Dr. S. Chellaiah is a professor, keynote speaker, IT consultant, and trainer. He has over 30 years experience in IT industry, academia, consulting, and training.

Customer Service has become omnichannel and it is truly a challenge to ensure high customer satisfaction. Automation is one answer, but it is both a boon and a bane.

Technology because of its demonstrated potential, increasing capabilities, and low cost, is replacing humans in many spheres. Customer service is one such prominent area. First, humans were replaced/assisted by IVR, then simple chatbots, voice bots, Intelligent Virtual assistants and now by Chatbots that use conversational AI.

The ubiquitous chatbots, are self-powered software applications that “chat” with customers like the way customer service representative would. Businesses are adopting chatbots in the guise of cost cutting and increasing customer satisfaction. But have they enhanced customer satisfaction is a question that can only be ambiguously answered?

Conversational AI refers to the branch of Artificial Intelligence that enables understanding of written and spoken human language. It involves Natural Language Processing and machine intelligence. By this, the software can understand what humans state, and “learn” as more and more conversations take place between the humans and computers; then it becomes “intelligent like humans”.

Conversational AI based chatbots using AI powered search can work across all channels presenting the same answer everywhere and every time; can function as agents and perform mundane or repetitive tasks allowing agents to handle complex queries and tasks; can be persistent and try many times to understand the chat; can be self-learning.


Chatbots understand the questions/comments posed by customer and search for an answer in the database and give the answer. Since they are not intelligent in the human sense, they are programmed to understand in one of two ways.

1) The chatbot reads the customer’s query/comment and tries to match it with those text in the database. When identical match occurs, it gives the answer. A customer wanting to know the balance in his account can ask, “What is the balance”, or “How much money is left?”, or “Tell me the remaining amount” etc. So, the program code has these variations built into it. This is a direct,linear match.

2) The chatbot uses machine learning techniques to understand the intent / context of the customer and then searches for that intent in the database to find answer. The chatbot must have as many possible ways of stating the intent as one can so that comparing is easy. After that, any additional
information that is needed/available can be extracted from the chat using named entity recognition (NER). Then the chatbot will answer the customer.


Most chatbots use English as it is the most popular language globally. But there are about 160 English dialects, various slang terms, abbreviations, and acronyms. This presents a complex challenge. For other languages, one can develop language specific chatbots or use a translation engine to translate into English in real-time using APIs from Google or Amazon Web Services. This translation must be as close to the original statement as possible. Inaccurate or incomplete translations introduce additional errors. A company called Language I/O does this and its product translates from 100 languages.


What if there are typographical errors? The Chatbot must be “able to correct the errors and understand the user’s intent correctly”. This is called normalization of the content. Only normalized content is fed to the Chatbot.


Voicebots can recognize human voice. The intonation, accent, and pronunciation, and dialects pose challenges in recognizing voice and translating it into text. Once this translation is complete then the rest of the process is same as for chatbot. Siri, Alexa, and Google Home and Google’s yet to be commercialized LaMDA are voicebots.

If customers find it easier to talk than type (due to limited motor skills like for senior citizens or people who are on the move like salespersons, or not so computer savvy), then voicebots are very useful.


With progress in voice recognition, real-time translation,advanced machine learning algorithms, and sophisticated processors, the voice bots will become intelligent enough to replace humans in the realm of customer service.