Welcome to a world where AI assistants have a personality, get your work done seamlesslyand often get mistaken for humans
NEW DELHI :
Amy Ingram is a secretary. She sets appointments and meetings. She works for top-level executives, including CEOs, in the IT, start-up and finance sectors. If you work in these industries, you’ve probably received an email from her.
But Amy is not a real person. “She" is the creation of x.ai, a US-based Artificial Intelligence (AI) start-up, and was built for one purpose: scheduling meetings. For people like Karthik Palaniappan, 38, CEO and founder of August Academy, a career counselling start-up in Chennai, who has to work with 40 clients a year and set up an average of 25-35 calls a week, the AI assistant is a bargain at $29 (around ₹ 2000) a month. “I run a consulting business and have online meetings with clients from across the globe in different time zones. That’s two or three iterations before a meeting is set, which is quite cumbersome. With Amy, I’ve offloaded this mundane task," he says.
Now all Palaniappan has to do is copy Amy on an email conversation with a client—the AI assistant takes over the scheduling part of a meeting seamlessly. Other than the ease, what he likes is that his contacts follow up with Amy thinking she’s a real person. “Ninety per cent of the time, if I don’t tell the other person that Amy’s a bot, they don’t recognize from her emails that she’s a digital assistant," he says. This makes clients feel he is making an effort to set up an appointment rather than delegating it to an app.
Every other day, clients who come to meet Ankur Warikoo, co-founder and CEO of Nearbuy, a hyper-local start-up based in Delhi, ask for his assistant Amy. “While almost 40% of the clients who come to my office to meet me ask for Amy, almost 70% of new employees we welcome ask for Amber, the lady who works in human resources," laughs the 38-year-old. Amber is an HR chatbot created by Gurugram-based InFeedo, a company that uses AI in HR functions. “Amber is a friendly spokesperson and a good listener, while Amy is polite and manages my calendar," he says.
As our phone screens, messaging apps and websites get inundated with chatbots, their “personality", or “conversational interface" as the user-interface design industry calls it, is the differentiator for brands and users alike. Amy’s personality (or her “brother" Andrew’s, who differs from her only in name) can change from courteous to friendly, depending on the situation. This is a major reason why people, too, respond with polite “thank yous" and “dear Amys". For Dennis Mortensen, CEO and founder of x.ai, a “thank you" mail is a marker of a job well done by the AI assistant. “We decided early on in our product (development) that we needed to humanize Amy and Andrew, give them names, key character traits, to the point where they even have LinkedIn profiles so the host and guests interact with Amy or Andrew as they would with a human assistant," says Mortensen in an email interview.
For this, his team hired writers with performing arts backgrounds to apply the right level of human-like empathy to the AI bots so they could understand and respond, and, most importantly, pass the Turing test. Developed by Alan Turing in 1950, the Turing test gauges a machine’s ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human. This turned out to be a great selling point for the assistant AI in x.ai’s early days and remains a talking point even today among early adapters like Warikoo and Palaniappan.
85% of all customer service interactions by the year 2020 will be powered by chatbots- Gartner, 2016 study
Giving chatbots a character and human personality helps create a connection with users, says Aakrit Vaish, CEO and co-founder, Haptik, a Mumbai-based start-up that builds chatbots for companies. Conversational interfaces are a popular requirement for most of his clients, whether they want a chatbot for their app or website. “A dry, staccato, formal chatbot, even though efficient and logical, fails for most of our clients," he says, adding that bots with a warm and friendly tone, which endear themselves to the user, are popular. It’s also less intrusive for users when they pop up on their phone or laptop screens.
Customers like Aurko Bhattacharya, co-founder of Mumbai-based instant credit facilitator ePayLater, prefer interacting with chatbots, rather than customer care executives, for mundane tasks like banking as AIs are “always pleasant", even if the task doesn’t get completed. “With a human customer representative, this can vary, as simple things like greeting, or how a query was answered or not answered, can make a huge difference," he says. Bhattacharya uses bots to pay bills, schedule meetings, find locations and interact with brands—and hopes to reach a future where AI chatbots come with a human face.
This may be just around the corner. In May, Google revealed its Duplex project—an AI assistant that not only mimics voice and speech patterns but adds pauses and fillers like “umm" to make speech sound convincingly human. Amazon’s Alexa Prize, an annual competition to advance human-computer interaction and build socialbots that can converse and engage humans on varied topics, has a whopping $3.5 million prize this year, demonstrating the importance of AIs with personality.
Bengaluru-based Niki, a shopping chatbot company with over two million users, researched how Indians want to interact with chatbots. It found that most customers, especially those from tier 2 and tier 3 cities, interact like they’re WhatsApping friends and prefer long, conversational messages. “The whole idea for us in creating Niki was to turn the cumbersome process of buying product and services online into a natural conversation with a chatbot, like you would have with a friend or talk to a shopkeeper," says its CEO, Sachin Jaiswal, crediting the chatbot’s conversational tone for its popularity.
For Warikoo, Amy’s name is the only thing about her that sounds fake. “It sounds like an alias and doesn’t induce confidence," he says, adding that he has already raised this with x.ai, but the company doesn’t have customizable names on its feature development list yet.
Though Amy’s error rate is quite high, at 20%—it can go into a loop and confuse clients—Palaniappan still prefers the convenience of a chatbot. “People take time off, have a life, and get bored of repetitive jobs. Amy schedules for me 24x7 and is polite," he says.