Lockdown Scenario 1: Barely two days into the COVID-19 induced lockdown across India, no online transaction could be carried out with a newly-arrived bank debit card. Panic set in as online bank payments were going to be the lifeline during the long lockdown period ahead. Frantic calls to the customer support helplines went unanswered; so did direct calls to the bank branch.
As the hapless cardholder was scanning the FAQs available on the bank’s website for a chance troubleshooting suggestion, pop came a chat window on the screen with a reassuring message: “Hi, this is Nisha. How may I help you?” After a brief exchange of three conversation pings, the cardholder was directed to the relevant section of the website from where he could activate an obligatory limit for online transactions – a recently introduced RBI mandate – and that made his new debit card operational.
Lockdown Scenario 2: The self-employed designer working from home woke up one fine morning in the middle of the lockdown to find out that certain adjustment features on her expensive DSLR camera were not showing up on the control touchpad. After several futile switching off-and-on attempts, the manufacturer’s toll-free helpline was the final recourse. Every line was busy and even when the designer could finally get through, an automated message greeted her to inform her that owing to the unforeseen COVID-19 situation only “emergency service” calls were being attended to, the details for which were available on the company website.
Wondering what kind of emergency services a camera brand could possibly offer during a pandemic, the designer skeptically accessed the manufacturer’s homepage and navigated to the customer support section. Much to her surprise, there was a link called “Chat with Our Support Executive”. The executive on the other side of the chat window pinged a series of diagnostic questions and then guided the designer step-by-step to reset her camera. Upon rebooting, all her familiar feature icons sprang back to life on the control panel.
Both the scenarios above have been taken from real-life situations faced at Kolkata during the ongoing nationwide lockdown. Both the complainants got their issues sorted out through online chats. And in both cases – hold on dear reader – there was no one on the other side of the respective chat windows; no bank executive or camera mechanic, in fact no human at all! The solutions were offered by a machine algorithm, an AI-powered Chatbot.
Chatbots are just what the name suggests: chatting robots. By “robot”, here we mean an algorithm capable of natural language processing and applying AI logic to suggest a solution to the issue being faced by the customer.
At the back end, there are two massive repositories regarding the product or service in question – one comprising all possible problem terms a customer might key in and the other made up of all probable solutions being offered by the company in way of troubleshooting, based on their in-house technical know-how. The AI algorithm processes the complaint being typed by the customer in the chat window, identifies the key problem-terms in it, and compares them against all available solution labels from their repository to come up with the closest match.
The beauty of the entire thing is that the conversation almost perfectly mimics normal human speech – with natural greetings and grammatically perfect complete sentences. Responses are framed in a way that the complainant is encouraged to open up, thinking a human representative is behind the tool – thus providing enough diagnostic keywords that help in accurate troubleshooting. So, although there was no “Nisha” behind that bank chat window, the customer came away with the sense of a positive and personalised service response.
Such AI chatbots are results of painstaking data gathering, and phenomenal progress in natural language processing algorithms. On one hand, there is the specific technical aspect in which customer feedback and complaints are gathered over time, analysed to identify diagnostic keywords and fed into the system. On the other hand, there is a more ambiguous aspect of capturing the essence of human verbal interaction – anticipating possible customer reactions and preparing a set of pertinent responses for every conceivable reaction. The potentials are alluring, and the benefits may be staggering indeed. It is not for nothing that technology giants, including Google and IBM, are investing heavily in AI chatbots.
As the pandemic situation prevented tele-support executives from reaching their designated call centres, and as customer-care numbers across the globe went unanswered – chatbots did save the day. So, can they totally eliminate the need for human support? Are call centres going to fade away as more and more chat windows start popping up all over our screens? Not yet, perhaps.
While natural language processing still awaits a multitude of refinements to fully grasp all finer shades of human conversation, there is also the issue of linguistic diversity. In a country like India, with over 26 formal vernaculars and hundreds of variations of each, linguistic algorithms still have a long way to go before they can cater to everyone. And, for problems where critical judgement and situational adjustments are necessary – there is still no substitute for human intervention.