AI chatbots for hotels: An opportunity for better digital experiences

By Iris de Jong
A friendly looking robot looking directly at the camera
Strategy & innovation

Research into the digital customer experience offered by hotel websites showed that 40% of hotels made it difficult to find all information about the hotel, its amenities and policies. This isn’t to say this information wasn’t there, more that it wasn’t easy to find, and it created a blocker for those looking to book accommodation. But this reveals a key opportunity for AI chatbots for hotels.

AI chatbots can solve that source of friction. Instead of having to navigate through FAQ sections, click on every single link on the website, or wait who knows how long to for an email response, the question could be posed to the chatbot, which–as long as the information is accessible to it–will be able to provide an answer on the spot.

Not only can they help provide a better digital customer experience, AI Chatbots can also help increase the efficiency and decrease the cost of the customer service team.

They can:

  • Prevent easy to answer, often repeated questions–like what time the gym is open or whether there’s a fee for parking–from being sent to the customer service team, allowing them to provide more meaningful services in other areas of the business.
  • Qualify more complex requests by gathering required information before handing over to a human agent, who can then enter the conversation better prepared, reducing resolution times and increasing customer satisfaction.
  • Serve as an internal knowledge base for hotel staff, support training activities, and identify upcoming trends in what hotel guests are interested in, and
  • Be integrated with the hotel’s booking system to make it easier for people to start and complete their booking journey, or even amend reservation details, delivering an overall better customer experience.

But for the opportunity of AI chatbots to be realised, hotels need to implement a chatbot that has the required functionality and ‘smarts’ to provide the desired digital experience–both for the end user and those managing the chatbot.

So, Inviqa’s engineering team decided to test-drive three AI Chatbots, Ada, Algomo and Microsoft Copilot Studio, chosen for their different value proposition and promise to address distinct business needs. Our engineers took each bot through its paces in a proof of concept (PoC) to get a feel of their capabilities in real-life scenarios and see how they performed against some of the key criteria that would make them useful to the hotel industry.

These criteria were:

So, let's dive in...

How easy is the AI chatbot to implement?

Not to blow our own horn, but our engineers are pretty tech savvy. However, not everyone needs their level of technical prowess. So, each chatbot was tested to see how easy it was to set up and the amount of technical knowledge and effort required to get it up and running. And–potentially thanks to how today’s AI chatbots function–this process was pretty quick (1 to 2 hours) and easy for each of the three chatbots.

More technical requirements like creating customised actions, responses, flows and some integrations would take more time, but are mostly still able to be performed by non-technical people. 

And in terms of getting a minimal viable product (MVP) up on a hotel’s website to start getting an understanding of how web visitors use the tool, this could be done relatively quickly. This is ideal for testing whether the tool is useful for the target audience and performing as intended without having to dedicate a lot of time and energy to get it up and running. We do, however, recommend allowing sufficient time to test the chatbot for accuracy of answers and basic functionality like handing off to a human to make sure this all works before setting it live.

What are the options for connecting information sources to the bot?

One of the reasons that AI chatbots are relatively quick to set up is thanks to the beauty of modern AI and the ability to ‘teach’ them on a hotel’s own data–often information that’s already on hand on the website or internal knowledge base. But how easy is it to connect these data sources to the AI Chatbot?

Each of the chatbots we tested can be connected to various information sources or have the ability for information to be uploaded directly into the chatbot’s system. The chatbot then ‘learns’ this information and uses it to answer any questions sent its way. This means you don’t have to set up conversational flows or logic for the chatbot to function.

We did find that when connecting or uploading information to the chatbot, consideration should be given to how that content is structured to ensure it gives useful responses–like all information on a specific topic housed in its own section with a clear heading. Plus, this information must be kept up to date if (as an example) policies change or a promotion ends, as you are responsible for the answers your AI chatbot provides.

When each chatbot was tested on how it handles content updates, with Ada it was found that if the content is updated on a website that has been imported to the chatbot, it will need to be reimported for the changes to be pulled through. For Algomo and Microsoft Copilot Studio, you just need the ‘refresh content’ button to be hit for the updated data to be accessible to the system.

What are the integration options?

For the hotel industry, it can be particularly useful for chatbots to be integrated with third-party systems like customer support solutions or booking platforms to deliver a more personalised experience (even greet them by name!) and help guests complete the task they came to your website to do – whether that’s simply finding information, booking accommodation, or amending an existing reservation.

When assessing a chatbot, you want to check what native integrations it supports and its ability to support custom integrations. While not all chatbots will integrate with all systems, don’t discount a chatbot that doesn’t natively connect to all the ones you currently use if it has the functionality to support custom integrations.

