22 Jul Chatbots – When a robot answers questions

Have you ever chatted with the customer support of a company? The probability that your questions were answered by a robot is quite high. Whether on the company website, in online shops or social networks, automated dialogue systems are being used more and more. By 2020, 80% of companies will already be using chat bots – according to estimates by experts.[1]

What is a chatbot?

A chatbot (a combination of chat and robot) is an online dialogue system that can answer questions and initiate additional actions, depending on its characteristics. After initial programming, all this happens without any human intervention and is therefore also used as a cost-effective relief for customer support. Chat bots work in real time and can therefore react to questions late at night. In some cases, they are clearly identified as chat bots, e.g. by corresponding avatars. But since they basically act like humans, they can hardly or not at all be distinguished from them without such a distinction.

Chat bots in social networks are also called social bots (or social networking bots). They can generate greetings or respond to comments, and even posts or tweets can be created automatically using social bots.

Voicebots are voice-based dialogue systems. The input and output is done via natural speech and the processing can be realized thanks to Speech-To-Text and Text-To-Speech technologies.  The speech recognition software of Alexa and Siri is also based on this system.

For the sake of simplicity, the word chatbot is used in the following to explain the general functioning and possible applications.

From the request to the answer

If a chatbot receives a question, it is processed and answered using keywords, internet research and partly also artificial intelligence (AI). At first the text request is decomposed and analyzed according to programmed rules (e.g. if-then-commands), possible syntax errors are also corrected. From this text, the respective question is then filtered out and the answer is then searched for in knowledge databases and on the Internet. If a suitable answer could be found, it is subsequently prepared for output and possibly supplemented by additional data. After a final grammatical correction, the answer is then output to the question poser.

If a chatbot cannot process a question, the customer support will be informed or the question will be forwarded directly to them.

How intelligent is a chatbot?

But how much can the chatbot do on its own and how intelligent is it actually? The answer depends on the type of chatbot used.

If it is a rule-based chatbot, it can only recognize defined keywords and give predefined answers. Synonyms must also be taken into account so that, for example, both online shopping and online shopping are recognized and understood. However, if keywords appear for which no suitable answers have been defined, the rule-based chatbot reaches its limits and cannot process the request.

Intelligent chatbots are supplemented by artificial intelligence and machine learning and are therefore self-learning. With the help of Natural Language Processing (NLP) the natural language of humans can be understood and the actual intention can be explored. Through continuous training and analysis of the feedback, increasingly complex reactions are possible.

If it is a voicebot, AI is also used to achieve a voice output that sounds as real as possible. More and more, robots can then no longer be distinguished from the voice of a human being.

Possible applications of chatbots

As already indicated at the beginning, chat bots are now used in many different ways and implemented both on their own websites and in messaging systems. In the area of sales, they can be used to increase sales and acquire new customers and can be used, for example, to answer product questions or coordinate appointments.

Chatbots are increasingly being used in customer service, as in the long term they represent a more cost-effective alternative and can additionally relieve the burden on customer support. Simple standard queries can be processed completely automatically – even at night. Another advantage is the fast reaction of the bot, because companies that respond within 30 days have an average reaction time of 42 hours. [2] During this time, customers can become impatient, resulting in valuable leads being lost.

In marketing, the dialogue systems can be used for example in brand building and competitions and ensure high conversions.

Conclusion

In many areas, chat bots give the opportunity to help out, relieve customer service and lead to savings. A major criticism of the robots, however, is that their use in the appropriate places means that jobs are lost. However, it should also be mentioned that thanks to the bots, new jobs can be created in new areas simultaneously. It is certain that the dialogue systems offer a lot of potential for companies and 70% of Germans consider them to be useful for supporting ordering processes.[3]

[1] Kinsta:E-Commerce-Statistik für 2020

[2] Harvard Business Review: The Short Life of Online Sales Leads (2011)

[3] Statista:Umfrage zu sinnvollen Einsatzgebieten von Chatbots in Deutschland 2018