Robots are taking over the world because they have simply become too intelligent. They have constantly evolved and were finally superior to humans. This is a future scenario that can be seen in many science fiction blockbusters, but to what extent can one refer this to be the truth? After all, the relevance of chatbots and artificial intelligence (AI) in our everyday lives is increasing steadily: whether with Alexa, Watson or Siri, in marketing and sales or in medical technology – AI is taking place in more and more areas and supporting decision-making and personalised communication. Above all, the ability to evaluate huge amounts of data and recognise patterns in it within only seconds distinguishes the technology from humans.
“Our intelligence makes us human and AI is a complement to that.” Yann LeCun, Chief Artificial Scientist at Facebook AI Research
When having a look at Karl, Microsoft Office’s animated paper clip, it becomes clear, that communication and speech are more than just an exchange of information. Inaccurate orders by voice assistants or amusing responses by digital assistants – the use of AI does not always seem to work smoothly. So how intelligent and helpful are chatbots really?
Not all chatbots are the same
A chatbot is a computer programme with which humans can interact, imitating authentic communication in natural language. The first chatbots were based on a simple principle: the bot could identify individual keywords and then perform a predefined action.
Through the use of AI and natural language processing (NLP), chatbots today are capable of more and represent an intelligent interface between information, machine and human. The terms “chatbot” and “artificial intelligence” are not to be understood as synonyms. A chatbot is an application, while AI is the technology working in the background.
Not all chatbots are based on AI. The so-called rule-based chatbots are often used for simple applications, standardised questions and uncomplicated issues. By linking commands, rules and keywords, the appropriate answers can be provided for predefined questions. Thanks to NLP, the chatbot is given the ability to process, read and respond to human language naturally. Similar words and phrases that lead to the same answer are identified and the user is guided through a predefined decision path. If the complexity of the requests increases, it becomes more difficult to create appropriate rules for each scenario. Hence, the chatbot reaches its limits.
This is where AI comes into play: AI-based chatbots, or Conversational AI, are able to learn and search for information from one or more databases. First, the bot is provided with (training) data from which it identifies regularities, derives further answers, puts coherences into context and thus continuously develops.
Natural language, however, bears the risk of misunderstandings: dialects, emotions, spelling mistakes or irony are making it difficult to understand. Deep learning is an area of AI that makes chatbots smarter the longer they are in use. With every data input from users, they learn, improve and can react individually to specific situations. In addition, Conversational AI uses Natural Language Understanding (NLU) – with which it can recognise the language based on keywords but also understands statements contextually. Moreover, it identifies the mood of the user. The goal here is to be able to respond to the customers in the best way possible and thus offer him or her an optimal experience. If Conversational AI cannot answer the question, for example, the customer is directly forwarded to a service employee. If Conversational AI already knows all the answers to customer queries, the interaction with the chatbot is continued and valuable resources are saved in the service team.
The best of human and AI: a hybrid model
The hybrid approach offers a combination of the benefits chatbots offer and the capabilities of humans and Conversational AI. A hybrid chatbot first uses rule-based components such as predefined keywords and processes. Furthermore, interactions between agent and user are analysed by the hybrid solution from the AI and used as training data. In this way, Conversational AI learns with each conversation and can then automatically conduct the dialogue with the customer. If a request exceeds the knowledge base of the chatbot, the customer is forwarded to a suitable employee. Thanks to the collected data, the agent can directly follow up on the previous dialogue and provide appropriate solution suggestions in the live chat. The employees have full control at all times over whether and when they want to take over the chat.
In this model, human and machine work hand in hand. While the chatbot acts independently and answers standard questions automatically, the agents can concentrate on more complex customer enquiries. The workload for employees is significantly reduced, customer satisfaction is considerably improved and the quality of the answers remains continuously high.
In addition, thanks to artificial intelligence, the chatbot can collect valuable metadata about the customer, give them personalised product recommendations and pre-qualify leads for sales purposes. Thanks to AI, huge amounts of data can be analysed, networked and filtered, from which valuable patterns and correlations can emerge for corporate strategy. Future scenarios can then be derived.
Intelligence based on language and knowledge
Speech recognition and knowledge modelling make apparent how intelligent a chatbot actually is. Only when the intentions of the dialogue partner are recognised, corresponding information can be extracted from the database and delivered to the customer. Thanks to artificial intelligence, chatbots are able to analyse interactions and learn from them. An intelligent chatbot can only be as good as the data it is trained with. Further development can be achieved with the input of new data, which is provided by humans.
Chatbots support customers and employees in their everyday lives by recognising correlations that would not be apparent in manual evaluations. AI-based chatbots save time, nerves and money through automation. Moreover, they create customer loyalty due to fast communication and accessibility at all times. AI currently has no emotional intelligence and cannot replace humans in tasks requiring this characteristic. The true benefit of AI is therefore not to replace humans but to automate routine tasks efficiently. AI-based chatbots remain a technical tool for humans that offers the best possible support so that humans have more time to devote to tasks requiring emotional intelligence, empathy and experience. The perfect symbiosis between human and machine.