Don’t Call Her a Chatbot – the Story of Amelia

As Artificial Intelligence (AI) continues to evolve, businesses are finding it can complete some very complicated tasks (like foretelling the future!). For the hospitality industry, one of the major ways AI is typically considered to be relevant is for guest interaction, such as answering FAQs. However, the technology has now come to the point where it can trick guests into thinking they were interacting with human staff members instead of AI-based software programs. And like a human, AI now has the ability to actively learn and continuously improve.

For hotels, this means that an AI-based solution can now become the primary point of contact with guests via any channel: phone, chat or social media. IPsoft’s Amelia is one example of this type of technology. To learn more about how she functions and how her technology could benefit the hospitality industry, we spoke with Jonathan Crane, Chief Commercial Officer of IPsoft.

(Don't forget to check out other stories from HT on the topic of Artificial Intelligence.)

What are some of Amelia's more unique capabilities?

Amelia is the result of decades of research including discrete mathematics, cognitive neuroscience, autonomic systems and embedded language models. We have developed her to be capable of understanding complex natural language, anticipating context switching and recognizing intents humans express in various ways, as well as users’ emotional state. By using unique programming and algorithms, we have modelled her on human intelligence, understanding and empathy, so she’s utilizing each of the human brain’s eight core cognitive skills to deliver unparalleled conversational capabilities. As a result, she surpasses mere chatbots – which only provide pre-defined answers based on certain keywords, as well as RPA robots, which can only follow linear decision trees – in every way.

Where did "her" name come from?

Amelia has been named after U.S. aviator and pioneer of the sky, Amelia Earhart (1897-1937). Amelia is a pioneer herself in the field of cognitive computing, a type of computing which learns by experience and/or instruction, just like a human. With Amelia, organizations are able to explore the possibilities of conversational AI and discover new ways of providing optimum services to clients and employees.

What type of hotels does she service best?

With the wide variety of use cases Amelia can be applied to, she’s an asset for any hotel. For example, at a budget-oriented venue that’s aimed at maximum self-service in order to keep costs down for guests, she can be trained to handle the check-in and check-out process all by herself, including answering guests’ questions for sight-seeing or restaurant recommendations.

At a hotel that is aiming to provide first-class personalized services to a rather demanding clientele, you would probably employ Amelia as an enhancement and support to your expert staff. She could take on mundane services in the background, including room service orders, booking a taxi, and so on.    

What types of limitations does she have?

Limitations will depend on the breadth and depths of training and integration with other backend systems. If a guest asks for the shortest way by foot to the next park, and she’s not integrated with any kind of maps service/database, she will not be able to answer such a request.

But for example, if she’s integrated with your well-kept CRM and your booking system, she can proactively offer services that meet a guest’s preferences and needs. If she’s trained to take food orders for room service, and provided with additional data regarding ingredients, she could provide a guest registered as a vegan with only the suitable menu options. But of course she can’t do that out of the blue, she needs the data and the training – just as any human does.

What do you hope she'll be able to do in the near future?

The goal of Amelia is to augment the work of humans, allowing them to focus on higher-level, higher-value tasks, all the while providing a great experience to those who interact with her. In the future I would like to see her go one step further and aid humans in completing unfriendly and/or high-risk tasks. Think about the potential of leveraging Amelia to work in a mine, or using her to be one of the first digital colonists on Mars, helping to build up a livable infrastructure. This is very forward-looking, but if you think about where AI has taken us thus far, it's not that hard to imagine this future.

How is she currently being used in the hospitality space?

Ecole Hótelière de Lausanne (EHL), a world-renowned hospitality school in Switzerland, uses Amelia to automate three specific IT and administrative tasks.

The first allows guests to use Wi-F securely without the problematic nature of passwords being sent over SMS. This was especially important during job fairs, where more than 400 companies come to the school looking to recruit new talent. With so many recruiters and students in need of Internet access, the school was previously forced to assign a specific employee to handle the existing Wi-Fi system during these events. With Amelia, this process has been optimized and automated.

Amelia also helps students regain access to their accounts via an SSO on their PC or through a physical kiosk at the front of the service desk office. As a result, the service desk does not have to manually process these requests.

Last but not least, Amelia has taken on the role of a virtual admissions agent. She lives on the school’s website and assists prospective or future students with answering more than 200 questions through her Natural Language Interface. If she can't answer a particular question, she'll connect the student to an appropriate human agent. She'll then maintain a record of that conversation to better assist with additional inquiries down the line.

The school has also integrated Amelia’s development into its educational mission so students can utilize cognitive AI throughout their careers. Students eagerly line up to take part in Amelia's development at various stages within the school.

Does she require a mobile app download to interact with staff/guests?

While it’s possible to integrate Amelia with a client-specific mobile app, it’s not a requirement. She can be made available on a website or via channels like WhatsApp, Messenger and Slack, to name a few. She can also be offered via kiosk installations or integrated with mobile service robots, as our partner Global DWS has done to enhance their disinfection robots for elderly care facilities with Amelia's conversational AI capabilities.

What is one of the more complex tasks she is able to handle?

Amelia can have a multi-term intelligent conversation where she can handle humans who don’t follow scripts but can, in 90% of the cases, help them get what they need. Consider this transaction: ‘I want to make a transfer via ACH of $500 to my checking tomorrow.’ That sounds simple, right? But what happens if, when asked to confirm the transaction, you say, ‘Make it a Zelle on Monday.’

Think about that – what did you actually say? Did you answer the question? Amelia asked you to confirm or deny but you didn’t actually answer. But you did imply that you wanted to change the date and wanted to switch from ACH to Zelle. Now imagine having to write a rule for all the different ways a person might say these things. It gets very complex. Because of that, most chatbots have very low coverage, meaning they’re not really able to help you solve a large percentage of your problems. They’re very FAQ-based. They’re not conversational, where you can go back and forth and have interactions.

With Amelia, we realized that to solve the complexity of human variance, we need to combine probabilistic AI, rules-based finite machines and deterministic AI. The combination of those can start to better mimic all the different facets of human intelligence, allowing her to learn observationally to understand humans and carry on an actual conversation.

Any fun stories of guests interacting with her and thinking she was a real person?

Interestingly, 45% of users for one client got the impression Amelia was human. When they were transferred to a human agent, they asked: ‘Who was that Amelia person?’”