USF and WVU study: guests want human contact even with an AI concierge

What the study is and who conducted it

This is not an advertisement for software and not a provider’s report. The University of South Florida (USF) published a news item on 9 February 2026 about an academic article in the Journal of Hospitality and Tourism Technology. The article is by Luana Nanu of USF and by Ajay Aluri and Amelia Szczesney of West Virginia University (WVU). Its title, in translation, is “A study of service and engagement in the hotel industry: from high human contact to conversational AI”.

The method is mixed: group discussions with people from the industry, then surveys of hotel managers and of guests. The university’s news item states a sample of 145 participants: 44 practitioners from the hotel industry and 101 consumers with a recent hotel stay. The surveys were conducted in the south-eastern states of the US. The news item does not state the date when the data was collected.

This matters for the scale of the conclusion. The sample is small, there is only one region, and the results describe attitudes, not real behaviour when making a reservation. The full text of the article is paywalled and was not read for this publication; it relies on the university’s news item and on the public summary of the article. Anyone who wants the exact methods and numbers needs to open the primary source.

What was found

The clearest finding is the difference in enthusiasm. Hoteliers and staff are considerably more positive about an AI concierge than guests are. The reason is understandable: for the team it is a relief when the workload is heavy and there are not enough people.

According to the study, guests prefer a person when the occasion is personal or emotional. The text’s conclusion, not the study’s, is that for quick, routine questions the assistant may be enough. An example from the study: suggesting a restaurant for an anniversary or arranging tickets for a local attraction.

There are two obstacles. The first is the lack of emotional authenticity: the guest does not feel empathy. The second is privacy and trust, especially in voice conversations in a public place. According to the university’s news item, 81% of the participants named emotional authenticity as a critical problem, and 76% had concerns about privacy and trust.

The authors do not say “do not use AI”. They recommend a mixed model: AI in apps, on the website or on a screen, with a person available for complicated matters. Nanu describes it in such a way that AI leads the conversation until it becomes more appropriate for a person to join.

What this means for the small property

In a big hotel, “a person” is a whole department. In a guest house with six rooms, “a person” is the owner, who is at the same time cooking breakfast and welcoming a guest. That is why the lesson is simpler and easier to apply.

Divide the questions into two kinds.

• Quick and repetitive: breakfast time, parking, check-in time, how late the kitchen is open, whether there is Wi-Fi, how to get there. Here the answer has to be instant and identical.
• Important and personal: an anniversary, a diet for a medical reason, a complaint, a change to the whole reservation, anything where the guest wants to hear that they are understood.

The first kind is for the assistant. The second is for a person, and the person has to be easy to find. The guest should not reach “I cannot help” without a next step.

Three practical steps

• Write down what you answer most often over one week. If a question comes up every day, it is a candidate for an automatic answer.
• Decide what is never for a machine. Write the list down for yourself: a complaint, a health problem, a price change, a special occasion.
• Put the phone number in a visible place next to the answers: on the website, in the menu, in the guest handbook. The guest should see the number before they have started looking.

Where the assistant comes in

ISKA Assistant follows a similar principle. It answers from the property’s approved information, and when there is no data, it directs the guest to the property’s phone. There is no live chat with a person, and it does not replace reception. If a guest wants to reserve a room, the assistant collects dates, guests and contact details and records a room request, which a member of your staff confirms, by default by phone. It does not accept payment and does not confirm anything on its own.

This is not proof that guests will like it: the study shows only what the people in the sample prefer. It is a reason to set it up as a helper for what is quick and not to present it as a replacement for a person. Prices are on the pricing page.

Finally

The most useful thing in the study is not the word “AI” but the boundary. Guests do not reject an automatic answer to a simple question. They reject the absence of a person when they need a person. If every question has a quick answer, and for what matters there is a phone, both sides win: the guest does not wait, and you devote time to the conversations that really require it.