Skip to main content

How Generative AI Maximizes the Direct Channel with Tata Crocombe

Crocombe offers hotelier experience stemming from the firsthand results of experimenting and deploying GenAI at his private island resort and two beach resorts in the Cook Islands in the South Pacific.
Advertisement - article continues below

The public debut of ChatGPT in late 2022 and the subsequent hype cycle throughout the first half of 2023 certainly spurred a lot of thought from hoteliers about to use these large language models (LLMs) to improve the hotel business on multiple fronts – the onsite experience, BOH efficiencies, recruitment and, for today’s case, getting guests to book direct in lieu of third-party channels.

Perhaps no hotel owner understands just how revolutionary generative artificial intelligence (GenAI) will be more than Tata Crocombe. Rather than just a vendor promoting its products into the vast expanse that is the hospitality technology marketplace, Crocombe offers hotelier experience stemming from the firsthand results of experimenting and deploying GenAI at his private island resort and two beach resorts in the Cook Islands in the South Pacific, respectively the Aitutaki Lagoon Private Island Resort, The Rarotongan Beach Resort & Lagoonarium, and Sanctuary Rarotonga.

We first sat down with Crocombe at HITEC this past June in Toronto, Canada, after he delivered a presentation on the seemingly innumerous ways that GenAI can be iteratively embedded into various operations and technologies. Crucial here is the word ‘iterative’ as contemplating the sheer power of ChatGPT and its ilk can be overwhelming insofar as rethinking every entrenched process at a hotel in one fell swoop.

Rather, Crocombe’s presentation demonstrated the various ways that he has currently implemented GenAI as well as the other areas where he’s charted a course for successive rollouts over the coming quarters and years once the initial use cases have been ironed out. As he emphasized, however, probably the most profound way that a hotel can deploy GenAI right now is by using a machine learning (ML) chatbot or conversational AI (for the voice channel) to enhance the website booking experience as a means of driving direct bookings.

All other variables held constant, quickly scaling your technologies to support the biggest possible channel shift is the best value to drive profitability in the next few business cycles because of the resultant bump in net revenues with marginal opex increases.

The Generative Game Is Afoot

In the pursuit of greater and greater GOPPAR and, ultimately, net operating incomes (NOI), there is no clearer place to look for a quick win than in growing net revenues by reducing customer acquisition costs (CAC). This objective is achieved by increasing the percentage of direct business significantly by winning guests to book online direct with the hotel rather than a third-party distribution partner, and in the process reducing commissions to the OTAs that can be as high as 28% down (comprised of a 15% standard commission, a 3% preferred commission and a 10% members commission) to the direct CAC which can come in as low as 2%.

Every prudent executive committee, marketing director or revenue manager has likely already given some thought to their channel conversion strategy and how chatbots figure into this mix of tactics. But with Booking Holdings and Expedia both rapidly incorporating ChatGPT into their travel planner toolkits to help keep more customers loyal to their platforms, it’s now or never for many hotel brands.

This is an arms race that hotels cannot lose grounds on because, when combined with their own loyalty programs, the OTAs are building defensive moats where it will be hard to reverse the value they are extracting from the global hotel industry once their GenAI tools are ironed out.

Booking Holdings and Expedia have set the current standard for online bookings, though. If hotels want to increase their direct booking percentage, they have to out-service these competitors by making the reservation experience more efficient and pleasurable than that on an OTA. This requires hoteliers to never accept the status quo for how their website or booking engine works, as guests may want in their hearts to book direct but end up going back to an OTA because the hotel’s was poorly built or more expensive than the third party.

This last aspect related to channel management touches on an important point before getting into the specifics of chatbots – concerted efforts. An AI agent is not a panacea; it must be deployed in tandem with an array of other tactics, some online and others not, that will together make for a better reservation experience.

To give you a specific example here, Crocombe developed for his properties a landing page called What Room is Best for Me that has helped guests work out for themselves where they want to stay, adding an element of gamification and ‘sense of discovery’ to the customer journey. This was combined with a thorough upgrade to the website with lots of new information, imagery and virtual room tours, as well as direct booking promotions including 15% off on accommodations then 10% off on restaurant, bar, spa and gift shop purchases.

Next, even before testing and launching the chatbot, Crocombe observed that there are guests that want to deal with a human and those that are happy to parlay with a bot. So, his team connected a 24/7 live chat and recruited a reservations center in the Philippines so that going forward this blended model would facilitate flexibility for any guest’s preferences.

How Direct Chatbots Help You Win

Crocombe offers a shimmering ray of hope, though, in that a well-trained LLM-based AI will always be superior to whatever offering the OTAs present because no third party will ever be able to give the same level of detail, the focused passion, calibration of responses and connectivity into all hotel operations as a first-party system can.

Significantly, this new generation of AI tools is ‘democratized’ in that they are available to any property for sometimes as low as $20 per month, thereby leveling the playing field and allowing any independent hotel or vacation rental a better-than-fighting chance against the multinational brands or, our industry’s common enemy, the OTAs.

That said, the training and interfacing for any chatbot for conversational AI falls squarely on the hotel’s marketing and IT teams to elevate it through testing and retesting to that 95% confidence interval wherein guests are always satisfied with the GenAI’s answers and come to trust the hotel’s machine for all requests. Oftentimes, once fully trained, the accuracy of the focused bot can be greater than almost all of the staff besides a handful of senior managers who have lived and breathed a hotel for years. In Crocombe’s case, the system is split between general information and contractual information relating to a reservation or other exchange of monies, with the chatbot for general information and the structured booking engine for reservations.

