Chatbot Tracker: A Humorous Chatbot?

Be prepared for a chatbot that responds with “Thanks a latte” even if you order a mocha, Americano or just an old-school, plain black coffee.

Starbucks recently announced that it has the cogs churning on a new chatbot called My Starbucks Barista. Circumventing or enhancing (however you see it) the traditional barista interaction experience, the customer speaks into the phone and subsequently interacts with the barista chatbot through a bunch of follow-up questions.

Get ready to place your order for a “double upside-down macchiato half-decaf with room and a splash of cream in a grande cup.”

Or just “a tall Pike Place.”

Whatever your order is, experts say this is soon to be the norm.

“With more than a third of millennials selecting chat as their favorite way to interact with companies, Starbucks is smart to incorporate bot technology into its mobile app,” said Scott Horn, CMO of [24]7, who added that he views 2017 as a big year for bots, especially when it comes to customer experience. “Rather than focusing on making bots seem human-like, companies would be wise to focus their efforts on creating a bot that is as intelligent and efficient as their best-performing human agent.”

Horn said some companies over-invest in giving bots a personality, but a bot shouldn’t pretend to be a person.

Not everyone agrees. Other experts say the human-bot experience should aim to be as fluid and as “human” as possible. The concept of “friction” when it comes to chatbots may arise in the sheer fact that chatbots are, at their core … not human. Thus, the goal to have chatbots become closer to humans, especially in customer service interactions, could truly be ideal and efficient. Even if this poses a hurdle and a challenge that will take time, plus a recipe of blood, sweat and tears.

“Making chatbots human-like is a huge challenge, and something that is far away today,” said Rurik Bradbury, global head of research at LivePerson. “And regardless, 80 percent of people say that they want to be told, clearly, whether it is a human or bot they are talking to. They don’t want to be tricked.”

But being tricked may or may not be an issue. To humanize something is to also infuse emotions and personality elements, such as humor. It’s not knock-knock jokes per se but rather clever responses that may warrant a giggle. But it could indeed be little jokes within the bot conversation.

Weather app Poncho has used a touch of humor in its chatbot. News site Quartz has also made a humor attempt. There is chatter around Sage’s accounting-focused chatbot being “funny.” And, apparently, makeup brand CoverGirl’s chatbot is funny.

“Chatbots should have personality. Even if the bot only offers a simple service or utility, it still engages in real conversations with its users,” said Jonathan Shriftman, director of BD at Snaps. “Communicating with bots should feel like texting with a close friend. Smart bot builders will build their bots with a backstory and program it to respond with humor at the right time.”

Shriftman said a strong, consumer-facing and efficient bot can be achieved with well-written copy or AI. As a result, when users ask questions that welcome funny replies, the bot may be able to incorporate humorous GIFs or images or send witty intelligent messages, such as smart push notifications.

That said, there’s the question of “What is funny?” Something humorous to one person may not jive with another. And that’s where experts debate if chatbots can indeed be funny.

“It’s possible to make a bot’s script fun and lighthearted, but it still won’t pass a Turing test, because humor is the hardest thing to replicate using AI. IBM Watson still doesn’t tell good jokes,” said Bradbury at LivePerson. “Making a bot ‘fun’ also misses the point. The primary point of a bot is not to fool people into thinking it’s human (which we can’t do, and people don’t want) or to be ‘fun’ (which is not why people use them).”

Because the predominant purpose of a bot is to help more people more efficiently, and at lower cost, Bradbury added that typical unassisted chatbots have a CSAT (customer satisfaction score) in the low 70s, much lower than human-powered web chat (in the 80s) or messaging (as high as the low 90s).

“That’s a big satisfaction gap,” said Bradbury. “Closing this gap should be the focus of bot makers, and the way to do it is with higher-quality answers.”

And as a result, any distraction from practical benefits could be a net negative to consumer satisfaction outcomes.

While Bradbury may say “a pig with lipstick is still a pig,” a funny chatbot is not only no easy feat, it’s also “a funny chatbot.”