The owner of a small hotel or guest house often asks this question when deciding to launch an automated assistant for enquiries: should we tell the guest, or is it better to sound like us? The temptation to sound fully natural is understandable. The guest writes in the evening, wants a quick answer and does not care who is writing. But there is a big difference between “it does not matter to them” and “they will not mind if they find out”. Here is how to think about the choice.
Why a hidden bot costs more than it seems
Imagine a guest who asks whether there is parking and gets a warm, accurate answer signed “Maria from reception”, then finds out the next day that Maria is a program. The parking answer was correct. But the guest now asks something else: what else have I not been told? The distrust does not stay with the parking answer, it spreads to the whole property.
There is also a more practical reason. An automated answer sometimes gets things wrong or does not know. When the guest knows they are talking to an assistant, they read the mistake as a limitation of the tool. When they think they are talking to a person, the same mistake looks like carelessness on the part of the team. Transparency protects you, not only the guest.
What the EU regulation says
The European Commission explains that from 2 August 2026, under Article 50 of the AI Act, systems that interact directly with people must be designed so that the person understands they are talking to an AI, and already at the first interaction. There is an exception where this is obvious to a reasonably well-informed and attentive person, but the Commission says the exception is read narrowly. According to the same page, this obligation lies with the provider of the system, not with the venue that uses it. Do not take this as legal advice: check what applies to your property, and ideally ask the provider of the tool you use how the labelling is handled. Good practice, however, does not wait for the law; trust itself demands it.
How to say it in one line
You do not need long explanations and warnings. A short line at the start of the conversation is enough, for example: “Hello, I am the guest house's automated assistant. I answer from the information the team has prepared. If you need a person, I will tell you how to get in touch.”
This line does three things: it says what the assistant is, says where its knowledge comes from and promises a way out. It does not justify itself and does not apologise. Guests read such a line as a service, not as a flaw.
Tone: warm, but not pretending
Transparency does not mean dry language. The assistant can greet guests in a friendly way, address them politely and be brief. It should not:
- sign with a human name and role that do not exist;
- speak in the first person about personal experiences, such as “I tried the breakfast myself yesterday”;
- imitate a delay to make it look as if “a person is typing”;
- promise things that only the team decides, such as a discount, a room or an extended stay.
The simple rule: anything a person would say in their own name that is not part of the approved information, the assistant does not say.
The way out to a person is part of transparency
A label without a way out is a half-truth. If the guest reads that they are talking to a bot and then cannot see how to reach a person, they will feel locked in. So decide three things in advance:
- what the assistant writes when it has no data: an honest “I do not have this information” and the venue's phone number;
- which topics always go to a person: a complaint, a change or cancellation of a reservation, special conditions, anything that involves money or an exception;
- who takes over the enquiry and when, so the guest does not wait until the next day without an explanation.
If your small guest house has no staff member at night, simply say so: “The team is available from eight in the morning” instead of silence.
A small test for your text
Read the opening line and answer four questions. Will a guest who is not technical understand that they are talking to a bot? Does it say what the answer is based on? Is there a visible way to reach a person? Does the line sound like something you would say in person? If the answer to any question is “no”, fix the line; do not add more text.
Where ISKA Assistant fits
ISKA Assistant answers from the information the venue has approved, and when there is no data, it directs the guest to the venue's phone. There is no live chat with a person, so the way out to the team is precisely the phone and the room request, which the team confirms. How exactly the assistant will be labelled at your property and what the opening text will be is a question worth asking at the demo: you request it from the product page.
What to do this week
- Write an opening line of two sentences and read it aloud to a colleague.
- Choose the phone number you direct the guest to, and check that it is answered during the stated hours.
- Collect five questions the assistant should not answer on its own, and assign a person to each.
- Review the conversations from recent days: did any guest ask “am I writing to a person?” Such a question is a sign that the line is not clear enough.