AI seems to be the universal answer to all our problems and for the very first time, innovation might destroy jobs without creating new ones. That may well be the case, but as far as hospitality chatbots are concerned, they are not here to take your job, but rather, to extend your reach.
There are so many places where you are losing business because you aren’t able to properly engage your customers, such as on your website, your Facebook account, WhatsApp, WeChat etc. Chatbots give you the ability to scale your presence on the platforms that you currently can’t cover because you’re chronically short-staffed. They filter and manage recurring requests and call you when the human touch is needed.
Chatbots talk and may be equipped with artificial intelligence, but they are no Terminators sent from the future to destroy humanity. Their AI component simply allows them to recognize patterns in human language and match the need expressed by the client to a database of possible answers. We speak of NLU (Natural Language Understanding).
When your customers ask a question with spelling mistakes or in a form that your bot has never encountered before, its ability to answer depends on the quality of the content you have fed it.
- Generic content is available for free on the Internet and will help the bot understand basic expressions like “Hello”, “Thank you” and “How are you?”
- Specific content is what makes your bot capable of being relevant to your business context and challenges. For example, in the hotel industry, it has to understand what a check-in is. The best way to do this is from conversations in verbatim (transcripts of real conversations) that allow the chatbot to understand more and more precisely concrete requests with the actual formulations of customers. A proper training will make your bot capable of differentiating “I’d like to place a booking” from “I booked through booking”.
- API content lets your bot access other systems of your IT environment such as your booking engine so that it can search and show customer relevant information in real time. For example, room availability and pricing etc. The more APIs you’re able to connect to your bot, the smarter it will be.
Zoe, tell me a joke
Through training and conversation design, chatbots can become experts at responding to customers on specific topics. However, they don’t think like humans do and remain unable to respond as soon as the topic of conversation falls out of their area of expertise.
On one hand, a chatbot can answer very complex but frequently asked questions. On the other, it will not be able to crack a simple joke or tell you how it feels — which is OK, as chatbots are assistants for specific applications, not your new life companion. You wouldn’t ask these things either on a human chat. So why the heck would you ask a bot what it is wearing? Zoe answers to over 4000 hotel conversations daily and not once has it received questions out of the hospitality scope.
If you expect a quick result from your chatbot, I encourage you to work with an expert company specialized in your industry, that has already accumulated enough experience and verbatim. (More info on the learning curves of chatbots here.)
Conversation design is as important as intelligence
Conversation design is the ability to frame a conversation so that the customer immediately feels comfortable and understands what your bot can be helpful with. Some elements of design are essential.
- Welcome message: Your bot must be extremely clear about its nature and abilities. Most conversation failures happen right at the beginning. So, a well crafted welcome message can dramatically increase the success rate of your bot conversations.
- Funnels: Check-in time questions can be tackled with simple answers, others such as booking a room, require a more complex dialogue. This kind of interaction needs to be handled with a series of Q&As called funnels that gather all the information your chatbot needs to give a precise answer. Once the bot enters a funnel, any question outside of this funnel won’t be understood until the funnel is finished. So if a customer enters a ‘book a room’ funnel and in the middle decides that he wants to ask about the menu of the restaurant, the chatbot won’t understand. It is critical to properly identify what the customer wants before starting a funnel and to make the dialogue as short as possible.
- Fallbacks: These define what happens in case of conversation failure. Usually, the chatbot will try to put the conversation back on track and if it doesn’t work, transfer to a human being. Thanks to fallbacks, trapping people in an infinite loop no longer happens with decent chatbots. (More info on the different offers of chatbots available on the market here.)
Artificial intelligence is not comparable to human thinking. However, while chatbots are not able to improvise, they are very useful to scale repetitive customer interactions and trigger sales.
If you’d like to discuss more or have questions about chatbots and AI for the hotel industry feel welcome to contact me at firstname.lastname@example.org.
Originally published at www.quicktext.im.
About the Author: Benjamin Devisme
Co-Founder & Evangelist of Quicktext, a 24/7 AI-powered communication toolbox helping hotels increase sales while improving the guest experience.
Need help optimising your hotel revenue? Contact us today and let’s schedule a call to see how Catala Consulting can help you optimize your revenue performance in your hotel, apartment or restaurant.