- Most AI conversations involve guest requests.
- Bookings made up 45 percent of action requests.
- AI handled up to 30 staff hours of work monthly.
Canary Technologies’ report, “From Answers to Actions: What Millions of Guest Interactions Reveal About AI Inside the Stay,” examined more than 10 million AI-enabled guest conversations.
"Guests want their specific requests taken care of quickly and effectively,” said Gurtej Gill, Canary’s vice president for AI operations and adoption. “Delivering on that takes agentic AI that works across every team and system on property, so hotels can deliver great service, improve operations and drive new revenue."
Guests use AI to make requests
About two-thirds of the conversations involved a request for action, while the rest sought information. This means guests are not just asking AI for answers; they are also using it to get things done. Nearly one in four requests required information from multiple hotel systems to complete.
Check-in and checkout were the most common subjects among guests seeking information, accounting for 34 percent of those questions. Transport and parking made up another 16 percent, while booking information accounted for 12 percent. Guests also asked about rooms, amenities, billing and other hotel services.
Bookings were even more prominent when guests wanted an action rather than an answer. They accounted for 45 percent of action requests, followed by check-in, checkout and arrival-related requests at 17 percent. Another 10 percent involved guests asking to speak with a staff member.
At U.S. hotels, 68 percent of guest AI conversations took place during the stay. Another 24 percent happened before arrival, while 8 percent took place after departure.
Language and guest needs
Waiting for a response can be frustrating when a guest needs help. AI replied to guest requests in an average of nine seconds, and 99 percent of replies came within 30 seconds. Staff took an average of three minutes to respond. One in four staff replies took more than 15 minutes, while one in eight took over an hour.
The technology can also respond in different languages. The conversations covered more than 125 languages, with hotels in Europe, the Middle East and Africa using an average of five languages. The average was four in Asia-Pacific and two in the U.S.
Sixty percent of responses used details such as booking information, past conversations and hotel records to give personalized answers. The other 40 percent provided general information. For example, the system can use a guest’s checkout time, flight details and traffic updates to suggest when to leave for the airport.
There are also signs that guests are willing to engage with the technology. In a random sample of nearly 700 conversations, 27 percent of AI-only exchanges included an unprompted thank-you. The figure rose to 39 percent when both AI and staff were involved.
Revenue and efficiency gains
AI can also help hotels respond to booking enquiries outside regular front-desk hours. Forty-one percent of calls came outside 9 am to 6 pm, and 32 percent of those calls concerned bookings.
The technology can also take routine work off staff members’ hands. At the busiest properties using AI for voice calls and messaging, it handled work estimated to equal up to 30 staff hours a month. The conversations were not always brief, either. Some 23 percent included more than three guest messages and multiple questions or requests.
One example involved a guest booking a room for the same night at a U.S. waterpark resort. The AI checked availability, provided two room options and their prices, confirmed the guest’s details and sent a secure booking link. The guest completed the booking without a staff member joining the call.
Responding to these enquiries promptly gives hotels an opportunity to secure reservations that might otherwise be delayed.
A separate RoomPriceGenie survey found that more than half of hotels use AI to help set room rates, but only 10 percent trust it with most pricing decisions.
