A chatbot is a computer program that simulates human conversation either through voice commands or text chats. It is an artificial intelligence (AI) feature embedded in applications and digital platforms. The automated program interacts with customers like a human representative of the company would. The more ‘human’ a chatbot, the better it will be at interacting with customers and resolving their queries.
With businesses increasingly moving to digital platforms, many of us have, at one time or the other, interacted with chatbots. Our experience may have been varied. Earlier, chatbots operated within fixed guidelines. But now, with advanced AI technology and machine learning, chatbots can go beyond fixed instructions and adapt by using algorithms, statistical models and data.
Efforts to further improve chatbots are ongoing. At IIT Delhi’s Department of Management Studies, research scholars Amit Kumar Kushwaha and Prashant Kumar, led by Prof Arpan Kar, have listed factors that can improve the experience of chatbot users.
They analysed over 2.5 lakh users’ posts in B2B businesses to understand the experience creation process in chatbots. They distilled the data to pinpoint the parameters that influence customer experience and identify significant trends and emerging patterns. Their findings have been published in the Industrial Marketing Management journal.
Kar and his team touch upon aspects that need attention from developers. They found that in AI-based chatbots, the quality of service is pertinent. If a customer is satisfied with answers, he/she will believe in the system and will use it in future.
Equally important is trust. This can be built only if the answers given by the chatbot are credible and reliable. The research revealed that the computer-based system must pay attention to the visual content. Thus, a good picture of a product with the descriptive points displayed may be more easy on the user than a long stretch of text that does not provide clarity. Speed of interaction is also important to gain users’ confidence in the system.
Privacy of users, the research revealed, must be respected. For training purposes, chatbot developers often use data from different users. Therefore, to earn the user’s trust, it is important to specify that customer data is used in metadata format, where privacy-related information is protected.
The AI-based chatbot must be constantly improved, based on customer feedback, through the co-creation process.
While most other research has focused on the algorithmic perspective to improve customer experience, the IIT research has also factored in management principles, including a firm’s related scenario from past data, which directly influences customer experience and brand trustworthiness.
So, what would be a model chatbot? According to the team, “For a better customer experience, the chatbot needs to focus on service quality, interaction speed, and the ability to provide answers and predict the query based on customers’ past questions.” A model chatbot, like a human customer service agent, should interac with the user, gain information and modify its responses accordingly. In fact, one of Kar’s team members, Shagun Sarraf, is working on humanising the chatbot even further.
“We are working on similar deep-learning projects with some Fortune 500 companies,” says Kar, but declined to share further details at this juncture.