The chatbots we tested varied in what integrations they supported natively, but all provided options to expand these integrations with more custom solutions. Both Ada and Microsoft Copilot Studio support custom integrations using APIs, while Algomo is connected to Zapier. This allows the platform to be easily integrated with the over 7,000 apps already available on the platform, as well as through Zapier’s custom API functionality.

Does the chatbot support multiple languages?

With guests potentially coming from every corner of the globe, and for those hotels with a presence in multiple regions, having a chatbot that speaks the language of your guests can be especially helpful.

Each of the chatbots we tested comes with multiple languages supported through their interface, though the scope and ability to offer on-the-fly translations differed from chatbot to chatbot. One of the chatbots we tested, Ada, uses Google Translate to translate the learning data into any enabled language, plus it allows you to upload learning content in languages other than English. This can be used to answer questions that are asked in that language. It’s important to note, however, that if the answer cannot be found in the question’s language, the chatbot won’t look through the rest of the knowledge base to see if the question has been answered in another but rather ask the user to clarify.

And there are a couple of watchouts with using this multi-lingual capability. The first is that you will probably need to have humans on hand that are proficient in the languages you enable, in the event the chat needs to be escalated. Second, consider whether having multiple languages in your learning content is actually useful or could cause issues down the line.

Does the chatbot support different tones of voice?

As well as being accessible in the user’s language, having the chatbot ‘speak your brand’ is another way to deliver a more cohesive customer experience. After all, if your hotel’s tone of voice is friendly or quirky, you don’t want the chatbot sounding too formal (or vice versa).

Two of the chatbots we tested allowed you to specify a tone of voice – friendly, formal, neutral, etc. But it's good to know that the content you provide as learning data will also go some way to determining the tone used in the chatbot’s answers. So, if it isn’t already, it may pay to review any learning content for tone of voice before linking it to the chatbot.

We did note that all the chatbots also allow for the customisation of rote answers like ‘I’m sorry, I couldn’t find an answer to that question’ and welcome messages to match your preferred tone of voice.

How easy is it to see chat data and analyse conversations?

Another particularly useful function of an AI chatbot is that it can help you fill knowledge gaps – helping you gain insights into why potential guests are leaving your website without completing a booking, and highlighting information gaps where the AI chatbot is coming up short.

Often–as mentioned earlier–people will use a chatbot to get information they can’t (easily) find elsewhere. The chatbot’s chat history or analytics can therefore be used to help uncover questions that go unanswered and become a blocker to people booking. Not everyone will take the extra step of calling or sending an email when it’s often easier to find another hotel.

Using the analytics provided by the chatbot, you can fill that knowledge gap, adding additional questions and answers to your FAQs, or adding details like gym opening hours or parking fees to relevant sections of the website. Plus, you’ll have gained insights into why people may not have been converting, which you’re now able to address.

While Ada was the only chatbot that provided conversation tracking, performance reports and the ability to filter by topic out of the box, Microsoft Copilot Studio's analytics and topic trends allow you to get to these insights, and Algomo’s chat data can be exported to be analysed by a third-party tool.

All three chatbots did provide comprehensive analytics and reporting on volume, conversions, unresolved conversations, customer satisfaction and more.

What happens if a human needs to become involved?

While conversational AI chatbots have become more sophisticated, this doesn’t mean they’re infallible or that a human touch won’t ever be required. With this in mind, we looked into what would trigger a chatbot to suggest connecting an enquirer to a ‘human agent’ and what that handover looks like: does the conversation stay in the chatbot widget, or does the conversation switch to email or another messaging platform for a seamless handover?

All the chatbots we tested provided ways for chatbot-based conversations to be handed over to a human via email for follow up, as well as through customer support platforms like Zendesk and Salesforce and even Slack where a ‘human’ agent can interact directly with the customer and continue the conversation.

Each of the AI Chatbots we tested offers customisable functionality to dictate scenarios where someone will automatically be prompted to connect to a human and ensure they don’t end up in a loop of, say, trying to amend a booking when that’s not something the chatbot can help facilitate.

So, while AI Chatbots in and of themselves won’t be able to solve all booking blockers, they do provide another channel that can help alleviate customer pain points, provide a better customer experience, and increase booking conversions.

A key thing to remember is that the chatbot will only ever be as good as the information that you feed it, and it’s not a set and forget channel. If, for example, a policy changes, this needs to be updated in the ‘training’ data, while chat data and analytics should – just like any other channel’s data – be checked regularly to identify sticking points and how the experience can be improved.

With the amount of AI chatbots available, it’s also important you clearly outline the requirements of the tool to help you narrow down the selection: not every chatbot is created equal or will have the functionality you need. 

If you’re unsure where to start, don’t hesitate to get in touch. We can help you uncover your requirements, support you through your selection journey – and even implement the solution on your site.