What really happens when you turn on a chatbot in your hotel?
Do you know what happens when you turn on a chatbot in your hotel? Over the last 12 months, hoteliers have massively embraced chatbots in an effort to drive up direct sales by engaging customers.
Hotel chatbots are quickly becoming the norm as chatbot technology has become truly excellent. However, I feel that too many hoteliers do not leverage the full potential of their chatbot. This is simply because nobody took the time to explain how it worked.
Hiring a chatbot for your hotel is a great idea but there are 3 things you must know before you do. This article will help you understand what happens the day after you turn on a chatbot in your hotel.
Chatbots are stupid and that’s OK
Sometimes hoteliers ask me: If we say very mean things to your chatbot does it become antisocial or something? Sorry to disappoint but the answer is No. Everybody remembers these social media experiments where some bot went rogue. The reason for this was not linked to the artificial intelligence inside them, but to their unsupervised machine learning process. Basically, a lot of jerks teaching it to be hateful. This cannot happen with hotel chatbots because we use supervised machine learning. This means that everything that the bot learns has to be approved by a system administrator or the hotelier.
Chatbots do NOT have the ability to think. There are different kinds of AI and I hate to break the dream but what chatbots use is called NLU (Natural Language Understanding). In short, they are stupid machines that are trained to use semantic analysis and recognize speech patterns that they associate with specific answers.
For Zoe (my chatbot), airport transportation is Dialog 11-01. When a customer asks her how he can get from the airport to the hotel, at no moment does Zoe understand any of this – and that’s perfectly fine. You just need her to be really good at detecting the semantic patterns that trigger dialog 11-01 and delivering the associated answer. Chatbots really are dumb! But they can be very useful when used for specific missions such as answering hotel customers’ requests. Since the same questions keep popping up over and over again, they often deliver the right answer faster than humans.
The number of requests that hotels can get is finite so you can train a chatbot to answer 90% of them. It’s a numbers game. When you reach a certain volume of interactions, you’ve mastered the learning curve. Your chatbot becomes really good at matching questions and answers.
How do you make your hotel chatbot smarter?
When you turn on a chatbot in your hotel, realize that AI is only part of the success of your bot. I recently received an email from a hotelier telling me that Zoe was not answering customer questions properly. We had just moved to a new version 6 of Zoe so glitches were entirely possible. However, I quickly realized that the issue was actually the answers inserted by the hotelier himself. When you set up Zoe you have to load your answers in the chatbot’s brain so that it answers properly.
Let’s do a quick thought experiment. If you enter: I like vanilla ice cream in Dialog 11-01, which is supposed to be about the possible transport from the airport, guess what your Zoe will say when a customer asks about airport transfers?
More seriously, you need to make sure that the answers you insert correspond to what the customers usually want to know. They must exhaust the question. Otherwise, customers will keep asking for additional information and the chatbot will fail. When this happens, remember that this issue is not related to artificial intelligence. Take a deep breath and adjust your answer to Dialog 11-01. Repeat the operation as many times as necessary until you get it right.
Case 1- Airport transfer
After some time you’ll see that glitches are not that frequent anymore and that most conversations go just right. So when you turn on a chatbot in your hotel, remember that this is just the start. There is a learning curve and you are at the bottom of it. For the first 3 months, you should really let your chatbot interact as much as possible with your customer. It’s just like when you’re learning to dance, there will be mistakes. When it happens, take over and remember to adjust the predefined answers so that the same mistake doesn’t happen again. Sorted.
Case 2- Promo code
We have equipped over 1200 hotels with the exact same technology. The difference in performance is 80% correlated with hoteliers not understanding that AI is about matching customer requests to the right dialog and that what’s in that dialog is actually up to them. To be fair, the variation in performance is also 20% correlated to the language of the customer. Some languages are more common than others. So, it is easier to train Zoe to understand English than Czech. We just need to get more hotels in the Czech Republic!
Chatbots are not practical, they are convenient
While I was doing a demo of Zoe, somebody told me that booking a room through my system took longer than going to the booking engine. It is true. If you want to book a room, the quickest way is to click on the book now button. So why do so many customers click on the Live Chat instead?
Many customers don’t seek practicality as much as they seek reassurance. The main idea behind traditional websites and booking engines is for the customer to search for what he needs and bear the weight of the purchase on his shoulders. When you turn on a chatbot in your hotel, it effortlessly provides the right information on demand. Chatbots exist to reduce the mental effort associated with booking a room. They also make sure that hesitant, lazy and mentally exhausted customers book direct. Most of us have way too much going on so we will gladly go for the hotel that takes us by the hand. It is not because there is a booking button on your website that everyone will want to use it. A chatbot is an additional hook to help lazy customers spend their money on your hotel.
People that visit your website are either your competitors doing some benchmarking or potential customers that are traveling to your city. The question that the latter is trying to answer is whether they are going to your hotel or the one in front of you. If for some reason your chatbot didn’t convince them, do not hesitate to call them back if you think you could close the deal. Most customers share their name, email and mobile number with Zoe so use this information! Remember that in the eyes of your customer, the difference between your hotel and the next is usually very tiny.
Chatbots can successfully manage most interactions, but never ever miss an opportunity to get back to a potential customer. They wanted to hear from you in the first place, so go for them. Because the main thing that sets you apart from your competition is how convenient you make your booking process.