Chatbot: How does this conversational agent work?
Have you ever had a conversation with a computer program? Probably, and you may not even have noticed! Indeed, chatbots, or conversational agents, are increasingly present on many websites. These are computer programs with or without artificial intelligence, which can lead to a written or even oral conversation. They are developed to imitate the behavior of a human in a discussion, as accurately as possible. A chatbot can be a good customer service tool, for example, to respond to requests for reservations or information.
A little history
In 1966, engineers from MIT developed ELIZA, a computer program that tried for the first time to conduct a conversation like
a human. Although this program uses a fairly simple possibility tree, its answers were enough to fool some testers. Its development and structure served as the basis for all of the following.
In 1994, the term “ChatterBot” which later became “ChatBot” was used for the first time to describe the type of program.
Since then, the technology has advanced well, and more and more companies use chatbots, or conversational agents, for their internal management or to communicate with their customers. According to a 2016 Business Insider survey, more than 80% of companies would like to integrate a form of chatbot use by 2020.
Chatbots and AI
These conversational agents are often associated directly with the concept of artificial intelligence. However, most only offer predefined answers in response to certain keywords, which does not imply any notion of intelligence. They are programmed on the basis of several scripted question/answer scenarios in order to best respond to basic and repetitive or answer the same question dozens of times.
user needs. They cannot, therefore, adapt their response to the level of complexity of the request.
Despite their effectiveness, these chatbots reach their limits fairly quickly. Indeed, everyone who has already spoken to a chatbot knows that they are not foolproof and that they reveal their robot identity quite easily. Nowadays, users don’t expect chatbots only to give them ready-made answers. They want communication tailored to their requirements. To reach this level, chatbots must be able to learn and analyze user needs. They are therefore increasingly integrating some elements of artificial intelligence, in particular, machine learning.
Machine learning techniques allow a program to learn on its own and evolve without human intervention. Thus, a chatbot using machine learning will be able to analyze and process the data it receives. Through this analysis, he will be able to better understand user questions and provide better answers. The automatic generation of responses, made possible by this data processing, gives chatbots significant added value and opens up great possibilities.
Usefulness and challenges of setting up a chatbot
A chatbot has the advantage of being able to respond to user requests 24 hours a day, 7 days a week, and very quickly. It can even support all users at the same time. Communication is personalized since the customer obtains a direct answer to his question. From a customer service perspective, this is ideal! What’s more, according to a HubSpot survey, 40% of consumers have no problem being helped by a chatbot instead of a human.
Many companies use chatbots for time-consuming tasks like booking reservations or basic customer service assignments. This allows employees, for example, to avoid wasting time on basic and repetitive assignments, or to answer the same thing dozens of times to the same question.
At the moment, the majority of chat agents cannot handle overly complex requests. They can direct consumers to human contact for more specific or sensitive needs.
Before integrating a chatbot on its website, you have to ask yourself if it can really add value for users. Indeed, he should not be content with repeating information that is already easily found. In this case, it could become a source of frustration for users. Also, if a chatbot is dealing with anonymous users, it will likely receive messages that are not related to the business. He must be able to redirect gently, but firmly to the subject of the request.
Finally, Chatbots are fascinating tools, the developments of which surely have not finished surprising us!