Chatbots are a hot new technology for customer service, and their rise to power isn’t slowing. In fact, Gartner not only claims that over 50 percent of medium to large enterprises will have deployed them by 2020, but by that same year 85 percent of all customer service interactions will be handled by them. 2020 will be here sooner than we realize and with these predictions, one would think chatbots are seen as a positive direction for customer service.

Not quite. The truth is some opinions of chatbots aren’t spectacular and expectations vary in terms of how helpful chatbots really are.

Any new technology is going to have its growing pains and the naysayers that go along with that evolution. The right planning and characteristics mean a chatbot can provide all its promised benefits while sidestepping the drawbacks.

Goals

Businesses consider any new technology because it solves a problem. The first thing to do is understand what problem(s) exist that are best solved by employing a chatbot. Hint: it’s typically solutions to higher volume, more common issues that bog down customer service.

In addition to selecting the types of issues to solve, set some metrics. What percentage of the total service volume do you anticipate being handled by a chatbot? For individual issues, what is the breakdown of service by chatbot vs. other channels? How much more quickly are customer solutions solved with a chatbot vs. a live agent interaction?

Besides productivity measurements, be sure to consider customer satisfaction. When you initially roll out a chatbot, seek feedback from customers–and consider really overdoing it. With any new endeavor, it’s important to understand if the new service offering is on-course and working as intended on the business side while being an effortless, efficient, and perhaps even fun channel for customers.

Personality

Speaking of fun, all work and no play will make your chatbot dull and uninteresting so don’t let all that business stuff stifle it. Consider ways to bring your chatbot to life.

A good starting place is simply providing your chatbot with a name. Depending on your company or brand style, that could be playful or it might be more formal. Do you already have a brand spokesperson you could use? This will take some of the robotic out of this customer interaction channel.

Besides a name, what kind of voice should the chatbot have? Does it speak in a polite tone, or does it have a more relaxed manner (perhaps using emoji or slang)? Create an interaction style guide to ensure the chatbot’s scripted communication is consistent while avoiding anything that might be unclear or offensive to customers.

A great example of a named and expressive chatbot is Progressive’s Flo. Their chatbot builds on the popularity of the insurance company’s popular advertising character, Flo. While the Flo chatbot offers a combination of responses to both sales and service questions, you can also have some fun, such as asking her where she grew up and what type of car she drives.

Solutions

Customers want solutions fast. While they might take a moment to ask a humorous question, they want to get down to business quickly, find an answer, and move on with their day.

Much of the frustration cited in the articles above stems from chatbots being unable to solve problems quickly. Hand-in-hand with setting goals for the chatbot, they should always offer proven answers and get to them promptly. It’s for this reason that successful chatbots are those that focus on resolving a defined set of problems with proven solutions.

Solutions offered by chatbots can tap into other self-service options available. Knowledgebase articles are easier for customers to consume than having the chatbot parrot back a list of steps. Directing to automated solutions–forms to submit information or to perform other automation–are also ideal.

And just as a chatbot should be coded to solve problems, it should also be coded to throw in the towel when it doesn’t know an answer. If the customer’s problem lies outside the defined set of answers, the chatbot must know its limits and perform a handoff (more on that in a moment).

Learning

Chatbots are scripted to function based upon keywords and phrases used by customers. The marvel of language is the many ways in which humans can express themselves, and that creates a large limitation for chatbots–they don’t understand all those nuances. But they can be taught.

Periodically review chat transcripts of both successful and unsuccessful interactions. What were the words and phrases used? Use that information to improve the connection between how customers are conversing with the chatbot and the solutions it can provide. Interaction logs also provide a great basis for solutions to consider adding in the future.

Teamed with humans

Chatbots aren’t at the point yet to reason and troubleshoot as a human being can. Live customer service agents should be standing by and ready to step in. By not taking the customer in circles or asking unrelated or redundant questions, this will limit the customer feeling like their time is not being valued.

But it’s not enough to simply transfer the customer from robot to human. All of the interaction the customer has had with the chatbot–name or identifying information as well as a transcript of the interaction including any possible solutions offered–must be handed off to the customer service agent. To do any less is unacceptable because it would require the customer to repeat information they have already spent time sharing. It also gives the agent the background while preventing them from suggesting resolution steps already proposed.

No agents available due to high volumes or chats occurring outside business hours? No problem. The chatbot should offer to create a case for follow-up by a customer service agent at a later time. Expectations should be set as to when the customer can expect that contact.

The future is here

Robots are among us. The last few years have seen chatbots go mainstream in customer service, and they are working alongside agents to help solve customer problems. They will continue to take more and more of the mundane work off the shoulders of customer service and over time, they will be able to solve more and more problems and even diagnose and solve new issues.

Like any new tool deployed in customer service, a chatbot can easily fall below expectations without planning and purpose. By taking a thoughtful approach and imbuing it with the right traits, it will quickly become a productive and valuable member of the team.