While other related tactics to encourage direct bookings may include the strategic sequestration of inventory and information sent to the OTAs so that the richness of content is always greater on the hotel website, generally most customers will go wherever the experience is the most frictionless (speed of service) and the most personalized (quality of answers). Direct integrations into the restaurant or activity bookings are one such elaborate way to increase the level of convenience as is offering more loyalty perks or remembering details from past interactions with the website.

Still, getting an AI from the sandbox testing phase to the 80% confidence interval then to >95% calibration requires a lot of work by the entire team that cannot be understated. However, this can be achieved relatively quickly, as in weeks and months rather than years or decades, with focused testing prior to launch and in the early days of being online with real guests.

With the arms race already underway, Crocombe advised that, in the case of ChatGPT’s applications, hoteliers need to ‘move fast and break things’ because getting to the 95% interval won’t happen by itself, but it can be achieved relatively quickly.

These narrow AI use cases are called ‘generative’ because they take ‘generations’ to perfect; unlike humans where every generation is about 30 years, here each successive iteration can take days or even hours. Moreover, beyond competing against the OTAs, it’s highly likely that one of your competitors is about to (or has already!) launch its own advanced chatbot, allowing it to jump to the top of the comp set over your own brand, particularly in converting lookers to bookers direct rather than through a third-party distributor.

webpage displaying virtual assistant technology using AI
webpage displaying virtual assistant technology using AI

Estimating the Impact on NOI

To demonstrate the longtail of what these Gen AI tools can do, let’s consider a simple hypothetical situation for a leisure-dominant, independent resort with 200 keys, average yearly occupancy of 75%, ADR at $250, OTA commissions (including preferred and members additional commissions) at 28% and direct channel CAC at 2%.

Since reopening from the COVID-19, Crocombe has introduced a myriad of strategies and tactics such as website widgets promoting direct bookings, 24/7 live chat, room packages and new loyalty programs that have grown his direct business from 20% (pre-pandemic) to 45%. But he sees ChatGPT bots as offering the biggest potential to convert lookers to direct bookers, targeting 65% direct bookings with a full roll out of ChatGPT enabled chatbots online. With these numbers, we can then look to see the overall size of the prize in terms of increasing direct bookings from 45% to 65% within a fiscal year.

Applying this 20% delta, we can deduce the net revenues increase as follows:

  • 200 keys for 365 days/year at 75% occupancy = 54,750 room nights sold
  • 55% (100% minus 45%) of 54,750 room nights at $250 ADR = $7,528,125 revenues at start
  • 35% (100% minus 65%) of 54,750 room nights at $250 ADR = $4,790,625 revenues at end
  • OTA room bookings shifted to direct channel = $2,737,500 gross revenues
  • CAC difference (28% OTA commission minus 2% direct channel costs) = $711,750 savings

Yes, there are some assumptions baked in here, particularly as this supposes a one-to-one conversion of OTA to direct reservations with no attrition to total occupancy. Regardless, a half-million-plus boost to net revenues is something that should make any hotelier perk up. This is doubly true when you consider the unit economics of this channel shift where the vast majority of this savings in net revenues flows straight through to NOI.

Benefits Beyond Net Revenues

Using whatever market cap rate expansion or compression percentage you want, the long-term effect is more safety for owners in terms of asset valuation growth. For the Cook Islands, the cap rate hovers around 10%, which means a multiple of 10 for this NOI flowthrough savings that can increase the value of the hotel by approximately $7 million.

Moreover, we must also consider TRevPAR growth whereby it’s reasonably assumed that the guests who book direct versus the OTAs will also have a greater ancillary spend per guest – particularly for resorts – and will be more likely to contribute to the brand via return visits to the same property or others in the portfolio. Less readily quantifiable, but still significant, benefits of encouraging this channel shift include better guest satisfaction that will translate into more word of mouth to help grow occupancies and fewer chargebacks for which evidence points to the majority of these emanating from third-party bookings.

To close, this gain to net revenues from the embrace of GenAI should by itself act as a mic drop for any hotelier looking to optimize NOI in a profound way over the next few business cycles. But as Crocombe bolded and underlined throughout our many conversations, this is only the start.

You can point to every single line item on a P&L or every single process within every department and there is undoubtedly a way that AI can deliver a quantifiable ROI for any hotel. It’s just a matter of starting with one objective and seeing it through to completion then looking for the next application and the next until, before you know it, you have transformed the hotel into an AI-first self-learning and self-optimizing hotel.



Together, Adam and Larry Mogelonsky represent one of the world’s most published writing teams in hospitality, with over a decade’s worth of material online. As the partners of Hotel Mogel Consulting Ltd., a Toronto-based consulting practice, Larry focuses on asset management, sales and operations while Adam specializes in hotel technology and marketing. Their experience encompasses properties around the world, both branded and independent, ranging from luxury and boutique to select service. Their work includes seven books: “In Vino Veritas: A Guide for Hoteliers and Restaurateurs to Sell More Wine” (2022), “More Hotel Mogel” (2020), “The Hotel Mogel” (2018), “The Llama is Inn” (2017), “Hotel Llama” (2015), “Llamas Rule” (2013) and “Are You an Ostrich or a Llama?” (2012). You can reach them at [email protected] to discuss hotel business challenges or to book speaking engagements.


This article may not be reproduced without the expressed permission of the authors.


This ad will auto-close in 10 seconds