Was 2018 the year of the chatbot? Or at least the year they started to become a lot more commonplace in customer service? They’ve become such a routine service channel, I am now more surprised when I visit a customer service website that doesn’t have a chatbot available.
Over the last few years, advances in chatbot technology have made them significantly easier to deploy and much more effective. They are available anytime and on the customer’s device of choice. They can quickly address common issues, taking the burden off the contact center and allow agents to focus on more complex issues. Though some might argue that chatbots are an impersonal approach to customer service, a recent study claims nearly half (48 percent) of consumers have no preference who or what–human or chatbot–assists them. With so many benefits and growing acceptance by customers, companies have a lot of reasons to love chatbots.
But that’s not all. According to Gartner, organizations have seen up to a 70 percent reduction in live channel (call, chat, and/or email) volume after deploying a chatbot, as well as higher customer satisfaction. That lower utilization of live agent channels, in turn, helps to reduce service costs. These results all fuel Gartner’s prediction that by 2020, 25 percent of customer service operations will be using chatbots.
It’s obvious when adding a chatbot to customer service, a company can anticipate significant returns across several vectors. But to use chatbots only for fast solutions to customer issues at any time of day is to miss out on some additional benefits they offer. They are like having a detective on the team. How is that? Because as customers use them, they are collecting useful data–and not just for that one interaction, but to prepare for future interactions, as well.
Triage the problem
When a company makes the decision to use chatbots in customer service, time is spent determine what problems the chatbots will address. From there, they must be “taught,” employing techniques like architecting the conversations and using machine learning. Through this education, they can then quickly provide answers to the questions they understand.
As a result of how they work, they can’t solve every problem. Some issues are sudden and new. Others might require too complex of a solution. What’s a chatbot to do in these situations? They quickly recognize they’re at their limit and offer to call in help.
The first way your chatbot plays the part of investigator is to collect the details in these cases it can’t solve. Like a human detective scribbling clues on a pad, they can work through the standard set of questions and record the details. If they cannot solve the problem, they can hand all that they’ve learned over to a human agent who can quickly assess the information collected and get to work on solving the customer’s problem.
Identify the need for new solutions
What happens with all those problems the chatbot couldn’t solve that required human assistance? They are not forgotten.
No, once again, they go into the detective’s notes–well, the chatbot’s log, actually. Over time, that log will start to display patterns as to what additional high volume problems customers are experiencing that the chatbot can’t solve. The appropriate cases solved by human agents can be fed to the machine learning algorithms powering the chatbot or new conversations can be built. Either way, the chatbot “learns” new solutions.
As stated previously, chatbots do have their limits. They are typically confined to addressing problems that have simple answers. The better chatbot options, part of modern customer service platforms that also offer powerful workflow, can also deliver solutions using automation.
Understand the language of the customer
Today’s chatbots don’t really understand language as humans do. They are parsing a series of words to determine intent. We can thank the continuing advancements in technologies such as Natural Language Processing (NLP) and Natural Language Understanding (NLU) for this.
Yet NLP and NLU are still maturing technologies. Anytime you’ve queried a personal assistant on your smartphone that has either misunderstood words or provided an answer completely unrelated to the question, you’ve run up against the limitations. As painful as those failures are in the moment, the good news is they serve to fuel future improvements.
Just as with personal assistants, when customers interact with chatbots, its detective nature is at work recording the words and phrases you type into that log mentioned earlier. The chatbot might actually have the solution to the customer’s problem, but when unfamiliar terms are used, it lacks the ability to connect problem to solution–that is, until the next time, after it has been taught these new terms.
Help for today and the future
The appeal of chatbots for addressing common customer issues is clear and their adoption continues to rise. The predictions of their continued growth might even be somewhat conservative. But they have more to offer than a cost-effective means of providing answers to today’s problems.
Chatbots can serve as the intermediary, interviewing customers to collect information for agents (reducing interaction times) for issues they can’t solve. Even more important, they can gather the clues necessary to better understand the customers’ language and what problems are ideal additions to their growing library of solutions. With this kind of sleuthing, even more cases won’t go unsolved for long!