Chatbots are becoming a common part of our everyday online interactions. They help us with directions, to order food, and increasingly, they provide customer service. Many commercial websites have a chatbot cheerfully greet us on landing, standing ready to respond to our every need. But are these chirpy programs actually providing good customer service? If you’ve ever interacted with one, you probably know the answer is no.
Poor interactions with chatbots are frustrating, and according to a 2018 survey, we have a clear preference for human customer service agents over bots. While having a bot available to answer simple questions at any hour of the day can improve the customer experience, they can be a liability when dealing with more complex interactions, or when they don’t pick up on what the customer is asking.
The first chatbot, ELIZA, was developed in 1966, and simulated simple human conversations with pre-programmed responses. While bots these days are arguably much smarter than ELIZA, today’s bots still work on this same principle of canned responses to pre-determined queries. Every possible question, phrase and its variants need to be coded, and anything outside its pre-determined decision tree will confound the bot.
Humans can be complicated and confusing, so it’s not hard to imagine how easily things can go wrong. A more complex query, or even simple corrections like “I meant a small, not a large” can be misinterpreted – “Okay, got it! That’s one small and one large.” At best, the customer will be frustrated while they persist with attempting to guide the bot to meet their needs. At worst, the frustrated customer abandons their purchase, and clicks away to visit a competitor site.
More sophisticated chatbots are developed with AI. They can learn, have access to more data, can process more complex tasks, and seem more human. Many customer service bots now draw on customer databases to hone their interactions—from our demographic details, to our purchase history and previous interactions with the brand. In theory, this data-driven approach should make the customer service experience feel more personalised, and drive sales with thoughtful product suggestions. The reality is often a deeply impersonal experience—chirpy bots regurgitating suggestions for a variant of a similar product we purchased 6 months ago, or something we clicked and decided we didn’t want. These experiences are often alienating, and generate distrust.
These problems are not insignificant. Customers are the core of a brand’s success, and leaving the important work of customer service to poorly-prepared software programs can hurt the bottom line. While we are increasingly accustomed to, and willing to accept interactions with bots, we still prefer human interactions for all but our most standard needs as customers.
While one bad experience with a customer service agent isn’t likely to deter us, we are much more distrustful of bots. One survey found that 73 percent of respondents would never use a chatbot again if they had a bad experience during their first encounter. That’s a pretty high fail rate, and should concern any company deploying chatbots as a core part of their customer service strategy.
Customer service and support is a key pillar of any brands business strategy—integral to winning and keeping loyal customers. Chatbots certainly have a role to play—they can be quick, simple solution to standard customer needs and cut the costs of resolving basic queries. And it’s not all bad news in this regard. Some organisations are getting it right. The University of Adelaide has had amazing success serving thousands of expected questions over Facebook Messenger. By adequately preparing their chatbot with scripts and contextual knowledge, the university reduced calls by 40% on that day of expected high volume, and increased customer satisfaction by 60% because they were able to serve answers to that particular query in under a minute, compared to 40-minute wait times in previous years.
Chatbots can offer a quick and immediate solution for simple questions, saving the customer time in waiting for a human to be available. It also frees customer service staff from repetitive work, allowing companies to redeploy attention onto higher value work. But you won’t be ready to deploy a chatbot until you’ve got a knowledge base already serving repeat issues successfully and the content is actionable by your customer.
Very soon, I’ll be launching an online course that will help you improve your self-service knowledge base and prepare you for providing customers a speedy and accurate chatbot option. You’ll learn:
- How to set the right strategy and culture, so your team keeps working the way you need them to.
- How to write & improve articles, so they’re findable and can be presented by a chatbot without extra effort.
- How to optimise self-service, so customers know it’s available.
- How to measure success, so you know what needs improving.