Testing chatbots: Virtual customer service for the connected car

More and more car manufacturers are equipping their models with digital services. Apps and mobile services provide the driver with lots of information, but this also leads to new questions about things such as configuration or operation. How well can these questions be answered by chatbots?

A team at Arvato has done its best to find out. In fact, this was part of a pilot project in which Arvato has given one of the world’s largest car manufacturers advice on the digital transformation of customer service. This involves, among other things, the communication channels through which customer contact will be made in the future.

Chatbots, in particular, were tested very thoroughly. These virtual advisors are supposed to give end customers quick and comprehensive information that addresses their specific concerns. At the same time, they can provide a car manufacturer with insights into what issues and questions customers have on their minds.

In the first phase of the project, Arvato and the client selected a chatbot technology that performed well and was suited to the task. Then the bot had to learn. Just as with service consultants made of flesh and blood, chatbots must also be trained before they can do a good job. Real customer issues that the service team had already dealt with and solved helped towards that. On the basis of these real dialogs, the bot became familiar with different scenarios and possible solutions so that it could deal with questions. Then the actual testing phase began, in which the virtual advisor had to prove its abilities.

The pilot project showed that chatbots can answer even technical queries well and efficiently when it comes to standard questions. Special issues — complaints, individual cases in which the connected services do not work, etc. — are better dealt with, as before, by human customer-service agents. Because they are capable of empathy they can put themselves in the position of the customer and solve the problem creatively.

Tips for a successful chatbot

Another finding of the project was that there are several points to consider beforehand to ensure that a chatbot can operate as well as possible.

• The firts step is selecting the subject area in which the bot will work. This area should be clearly defined and make it possible, with relatively little starting content, to achieve a high probability of success with automatic answers. In areas with many individual aspects and special cases, either the number of correctly answered questions or the cost of the bot’s initial training is very high.
• The second step is the choice of suitable technology. For example, it is important to define what interfaces the end customer will use to communicate with the bot, and in what way: written, as in a conventional chat, or verbally via devices like Alexa, etc.?
• The third step is the training of the bot, in other words, the “education” of artificial intelligence. And this step is absolutely crucial for the success of a chatbot solution. Because the technology is initially only a shell. The better this is filled, the better the chatbot can meet the expectations of customer service.

Author: Editorial team Future. Customer.
Image: zapp2photo – Fotolia/Adobe Stock