A persona at a coffee bar using thei digital workplace chatbot
PHOTO: Shutterstock

In 2017, only 2 percent of customer service departments were using chatbots. Research from Gartner says that 25 percent of customer service and support operations will integrate virtual customer assistants or chatbot technology by 2020. 

Essentially, chatbots are designed to field customer questions or to help employees find information more quickly. Unfortunately, as with any other tech, the data indicates that people don’t always engage with chatbots as planned.

According to a Pegasystems survey released this month, usage varies by age, country and prior experience with bots. The software company polled 3,500 people in the United States, Canada, the United Kingdom, France, Germany and Australia: 1,000 of the respondents were American. 

No matter where they lived, though, the preference to interact with people instead of bots was overwhelming. However, consumers aren’t against online chat, the majority (65 percent) just wish that conversations were with a human agent.

Related Article: Top 14 Chatbot Building Platforms of 2017

The Difference Between Chatbots and Online Chat

That’s not surprising, because as advanced as chatbots have become, natural language processing still isn’t perfect. This ‘language gap’ can make customer engagement klutzy and frustrating. But 63 percent of Americans who’ve never used one before said they would if the chatbot was connected to a person. Unfortunately, “people bots” don’t exist. Chatbots use machine learning to produce language, but they’re never actual people. When both parties are human, that’s just online chat. Therein lies the public misperception as to what chatbots actually are.

In part, this misperception is perpetuated by companies that have relatively few online chat portals clarifying whether you’re talking to a machine or to a human. “I would never advise a bot to pretend that it is human,” says Ashok Kumar, vice-president of digital at Verizon. Telling customers that they’re talking to a computer creates understanding, he explains: “Machines have limitations. [By] pretending that you are human, you're going to somehow, somewhere disappoint your consumers.”

The survey says he’s right, measuring precisely how disappointment drives away users. Pegasystems spokesperson Russell Dougan wrote in an email that, globally, consumers’ “top complaints about chatbots include not enough smarts to effectively answer questions (27 percent), lack of context in the conversation (24 percent), [and] robot-like engagement with few human qualities (14 percent).” American respondents were even more cutthroat, 53 percent would simple end a conversation when they felt a chatbot wasn’t “smart enough.”

Related Article: How Much Does It Cost To Build an Enterprise Chatbot?

How Consumers Really Use Chatbots

Perhaps this is why Americans use bots more post-purchase than pre-, as 66 percent turn to them to track an order and 41 percent to update mailing addresses. Given that the current trend is developing chatbots to answer product questions, these use cases are far different than what some companies are training for. And because of how natural language processing works, training a chatbot for one use case won’t necessarily make it communicate effectively for another. This, perhaps, is the best insight this new data can provide: How skilled does a chatbot really need to be at what?