today, and many of them are offered by existing products on the market.

What else have you seen, used, built, or wanted that represents an effective use of AI in hotels? I would love to hear about it and will plan an update in a future article. Please reach out to me by email (below) or comment on my LinkedIn page, where I will post a link to this article once it has been published.

‍