Matt Peoples

How to Stop Your Chatbot From Derailing the Customer Experience

Share this article

Remember when most websites were hard to navigate, hard to read and featured color schemes of yellow and lime green? When the web was new, businesses rushed to deploy the technology before thinking about user experience. And in many cases, it was a disaster.

We seem to be in a similar phase with chatbots. While these automated conversationalists can help customers get fast answers, they can also trap people in dead ends, unable to find what they need and getting more frustrated by the minute. If you want chatbots to improve customer experience (CX), don’t rush to deploy; instead, incorporate thoughtful, user-centered design at every stage.

Bot the Right Thing

It’s tempting to launch a cool new technology all over your business, but make sure it can add value before you pop it up. The most successful bots perform clearly defined functions such as tracking orders or answering basic questions. Capital One’s Eno, for example, focuses on fraud prevention, sending confirmation messages for spurious transactions or alerts about changes in buying patterns.

If you’re new to chatbots, start small and low-risk. Rather than employing them in the sales process or for marketing, hone your skills on post-purchase tasks, like shipping updates.

Lavish Time on Design

Most bots don’t use artificial intelligence (AI) and natural language technology to “understand” user input and respond appropriately. Instead, they scan for key words and use a pre-defined decision tree to respond.

The decision tree is the brains of your bot, so you have to spend considerable time developing and testing it. The initial greeting and follow up responses should help guide customers to use terms and phases the system is looking for. For example, instead of asking a wide-open question like “How can I help you?” the bot should offer suggestions, such as “I can help you with an existing order, delivery options or products details.”

Bot “conversation” designers need to spend most of their time on error management. Humans have a huge capacity to interpret one phrase in many ways. Your bot needs to handle misunderstandings gracefully, which includes helpings users connect with a person when needed.

Partner With Humans

While people love bots for quick, easy answers, CGS’s 2018 Global Consumer Customer Service Report found that 40% of U.S. respondents prefer speaking with a person when they need more than basic information.

And someone who starts with a simple chatbot question may quickly find their issue is more complex than they thought. Or they may just have a question outside the bot’s capacity.

You don’t want to force people in the middle of a conversation with your business to search for an 800 number, wade through the phone queue, wait, wait, wait, and then start over with an agent. Instead, the bots should link seamlessly to a person.

Unfortunately, it’s impossible to know if a person will be available when a chatbot user needs one. So pair bots with a callback system like VHT Digital, which lets people receive a call from you as soon as an agent is available or at a time the customer chooses.

And be sure to close the loop. MyCustomer research found that the top complaint about chatbots is typing a bunch of information into one then having to repeat it all to a representative.

Your callback system should share that chat data with agents so the conversation can pick up where it left off.

Keep It Real

Because systems scan for keywords, you have to know what people are really saying when they ask questions. Use as much real-language data as possible to build the initial set of commonly-used terms and phrases. Recorded live chat conversations can help.

Then, as the app rolls out, you’ll want to record and review its interactions to find expressions missed in the initial development. Even bots using natural language need training in the terminology and idioms associated with specific companies or industries.

Build an Honest Bot

A lot of customer frustration comes from unmet expectations. If people see a chat screen pop up, they might think it links to a person and can answer any question they have. The bot will inevitably disappoint.

So let users know right away they’re talking with a computer and explain what it’s designed to do. With expectations set properly, people are less likely to get upset that an application for delivery updates isn’t able to answer billing questions.

Embrace the Bot

According to Juniper Research, chatbots in retail, banking and healthcare sectors will save businesses $11 billion a year by 2023. And bots will conduct online sales reaching $112 billion in the same time frame.

These handy little apps are here to stay. There’s no reason to avoid them; just take the time to design for optimal CX and you’ll get the most out of the chatbot revolution. Your customers will, too.

Related Posts