By Russ Banham
Her name is Rose, and she is exactly the kind of person one would expect to handle customer queries at The Cosmopolitan of Las Vegas, a hotel known for its decadent tag line, “Just the right amount of wrong.” Listening to Rose speak, she conjures images that seem to align with The Cosmopolitan’s, well, suggestive ads.
But Rose is not a person—she’s a chatbot that made her debut last year as the hotel’s virtual concierge. And today, her persona matches the attitude of many looking to escape for a weekend in Vegas.
“If you called to ask Rose something and she couldn’t answer, she will say, `Sorry, I just spilled wine on my top and need to change into something more comfortable. Let me get you to a live agent,’” said Clara de Soto, cofounder of Reply.ai, the company that created Rose’s conversational strategy in partnership with The Cosmopolitan’s digital marketing agency R/GA.
Rose is new in the world of AI chatbots, yet she’s not only a smart algorithm. She’s also a personality. As more companies create voice-enabled bots to handle customer queries, they’re realizing the value in having virtual points of contact serve as a marketing extension of their brand.
“The reality is that you’re not just building a bot,” de Soto said, “you’re building a relationship.”
Rapport Is Money
For the Cosmopolitan, this relationship also has bottom-line benefits. Guests who engage with Rose spend 30 percent more at the hotel than those who don’t. And when these guests leave, they’re 33 percent happier, too, according to the hotel’s surveys.
Of course, it isn’t always easy to capture the right personality in a chatbot. “We equate it more to a long-distance run than a sprint,” said de Soto. “The reality is it takes time to do this right. In Rose’s case, the conversational strategy planning took us a solid two years.”
Most voice-enabled chatbots continue to be menu-based, whereby a caller asks a question and is provided with three or four possible answers to pick from. More sophisticated bots like Rose leverage machine learning and, in this case, natural language processing (NLP) to understand and interpret a question before providing an answer.
Certain questions, like booking a reservation at a restaurant or buying tickets for a game or a show, are easier than others. At present, Rose has been able to successfully answer a question 80 percent of the time without turning the person over to a live human being, de Soto said.
When compared to Facebook’s Messenger chatbot that reportedly fulfilled just 30 percent of user requests in 2017, that feels like pretty high functionality. Yet, de Soto continues her work to continually increase Rose’s batting percentage. “There are so many ways of asking a question and Rose is learning all the time,” she added.
Even questions that are salacious in nature, Rose can answer. “We made sure her response was not inappropriate, yet still reflected the brand,” de Soto said.
No Two Bots Should Be Alike
According to Juniper Research, an AI-focused research firm, by replacing human beings with pre-programmed bots, companies in the retail, banking, and healthcare sectors can trim 2.5 billion hours off of the time needed for humans to respond to customer inquiries in the next five years. What’s more, the firm estimates these companies will save an estimated $11 billion in the same timeframe.
Such a dramatic savings in time and overhead is reason enough for companies across industries to invest in chatbots. Another seems to be their reliability. Unlike people, they’re never sick, tired, angry, or rude—and, they don’t make the same mistake twice, thanks to machine learning.
“Machines do things the same way all the time until they learn something new,” said Kevin Kelly, founder and co-president of New York-based digital advertising agency Bigbuzz Marketing Group.
Potential downsides, of course, include difficulties in answering tough questions, causing customer frustration, and a lack of human connection. According to a 2018 survey by consultancy PwC, 64 percent of U.S. consumers felt that brands have lost the human touch in their automated customer experiences.
However, the same study indicates that more human-like interactions could tip the scale. “Automated solutions should learn from human interactions so those experiences can improve,” the study reported.
That’s exactly the kind of work that Reply.ai is doing. “No two bots should be alike,” said de Soto. “Each must be unique because each company is unique.”
While Rose serves as an extension of the Cosmopolitan’s brand strategy, she’s not alone. De Soto aids in the development of other chatbots who do the same (albeit without the sass), working for several insurers, for instance, that need bots who show empathy. For example, upon filing a collision claim, the chatbot will state, “Oh, I’m so sorry to hear that,” before moving on to asking the next question.
Liberal Arts Majors Rejoice
One problem with previous generations of chatbots was that the IT engineers who built the underlying technology also crafted the bot’s responses. De Soto, who has a theater background as a performer, instead enlists a combination of playwrights, copy editors, sociologists, and user experience (UX) designers to humanize the dialogue elements.
In writing the scripts, these creatives collaborate with other creative people at marketing firms to ensure message and tone consistency. “There’s a new title running around called a `conversation designer,’” said Kelly. “Their job is to design chatbot responses to customer queries that sound like real people—assuming people were always smart, helpful, interesting, and fun.”
Bigbuzz is currently working with Reply.ai to develop a full-service chatbot for a large provider of smart home devices. “[De Soto’s] team is presently working on the bot’s tone of voice to give it the right personality,” he said.
And personality may be just what is needed to keep callers on the line. Kelly predicted that in five years, chatbots will become so savvy and ubiquitous, human beings will rarely provide customer service.
“A subset of AI is machine learning, which will continuously guide chatbots to learn from each customer interaction, training them to solve problems in more accurate, faster, and useful ways,” he said. “In ten years, people will be telling their children about the old days when real, live people used to answer their silly questions on their mobile phones.”
Russ Banham is a Los Angeles-based technology writer.