Online chat has become an indispensable part of our daily lives. In 2018 just 15% of customer interactions took place via machine learning applications, chatbots and mobile messaging. Research company Gartner predicts that percentage to grow to 70% in 2022 and 80% of the companies that participated in an Oracle study indicated that they prefer to use chatbots to answer customer questions online.
There are numerous channels that cater for chatbots, such as WhatsApp, Facebook and Webchat, but it’s important to remember that being present on the channels where your customers are active is the key to successful interactions.
The three categories of online chats
Online chats can be divided into three categories:
• Human chat – Communicating with a customer directly
• Fully automated – Applications such as Alexa and Siri, which are very good at performing specific tasks
• Hybrid – A combination of technology and human
A human chat is relatively simple, if labour intensive and expensive. All you need to do is hire a person or company to handle your live chats. You’ll be able to deliver a great customer experience because you have a human to personalise your responses. A human-operated chat is especially good at answering more complex questions or responding to dissatisfied customers. However, using this kind of chat, customer wait times are often longer and the cost of using humans for every chat are often prohibitively high.
It’s not as easy to start fully automated chats, because you need to develop an efficient algorithm. At the same time, the customer experience is never flawless, because it’s almost impossible to have the right answer for every eventuality. One major bonus, however, is that the cost per chat is significantly lower than for human chat.
In the hybrid method, relatively simple questions are answered by a chatbot and more complicated questions can be transferred to a human. This results in lower costs than 100% human-operated chats.
The best of both worlds
We believe that the hybrid chat offers the best of both worlds. The frequently asked questions can be answered by a chatbot; these could involve questions about the status of a delivery, return of goods, or the availability of a different size/colour of a product.
If the questions get too complex for the chatbot to answer – for instance if someone has difficulty purchasing online tickets – the conversation can be transferred to a human.
The hybrid chat method also works the other way around. Suppose that you’re a customer in an eCommerce environment and you're engaged in a live chat with a person about a product that doesn’t meet their needs. If both customer and assistant come to the conclusion that the product has to be returned, the conversation can be transferred to a chatbot to guide the customer through the returns process. This reduces costs because by automating a binary process like product returns, you're freeing up the human to move onto the next customer more quickly.
Twice as many questions answered
One of the developments that’s improving efficiency within organisations is the ability for chatbots to reply to long tail keywords. The computer searches the databank to find out what the appropriate follow-up question would be and the chatbot replies with that question. This lightens the help desk’s workload and ensures they always have access to the correct answers to customers’ questions.
A good example is ANWB (Dutch roadside assistance). ANWB has been using Iris since 2012 and the power of this virtual assistant became apparent in the summer of 2019. Before this busy period started, ANWB did some research on how to create interactions between Iris and customers that were as human as possible. As a result, Iris was tweaked to ask more questions, including using positive affirmation in its replies. Subsequently, Iris answered twice as many questions about holiday preparations as in the previous years and the negative feedback was halved to just 6%.
The future of chatbots
Chatbot technology is advancing at a frantic pace. Thanks to Artificial Intelligence and Machine Learning, chatbots are becoming more and more adept at understanding language and sentiment and are consequently better at assisting people. A bot that can analyse sentiment can interpret the chat user’s mood and tailor the conversation accordingly. If your chatbot can read sentiment, there’s also no need to ask the customer how they felt about the conversation.
Another interesting development is conversational commerce, which includes instant payments in the chat. If a customer wants to know if a certain product is available in his or her desired colour, the chatbot can assist the customer and ask if they would like to pay instantly. If the answer's 'yes', the chatbot is able to redirect the customer to an integrated payment solution to complete the transaction
Apple is using this technology in Apple Messages for Business. Their customers can make payments with Face ID. Google is using RCS (Rich Communication Services, also called SMS 2.0) and a Google Pay integration.
There’s also been a lot of research on the predictive ability of machine learning to gain more insight into what people want in a chat. If, for example, someone uses the chat more than once to ask questions about a specific product; a specific brand of shoe, for instance. If you know that person’s shoe size, you’ll be able to give personal advice as well as targeted offers for a particular season or use case. Machine learning also works the other way around. If a person indicates that they’re not interested in promotions, then you can ensure they receive no personalised advice or offers. Personalisation goes hand-in-hand with privacy.
Regardless of how machine learning develops over the coming years, finding the right balance between humans and technology is crucial. Get the balance right and you have the perfect marriage of efficiency and customer satisfaction, with all the brand sentiment and advocacy benefits that go along with it.