Chatbots are seemingly everywhere. We interact with them, we talk about them, we complain about them, we often joke about them. Yet if chatbots are so widespread, why is it so rare to find successful examples?At DigitalGenius we’re asked all the time if we’re a chatbot, and/or how we’re different. Yes, we actually did build one of the first chatbots four years ago, and have learned a lot since then about applying machine learning and process automation toward solving the most painful customer service problems.
In this blog, we’re going to examine different elements of chatbots, to demystify their role and function as much as possible, and hopefully illustrate how best to apply machine learning to successfully automate processes in customer service.
A chatbot is an automated or pre-programed interface, typically found on a website or on messaging channels like Facebook Messenger, Skype or Slack. They can be powered by technology as simple as keyword-based rules and scripting, or as advanced as Natural Language Processing (NLP) coupled with dynamic dialog flows.
Therefore, chatbots are standalone conversational interfaces, yet it’s safe to say that not all chatbots are created equal. If the chatbot targets keywords, as most do, it will by definition not be as capable if it had an actual AI engine trained on a company’s own historical customer service data. It’s an important distinction, yet it does not fundamentally change what the chatbot sets out to do: independently handle customer conversations.
What’s My Purpose?
This question is not an exclusively human preoccupation; it ideally should be answered by chatbots as well. Chatbots must deliver a clear outcome; they must have delineated boundaries; and there has to be a defined raison d’etre. Is it there to help you make a purchase? To provide you with basic information, such as opening hours? To provide up-to-date order tracking? Tell jokes? You decide, but chatbots must have a purpose.
This is where chatbots often fail, by trying to boil the proverbial ocean. Providing a cogent and consistent experience for a single use-case is far more desirable than approaching conversations in one-size-fits-all manner, because there’s simply too much variation in human behavior and objectives. Chatbots cannot keep up with us.
Beyond this, a company needs to decide on the application for its chatbot, and all too often, the bot can’t be tailored for multiple use cases. Is it there for marketing, for customer care, or to enable a transaction? If it’s not clear, customers will - and do - ask the wrong questions, and will often be disappointed; or worse, frustrated and annoyed.
(Don’t) Just Do It
Companies are beginning to resist the urge to take on a chatbot project merely because of their hype, especially in the customer service arena. This is an area with incredibly high stakes for customer expectations, retention and competitive differentiation.
Research has shown that 49% of customers do not want to be served by a Chatbot at all, and that a whopping 67% of customers still prefer to interact with a living, breathing, human agent, whether by phone, email or chat. The people have spoken!
Experience is King
Experience will soon overtake price and product as the #1 reason customers buy; it’s projected that by 2020, 86% of customers will choose to pay more for a superior customer experience. Why jeopardize great experience with a chatbot as your primary means of customer interaction? Nearly half of customers don’t want a chatbot experience because they expect to be listened to, not pushed away (read more about deflection here).
Quality of service plays a key role in reputational integrity - 68% of customers recommend service to friends based on quality of service, above convenience and price. Deliver great service, and customers will come back - and because chatbots can often compromise customer experience, those turned-off customers, and turned-off potential future customers, are put at risk.
A Vision Without Strategy is an Illusion
The reason companies consider chatbots are that they’d very understandably like a way to serve their customers faster, with greater consistency, and make life easier for those customer service agents tired of never-ending, mind-numbing repetition. At DigitalGenius, we strive to make this vision a reality, too.
Yet ask yourself: what role do chatbots actually play strategically? They tend to be standalone interfaces that are uncoupled from a centralized view of the customer, as well as the customer’s purchase history, preferences, and so on. This means in a chatbot conversation, you’re asking the customer to re-introduce themselves every time, while valuable interaction data that should be automatically collected is lost. That just shouldn’t happen from a business nor an experience point of view.
Chatbots also scale poorly across languages and channels. You may identify a provider with bona fides in English conversational NLP, but what happens when you expand to new markets and languages? (read about an alternate AI-powered approach to language here). What happens when your customers in those markets prefer to communicate over email, which render the chatbot capabilities redundant? How will you cope with growing volumes with that standalone solution in place?
If the goal of launching a chatbot is to stand out or attract attention, publicity and new customers, that’s almost certainly not sufficient to qualify for a strategic customer play - one that serves to make your company’s customer service a competitive differentiator. It does, however, qualify as a gimmick.
It’s What They Don’t Tell You
Inevitably, there’s a great deal of work in store once a company’s chatbot actually goes live. There’s diagnostics and troubleshooting (and make sure you read the fine print; not all vendors provide support nor customer success resources), and then there’s the continuous maintenance.
You’ll need to identify and add new intents; fine-tune keywords, phrases and other triggers; and new dialogs will have to be designed, uploaded, tweaked and continuously refined. This work, in addition to measuring performance, collecting and saving relevant data, and gathering customer feedback, will likely need to be done on a weekly or bi-weekly basis. Or you can leave your new chatbot alone for months, and guarantee a terrible, unrefined user experience. Wait - you thought this chatbot was going to solve all your customer service headaches?
There is Hope
Engine, the service design consultancy, recently found that 62% of customers want simple and flexible customer service options, far above the number of customers who wanted new toys like chatbots (a mere 15%). Customers just want to have their problems understood, then resolved quickly and painlessly.
We believe the same, from learned experience with over 50 customers. Our platform is purpose-built to understand your customers, in any language, across text-based channels, and to fully resolve the most repetitive and process-driven queries in your contact center. Once queries become complex, they go straight to an agent, without delay or unnecessary repetition.
We want to make things simple for you, your customers and your agents. Machine learning models learn from historical data and get smarter over time. The approach here is geared toward saving time, scaling with your business and enabling you and your team to focus more on delighting your customers.
Chatbots have their time and place, but if you're looking to them as anything more than just another channel for very simple keyword-based queries or appointment setting, you're looking in the wrong place. Customer service certainly doesn't become your competitive differentiator with chatbots alone.
If you want to learn more about our own journey with chatbots, our success with Customer Service Automation, or are eager to discuss the core differences between them, please get in touch - we’d love to hear from you.