The Gist

  • Chatbots are not always effective. Although they're designed to help, the limited features of chatbots can frustrate customers and deter them from your brand.
  • Poor chatbot service lies in its design. Many consumers find themselves unable to reach a solution because chatbots cannot properly diagnose their problem.
  • Redesigning the chatbot experience — from both ends. Encouraging consumers to be candid and speak in their own words helps chatbots learn to more properly handle a problem.

Nothing tells a customer “I’m not really interested in talking to you" quite as bluntly as an interactive voice response (IVR). It’s the computer voice we all love to hate. Admit it, you've found yourself repeatedly shouting “agent!” into your phone more than once to get a human being to help you with a simple inquiry. The technology is fine for checking an account balance or finding out how much your next bill is going to be, but the communication fails when we need it to be interactive.

Despite the initial rollouts of self-service chatbots, which can use AI augmentation more effectively, it is intriguing to observe how brands are struggling to meet their customers' needs in implementing chatbots. Too often, the chatbot experience seemed to follow a similar model to IVRs — the only difference being that the client was typing instead of talking. But why? After what we've learned about AI over the years, how could a bad experience simply be replicated in another communication channel?

What's Wrong With the Chatbot Pathway?

Because it was done for the wrong reasons. When done right, self-service is a powerful tool to enable customers to ask questions and solve problems on their own terms, consequently improving the customer experience. Unfortunately, many brands saw in chatbots what they saw in IVRs: a means to cut costs by taking humans out of the conversation. This is ironic considering research by Vonage revealed half of consumers have abandoned a business when they reached an IVR, with an average loss to the business of $262 per customer. Because of this, we need to look at the root causes beneath poor chatbot adoption to avoid making the same mistakes (and losses of business).

Three important elements contribute to this poor execution.

  • You don't know your customer. First, many brands simply didn’t know their customers: who they are and what types of transactions they are trying to navigate. These brands often made assumptions about what a customer might want without any serious examination of performance data.
  • Following the "IVR tree." Second, brands continue to follow the “IVR tree” model of having a customer select the reason for the call (often from a small category), then prompting them through narrower subcategories after that. This is time consuming and often frustrating for customers, especially when they arrive at a dead end and can't back out.
  • Poor alternatives = frustrated customer. Finally, when they do arrive at a dead end, there is no alternative for the customer other than to restart the entire transaction and hope to find their ultimate destination through different answers to chatbot prompts.

The good news is that for each of these elements, there are three corresponding solution paths, which lie within the core skills and influence of customer experience professionals.

Related Article: Combining Self-Service, Chat and Phone Support: A Winning Strategy for Customer Service

Learning Opportunities

Solutions to Boost Customer Satisfaction

So what are the better solutions for a better customer experience?

  • Undertake a thorough analysis of unstructured voice of the customer (VoC) — particularly unsolicited, unstructured VoC. This is going to come primarily through call recordings, live chat transcripts and emails. Don’t rely on agent-entered reason codes or post-transaction survey responses. By nature, these responses are providing feedback to what they think should be important to your customers. Consumers are truly candid in their unsolicited inquiries; these sources also reveal the actual phraseology your customers are using, which leads us to the second solution path.
  • Empower customers to speak in their own words. With the always improving quality of AI to manage a conversation, there is no longer an excuse to ignore natural language processing (NLP) as the primary customer input to your chatbot. Trash the worn out IVR tree models you have been using and let customers type the way they talk. With just a modest amount of AI training, your chatbot can be adept in recognizing customer intent even when not explicitly stated. For example, “my plastic isn’t working” is a reference to a credit card problem. NLP empowers customers to go straight to their “job to be done” without the waste of time of layers of multiple choice questions that can lead them to a dead end … which is addressed in the third solution path.
  • Move over to a human agent at the first sign of customer confusion. Don’t allow your customer’s blood pressure to rise by having to “yell” for an agent if they get stuck. Set up thresholds and triggers to identify when customers have made multiple attempts to seek clarification or have expressed frustration. At this point, it is crucial to promptly transfer the chat session, along with its complete dialogue history, to a human agent who can resolve the transaction swiftly and effectively. The resolution of an inquiry is the most significant factor in determining overall satisfaction with an automated transaction. Customers will thank you, and you will learn more about just where to tweak your chatbot intelligence for future engagements.

Related Article: The Bittersweet Story of the Restaurant Kiosk and Customer Experience

The Future is Self-Serve

The concept of businesses pushing customers to self-serve began decades ago, particularly with gas stations. Few readers may realize there was a time when a driver never exited the car to refuel — an attendant did it instead. Now (except in New Jersey and Oregon), almost all refueling in the US is self-serve. And while the initial motives for this practice were to reduce costs, customers eventually evolved to prefer self-service to save time and avoid waiting on another human.

Therein lies the win-win proposition for chatbot implementations: instead of pushing customers to self-serve in order to cut costs, brands should aim to pull customers to self-service channels by offering choices that benefit them in the form of reduced complexity, cost and time savings. As a result, those brands will create more return customers, improve brand loyalty and, in possibly another bit of irony, reduce their costs of operation. Everybody wins.

fa-solid fa-hand-paper Learn how you can join our contributor community.