Pin on Pinterest

Chatbots have come a long way since the natural language processing program ELIZA debuted over 50 years ago. Designed to simulate a psychotherapist, ELIZA was soon followed by PARRY, a program inexplicably modeled after a schizophrenic, and the eventual, inevitable “CyberLover” bot that flirts its way into stealing identities. (Hey, we get it. Sometimes swiping on Bumble gets tedious.)

Somewhere in between falls the customer service chatbot, a simple yet effective piece of AI that delivers convenience for companies and consumers alike. Offering assistance at all times of the day, organizations can allocate resources elsewhere, and customers are spared frustrating wait times and inefficiencies.

What Makes a Good Chatbot?

There are two keys to building a chatbot that works. It must have a dedicated purpose, and it must be transparent. Those may sound fairly obvious, but what companies fail to realize is how much those go hand in hand.

“The reason why people feel like chatbots like Alexa or Google Home and stuff like that are not very personal is because they are built to do everything,” declares Alex Debecker, CMO of Ubisend. “You can ask it about Everest. You can ask it [to play] a song. You can ask it about a job. You can ask about a TV show, anything you want in the whole world, basically, at your fingertips or your lips.”

You may be asking yourself, “Wait a minute. Don’t people love Alexa?” That’s true, but that sort of technology is marketed around its sheer inexhaustible scale of knowledge. A customer service chatbot, though, is designed to replace a specific human interaction with which we are all familiar. Departing too far from that expectation is off-putting. With that said, it’s also essential to be upfront that the customer is in fact dealing with AI. “We try really hard to make it obvious that you’re actually talking to a chatbot,” says Debecker. “The very first message, we always say, ‘Hi, I’m a chatbot, and I’m here to help you on X-Y-Z.’”

EditGenie: A User-Oriented Chatbot

When Nixon Xavier, Vice President of Innovation at Katalyst Technologies, and his team developed a chatbot for EditGenie three years ago, he knew they needed to focus on the end-user. Naturally, they started by analyzing author feedback. “We get all these questions [like], ‘How do I do this? How do I do that?’ so we started [with] that as our base,” says Xavier. “What are the key challenges faced by the customer, and how can we make EditGenie’s chatbot more friendly and helpful to answer their questions?”

They began by covering simple, common concerns like text formatting and capitalization. When the chatbot was unable to resolve the customer’s issue, they realized areas in which the chatbot fell short. As those became apparent, they educated the chatbot further, while still providing the option for human help.

“Sometimes what happens in the user experience is the chatbot is trying to solve the problem without knowing what the problem is,” says Xavier. “We don’t take chances if it goes wrong on the first attempt. We take that and further it to the next route, an escalation server. Our support team can jump in at this stage and work directly with the user.” The innovation team at Katalyst is ready to assist you in creating and implementing chatbots into your own software and services.

Birth of a Chatbot

“When we first build a chatbot, its knowledge is limited,” says Joe Dixon, CTO of Ubisend. “In the early stages of its life, we aid learning by monitoring the conversations it has and manually correcting the cases where it hasn’t responded correctly.” Programmers must build the chatbot around the most common and practical customer concerns—things that can be easily resolved virtually. Once it operates more independently, users can then give feedback that indicates how helpful and accurate the chatbot was, and that educates it further. Subsequently, the chatbot can successfully interact with more customers and solve more problems.

For his part, Debecker doesn’t think there any customer service scenarios that a chatbot won’t improve. “If you call 911 you don’t want a chatbot answering your question,” he says. “When it comes to asking about a parcel, and all that sort of stuff, that that’s a lot less emergency-related.” There is a caveat, however. Not every customer will want to interact with a chatbot. “If your problem is really, really complex, and you really just want to explain it to someone, there should always be an [option] for the user to talk to an actual human.”

As for the rest of us, we hear CyberLover is single and ready to mingle. (Katalyst Technologies Inc is not responsible for any data theft or heartbreak.)

Source Link -

Recognize 354 Views