A few years ago, experts predicted a pandemic was coming. They saw the indicators and issues the warnings, and yet preparations weren’t made.
Credible predictions are difficult to make. Even more challenging is getting others to buy into what might seem a radical prognostication.
Several years ago, Gartner made a few bold predictions. In 2011, their analysts’ tea leaves indicated 85 percent of all customer service interactions would be “without a human” by 2020. This was followed seven years later by the prediction that in 2020 over 50 percent of medium to large enterprises would have chatbots deployed.
In the first few months of 2020, the pandemic has added to already high customer service volumes. Add to the mix work-from-home challenges and workforces affected by the virus and things only get worse. Inquiries about negotiating fees, skipping payments, addressing unemployment claims, resolving outstanding travel issues, or other topics, are keeping customer service busy. These common, high volume issues are exactly the types ripe for a chatbot to address quickly, while allowing customer service agents to focus on more complex issues.
On which side of that 50-50 adoption split Gartner predicted does your customer service find itself? If chatbots are in use, congratulations! If not, now is the time to kick off that project. In either case, consider the following tips for bringing them online and further refining them.
Identify problems and goals
Whether already deployed or starting to plan, it’s important to keep the chatbot’s scope well defined. A chatbot is not suitable for every customer problem. The higher volume, more common, and less complex issues that consume an agent’s time are the ideal starting point, such as those examples listed above.
Success metrics for the chatbot should also be defined. Examples include measuring the overall service volume addressed by the chatbot; the success rate for customers; and for problems the chatbot can solve, the breakdown between chatbot and live agents.
Customer satisfaction with the chatbot should also be measured. Some common questions to ask include: was the issue resolved efficiently? Did they enjoy using the chatbot? Would they use it again?
Customers prefer a fast, efficient solution to their problem. That doesn’t mean the experience need be robotic. It’s more likely they will consider using it again if it has some personality.
Simply providing the chatbot with a name is a great start. Check with your marketing team: company branding or style may allow something playful or dictate something more formal.
Also important: tone. The chatbot’s responses can be polite and professional or more relaxed. Again, brand and even the nature of the inquiry should influence how dialog is crafted. Create and maintain an interaction style guide to ensure the chatbot’s conversations stay consistent and on-brand while avoiding anything that might be confusing or off-putting to customers.
Even with the ‘bot more personified, be transparent with customers. Inform them they are not interacting with a human. They may get upset if they feel they are being deceived.
Powered by solutions
While personality is important, remember that customers want answers as quickly as possible. It is often the case customers’ frustrations with chatbots lies in their inability to get to the solution efficiently. Successful chatbots are those that not only focus on resolving a defined set of problems (the first consideration, above) but deliver proven solutions.
Solutions offered by chatbots can tap into other available self-service options. When many steps are involved, having the chatbot refer customers to a knowledge base article makes it easier for customers to perform them. Directing to automated solutions–forms to submit information or to perform other automation–is also ideal. Just be sure to perform regular checks that those solutions remain valid.
Chatbots watch for keywords and phrases used by customers to offer solutions they can provide. They don’t truly know words. Since humans express themselves in a myriad of ways, that can create issues for chatbots.
Modern chatbots can overcome these challenges with Natural Language Understanding (NLU). NLU allows a chatbot to understand customer statements by using models to determine what the customer wants to do by extracting relevant values from their input. This also allows the chatbot to offer a more natural-sounding conversation.
But NLU is not perfect. Not every chatbot conversation will end in success or may take longer than necessary. For these reasons, regularly review chat transcripts of both successful and unsuccessful interactions. What words and phrases did the customer use that the chatbot is unfamiliar with? Apply that information to improve existing conversations and for insight into additional issues to address.
Teamed with humans
Today’s chatbots lack the ability to reason and troubleshoot as a human being can. When scenarios arise the chatbot is unfamiliar with, it should recognize its limitations and offer to transfer the interaction to live customer service agents. Asking the customer unrelated or redundant questions does not value the customer’s time and leads to their frustration.
Simply transferring the conversation is not enough. Again, respect the customer’s time by passing through all conversation details. Name or identifying information and a transcript of the interaction including any possible solutions offered by the chatbot must be included in the hand-off. If an agent isn’t free or it’s outside business hours, create a case a customer service agent can follow-up on later.
Chatbot to the rescue
For the conspiracy theorists, one might ask if Gartner knew a pandemic was coming and chatbots would be the saviors of customer service. More realistically, Gartner was seeing the rapid technology gains chatbots were making and how well they would assist in delivering superior customer service by 2020.
If your customer service already relies on chatbots, congratulations! You are ahead of the trend. 2020 is less about deployment and should be more about refinement. If you have not yet put a chatbot to work, now is clearly the time. And once this uncertain period ends, it will continue to be a helpful addition to your customer service team.