AI Insights: Generative AI to manage short-term rentals

We want to help you keep up with how brands across the travel industry are exploring - and using - generative artificial intelligence solutions from companies including OpenAI, Google and more. So we're surveying technology professionals at leading travel brands, and we'll be publishing their answers periodically here.

David Jacoby, president of Hostfully, which helps professional short-term rental property managers maximize distribution and guest experience, is the latest to offer his insights on generative AI and its impact on the travel industry. 

We began working with generative AI in ... December 2022, after the release of ChatGPT. We spent a month trying different prompts and seeing what the large language model (LLM) was good at and where it struggled. What was clear was that this presented a new way for users to interact with computers. Put in a prompt written in different ways (sentence structures, bullet points, disjointed ideas, etc.), and the AI not only understood the request, but gave you an answer. 

That said, when it comes to hospitality and the guest experience, there’s a difference between being 90% and 100% accurate. We felt that building a feature where the guest could ask questions about the property left too much room for interpretation by the AI. If the AI misinterpreted a question and provided the wrong answer, it could lead to trouble, resulting in guests leaving a bad review of the host. AIs and LLMs now respond with such perfect language that a guest could easily assume it was a human-generated answer.

We feel that AI has a role in this side of hospitality. However, until the technology matures a bit more (or we learn to harness it better), it’s best to limit it to areas where it can only improve the guest experience, rather than risk a bad review.

Our current work with generative AI is focused ... on enabling hosts to provide unique moments to guests, tailored to their specific preferences. Our latest project, which released on July 10, is the ability for the guest to get a custom itinerary generated within the guidebook. The guest puts in parameters such as group size, the dates they are in the area and any restrictions they may have, and the guidebook provides a custom itinerary.

One of the reasons guests stay in vacation rentals is to experience an area like a local would. Unfortunately, answering guest questions and providing recommendations takes up a lot of time. Our guidebooks help by allowing hosts to provide such recommendations. However, hosts who operate in areas with many attractions still spend a lot of time answering questions, especially based on specific preferences of the guest, such as “What bike routes do you recommend?” or “Which museums are best for a family?”

While a guidebook should include the special secret spots that only the host knows about, it can't include every recommendation for every guest. This feature solves that pain-point. Even better, with our Viator integration, when appropriate, property managers can earn a commission from tours and activities that are recommended.

The biggest challenges for us related to generative AI are ... striking the right balance between prompt engineering and allowing the AI to “create,” and having the AI be consistent and reliable enough to be in dynamic guest-facing scenarios.

When we first started building the custom itinerary feature for the guidebooks, it became clear that the AI was great at coming up with really creative itineraries. But once in a while we’d get results that were unexpected or simply wrong. To fix this we’d add more prompt parameters, but that led to the answers being more like what you’d find out of a traditional Google search. So finding that right balance between which prompts to include and how to phrase them was a big challenge to overcome. 

The other big challenge for us is making sure that any AI-powered product we put out for our customers is 100% reliable. At the moment, if an LLM isn’t sure about an answer it still goes ahead and provides it. This could be disastrous in a scenario where it answers booking inquiries, or if a guest writes in about a subtle problem with the property. We’re still experimenting with AIs and LLMs and are confident that the technology will get to higher levels of reliability. Of course, a helpful interim solution is for AI to create drafts for a property manager to review and send, instead of having the replies sent automatically.

For the travel industry overall, we see the most potential for generative AI to … significantly reduce operational costs. If chat bots get consistent enough, they’ll be able to interact with the general public and flag edge cases for a human review. This will have huge implications for airlines, hospitality and local tour operators. 

One year from now we expect to be using generative AI for … more guest-facing activities. The pace of change in the AI space is so rapid, we’re confident that any problems with “hallucinations” will be dealt with. At the same time, there’s such a demand for AI in public-facing scenarios that it’s only a matter of time before a provider comes out with a feature that pings a human only in uncertain situations. We’re also really excited about integrating AI and machine learning with analytics. Large vacation rental managers sit on huge data sets that machine learning algorithms could help make sense of.