Actions and words are meaningless without a context. It is a common misconception that meaning is contained within words. Nothing could be further from the truth.
Think about the word ‘interest’. It may assume different meanings depending upon the context. In banking, the word ‘interest’ can be used to either refer to money paid at regular intervals for deposits or loans or, to express curiosity to learn about a product or service. Double entendre and other complexities associated with sentence construction muddles up conversations. It is only by remaining in context that anybody can make sense of it.
Similarly, in customer-brand engagements, contextual understanding and shared meanings are necessary to architect meaningful conversations. This is ever so vital in the case of chatbots. Unlike face-to-face interactions, visual cues are unavailable. Text and voice remains the only medium for a chatbot to analyze user sentiments. Hence, contextual awareness should be the root of its intelligence however quixotic that statement might sound!
From General Contextual Awareness to Specific Contextual Awareness A combination of operational factors like industry and geography function act as a premise upon which the product and service related knowledge is built. Basic domain modelling can endow chatbots with general contextual awareness. But that alone is not sufficient in giving a human-like conversational experience.
The difficulty that organisations face in engineering meaningful, intelligent and personalized conversational experiences is partly due to their limited understanding of contextual setting of a conversation. Every time a customer interacts with a brand, it is not an event happening in isolation. It is the residual impact of previous conversations.
Chat history is a crucial factor in shaping the landscape of customer-brand engagements. An enterprise chatbot can’t simply operate with a simplistic, need-based understanding of context. To put it in perspective, let’s analyze the following conversations between a banking chatbot and a customer.
The customer experience is far better if the chatbot could remember the previous chat and/or customer’s preferences. Such an approach can enhance customer experience drastically.
How to build a contextual chatbot? Context-aware chatbots look back into history and determine the most accurate response. Usually, the amount of history the bot can search can be configured.
Enterprises can harness customer information from third party systems such as CRM, sharepoints and chat logs to contextualize the conversation and quickly solve issues for an existing customer. In the case of new customer, there exist a plethora of tools that can identify a unique visitor.
Using Natural Language Processing (NLP) and Machine Learning, Chatbots like CogniCor understand what the user is trying to convey and generate precise response on the fly. Such bots do not follow a script/menu and are not keyword-based. With substantial amounts of training data and increase in conversations, contextual bots automatically self-learns what customers are asking and adds context to a bucket that the machine understands.
An enterprise chatbot need not be a super intelligent machine straight out of the science fiction. Rather, it should be an AI-powered system that has meaningful conversations with customers. Empowering chatbots with context awareness is a powerful way to engage customers with human-like responses, provide personalised information and drive higher conversions.
About CogniCor CogniCor is one of the fastest growing providers of cognitive chatbots for the financial services industry. A spin-off from Europe’s top Artificial Intelligence Institute, the company has done deployments across the globe and featured in Forrester, Gartner and IDC reports. Over the last four years, CogniCor has facilitated over 5 million interactions and delivered ROI worth $100m.
Originally published in Medium