AI in hospitality: the truth behind the buzz

Want a demo? Get to know Velma

Articles about AI in hospitality are usually more about sharing a dream. Not so much about giving a realistic view of the current state of AI technology. This is mainly because they are often written by people that are not actually involved in any AI project. Let’s stop watching sci-fi for a bit! Let’s see what value AI is able to create for hotels today and what its limits are.

Rise of the machines

Most of our misconceptions about AI, even specifically AI in hospitality, come from terms such as “artificial intelligence” or “machine learning”. Our intelligence is what makes us human. And we feel either enthusiastic or kind of uncomfortable applying it to machines.

There are 2 types of AI

  • General AI: Creating a virtual human being was the dream of engineers about 50 years ago. The idea was to replicate neuronal connections in computers. They later realized that what made the human brain so special was not just the existence of neurons. It was the fact that there were millions of them working as a network. And this was way too complex to replicate.
  • Narrow AI: In the 1990s, programmers realized that the glass was actually half full. They leveraged algorithms to teach computers how to identify patterns and solve very specific problems. These included –
    • Identifying specific elements in images. This helps Facebook recognize pictures where you appear.
    • Recognizing customer intents in texts and finding the right corresponding predefined answer in a chatbot database.
    • Finding the best itinerary in a defined route for Uberpool riders.

Machine learning 101

Traditional computing is about inputting a predefined command to get a specific result. Machine learning allows the computer to receive any input. It also helps leverage its knowledge base to determine if it corresponds to something the machine can understand because it’s close enough to what it has seen before. If yes, the machine responds with the right answer.

  • If the input is a text such as, “At what time does the check-in start?”
    • The chatbot will take each word into account. It will look into its database to see if it can find other occurrences where these words or synonyms are found together. Then, it will assign a matching probability.
    • If the chatbot has over 90% confidence that the customer intent matches “check-in time” and 40% confidence that the customer intent matches “do you accept checks” then it will display the answer with the highest score.
    • If confidence is below 90% the chatbot will ask to rephrase the question.

With this simplified process, you realize that the level of logic of a chatbot is important, but not enough. What really sets apart AI in hospitality, is its training. In the case of our chatbot, it took us two and a half years to train. We first collected thousands of verbatim (customer messages through a regular live chat called Quicktext). We then manually classified them in intents and fed our algorithm that we named Zoe. It took us a long time but it paid off. Zoe is now able to successfully handle over 77% of customer requests. This process is continuous and under the supervision of a human being. This is called assisted learning.

AI only develops where you can find big, organized data sets. Image recognition is one of the most advanced applications of AI. The reason is the availability of rich image banks such as SALLIE or Facebook that give you images and text description of each image characteristic. The more you feed your AI with this content, the more it is able to recognize these characteristics and identify them in a picture.

  • At this point, you’re probably wondering about the difference between assisted learning and deep learning.
    • With machine learning, you feed your AI with images enriched with a description of its characteristics or text organized in intents.
    • With deep learning, you just feed colossal amounts of data. And you hope that your AI will eventually find out on its own, the difference between a cat and a car. As far as chatbots are concerned, we stick to machine learning. This is because we don’t have enough datasets to expect the chatbot to learn by itself.

How do hotels profit from AI

A new generation of tools are available in the market. These help hotels harness the potential of AI to boost sales and increase customer satisfaction.

  • Big data. Some booking engines such as Avvio have developed Elora. Elora is an algorithm able to identify patterns in customer behavior. It can automatically personalize the display of the website and booking process in real time.
  • Bots help hotels leverage guest interaction for satisfaction and e-commerce purposes.
    • Chatbots are an engagement tool that can be used to help hesitant customers choose you. On property, they handle questions and requests, and also suggest services that will please the guest and trigger consumption.
    • Voice bots are much more action-oriented. They can help the customer perform tasks in his room. Currently, the voice bots for hospitality are still in their early days. We can’t wait to see what will happen with them in the next year or so.

Overall, the opportunity for AI in hospitality is not to replace the workforce of hotels with terminators. Rather, it is to remove existing points of friction so that more customers book your rooms and have a good time.

If you’d like to discuss more about AI in hospitality or have questions about our hotel chatbot, please visit our website or feel welcome to contact Benjamin Devisme at

Checklist: The Best Direct Booking Strategy
Free Download
10 Best Practices to Boost Direct Bookings
Recent Posts
Are you ready to increase your operational efficiency and direct revenue?
Subscribe to our Newsletter
Read more
Optimize your hotel's efficiency and direct bookings
Book a demo