The voice assistants Siri, Alexa or Cortana are now part of the everyday life of many users. Even if they don’t always understand everything straight away, they offer more than just the illusion that computers can now give intelligent responses to people’s enquiries and offer the right solution. More and more companies are taking advantage of this and are already integrating artificial intelligence (AI) today.
Shops and service providers offer their customers a competent contact person around the clock in the form of chatbots. Questions and problems do not just end up unanswered in an empty void, but clients receive feedback immediately.
The customer then feels noticed and taken seriously, which strengthens the relationship with the company. Repetitive questions can therefore be completely automated with chatbots. AI can immediately process the request in full. This not only strengthens customer satisfaction, but also relieves support teams that are sparsely staffed. On the one hand, companies can save costs, and on the other hand, customer service employees have more time to deal with individual enquiries and offer better care for customers. This way, the quality of customer service can be improved.
In addition to service and support services, chatbots are also increasingly being used for customer acquisition. The reason for this is that, here, too, many processes run in similar ways and can therefore be automated with predefined processes. When acquiring new customers, chatbots are a logical further development of mailbots that automate potential clients, but still address you on a personal level. Chatbots can thus cover first approaches to customers in lead generation, and the sales team can intervene as soon as the customer shows interest.
Chatbots can not only be used for communication between companies and private or business customers; they are also well suited for internal communications. For example, there are chatbots that are used for employee surveys or as a service help desk. The communication tasks involved here are comparatively simple, and the bots used can guide employees through predefined workflows. At the same time, people’s preferred way of communication has changed in the digital age. Chats and messenger services are the preferred communication medium today and have already replaced calls and emails in private communications. The popularity of chatbots is also on the rise among customers in corporate communication, and they achieve better results than circular emails. For example, chatbots also have higher response rates for surveys than classic tools.
The virtual psychotherapist ELIZA, which was programmed from 1964 to 1966, was the world’s first chatbot. It was a relatively simply scripted program that asked its users to keep talking or asked questions about the topics addressed using a thesaurus. Many people therefore considered it to be a real dialogue partner, at least for a short time. Only upon taking a closer look did the system’s apparent weaknesses become apparent. Nevertheless, ELIZA could not formally pass the Turing test, the measure of whether AI is equal to humans in its functionality. In the Turing test, the test person chats with a machine and a person at the same time. If, after intensive questioning, they are unable to determine which chat partner was the person and which one was the machine, the Turing test is considered passed.
The chatbot Eugene Goostman, which was developed in 2001, succeeded in this, at least to some extent. In a competition organised especially for it, a third of the test persons were not able to detect any significant differences between the human chat partners and the machine. Critics, however, stated that this was less due to its skills than the fact that the developers gave the bot the personality of a 13-year-old boy.
To date, no AI has been able to pass the Turing test. However, for chatbots, especially, that is not actually at all necessary. Society is increasingly accepting and using them. In a 2019 survey, 86% of those questioned stated that they would rather communicate with a chatbot on a company website than fill out a static contact form. Around three quarters of customers already do this on a regular basis. Many customers are now well aware of the advantages of this technology, and use chatbots even when it is clear that the person they are talking to is not a person, but AI. At the same time, positive experiences with the virtual assistants are increasing the demand for permanently available support. This can be seen, among other things, in the growing number of chatbots on offer. In 2019, around 300,000 different chatbots were available through Facebook Messenger alone, which were used by 1.3 million people every month.
Companies can get more information on the latest developments of chatbots and AI at the Chatbot Summit in Berlin and the Call and Contact Centre Expo in London, for example. On the other hand, the international congress for innovative customer dialogue CCW in Berlin or the Shift/CX conference in Frankfurt, for example, have focused on integrating AI for companies and service providers.
The easiest way of recognising the complexity of a chatbot these days is to ask a question that does not fall within the functional scope of the bot and does not appear in the knowledge database. Chatbot programs usually specialise in certain areas.
For example, a bookseller’s chatbot that helps you with ordering or can give book recommendations has no idea of current share prices, while the chatbot of a bank can immediately call up the relevant information requested. If you use company chatbots to request information for which they were programmed, you can only partially determine the difference.
Modern chatbots currently work in two different ways: on the one hand, there are scripts based on rules, and, on the other hand, there are also chatbots that are based on artificial intelligence that learn something new with each customer interaction. The basis for this is always the largest possible database of knowledge, also called a knowledge base. This helps the bot to recognise patterns and to find the right answer to them. The larger the amount of data stored, the better the chatbot will react to the many different text entries or requests.
The chatbots based on rules can only answer question-answer sets that have previously been added to the knowledge base. Therefore, they are not considered intelligent, because they cannot learn something new on their own, nor can they recognise unknown patterns.
Other chatbot models are based on artificial intelligence and use machine learning and, in part, natural language processing (NLP) to process the requests. This means that not only are text entries processed by a lot of rules, but the AI chatbots can access the knowledge base and give the impression that the dialogue partner is human. These models usually require large amounts of training data.
These AI-supported chatbot models are considered intelligent because the program learns something new over time and can thus recognise previously unknown patterns and respond accordingly. Over time, these AI chatbots can also learn to refine existing patterns and, for example, recognise different adjectives in relation to nouns. These AI chatbots offer an increasing rate of automation to companies and a continuous flow of conversation to customers.
However, since handling a conversation is a very demanding task and the technology of chatbots cannot yet automate 100% of requests, there are hybrid chatbot models as well. The aim is to combine the advantages of AI-supported chatbots, such as an increasing automation rate, shorter response times and a contact channel around the clock, with the capabilities of the human agent. This allows the chatbot to act independently and only activate the agent if the bot does not understand the question. This leads to a constantly growing knowledge base and automation.
Chatbots are already taking on a variety of functions and will continue to expand in the near future. Speech and pattern recognition will also improve further and thus increase the apparent understanding of programs. While chatbots are currently mainly used in messengers and on homepages, they will also “conquer” other media in the future and rise to become virtual assistants that will be available at any time. At the same time, dialogues are becoming more and more natural, and users are beginning to develop personal relationships with them. The expansion of specialised chatbots to versatile assistants who have answers to all questions has already begun.