What AI Can and Can't Do at Hotel Reception
AI can answer a large share of hotel guest questions on its own, any time of day. But it works best with people behind it, not instead of them.
At Sigma Hostel in Zurich, a 70-room hostel with no staffed front desk, guests contact reception through an AI WhatsApp system connected to the PMS. In the first six months (March to August 2026):
3,500 guest chats were handled
40% were resolved by AI alone, with no human involved
the rest went to the remote support team, with a 65-second average first response time
So four in ten conversations never needed a person. The other six did. Here is where that line sits.
What AI handles well at reception
AI is strongest on questions that have one clear answer and come up again and again. In a hotel, that is a big part of the day:
How do I get into the building and find my room?
What is the WiFi password?
What time is check-out?
Can I leave my luggage?
Where can I park, and how do I get there from the station?
Because the AI is connected to the PMS, it can also answer questions about the guest's own booking, not just general ones. And it answers at 3am as fast as at 3pm.
The key condition: the answers must be clear and standardised first. AI repeats what the property has defined. If the processes are messy, the AI will be too.
Where a person is still needed
The 60% that AI doesn't solve are usually the conversations that matter most to the guest:
Something is wrong. The code doesn't work, the room isn't ready, something is broken.
The request is unusual. A late arrival after a cancelled flight, a change to a group booking, a special occasion.
There is money involved. Refunds, extra charges, disputes about a payment.
The guest is upset. A complaint needs someone who listens and can make a decision, not a standard answer.
These moments decide whether a guest leaves a good review or a bad one. That is why a staffless hotel still needs people. They just don't need to sit behind a desk on site.
How AI and the remote team work together
At hostz, AI is the first line, not the only line. Every guest message follows the same path:
The guest writes on WhatsApp. One number, 24/7, for all buildings.
The AI answers if it can. Standard questions are solved in seconds.
If it can't, the conversation goes to the support team. A person picks it up, with the full chat history and booking details in front of them.
The team solves it remotely. Most issues can be fixed from a distance: resending access codes, changing a booking, answering a complaint.
On-site help only when needed. For the few physical issues, the team coordinates with housekeeping or maintenance.
From 20% to 40% in two years
In June 2024, we shared that AI was handling 20% of guest inquiries. At Sigma Hostel in 2026, it resolves 40% on its own.
The technology has improved, but so has the groundwork. Every real guest conversation shows which questions come back, and each one can be turned into a clear, defined answer. The more standardised the property, the more the AI can take on. A hostel with no restaurant and simple processes is close to the ideal case.
We don't expect AI to reach 100%, and we don't think it should. The goal is to take the repetitive work away from people, so they have time for the guests who need them.
Before you add AI to your reception
AI at reception is not a plug-in you switch on. Before you start, check four things:
Are your answers defined? Check-in steps, house rules, parking, luggage: if your team answers them differently, write them down first.
Is the AI connected to your PMS? Without booking data, it can only answer general questions.
Who takes over when the AI can't help? At 3am too. An AI with no human behind it frustrates guests faster than no AI at all.
How will you measure it? Track how many chats AI solves alone and how fast a person replies to the rest.
AI won't replace hospitality. Used well, it gives your team more time for it.
Read the full Sigma Hostel case study.
Comments