Voice assistants set to vacuum away chatbots in India

Sunny Sen January 15, 2018 14 min

Story Highlights

  • The number of voice bots users in the country is estimated to cross 135 million by the end of 2018, making India the hotspot for the new interface
  • Experts feel the adoption of voice assistants is a natural gravitation among consumers in a country that already tops voice searches – 28% – in the world
  • The main drivers for companies taking to the new technology are costs savings and quality of customer service — both of which were underserved by chatbots

Surabhi Sharma, a graphic designer in New Delhi, is teaching her 50-year old mother to book a taxi on Ola using Alexa. Sharma’s mom is not new to the internet – she uses Facebook and WhatsApp but is still reluctant to make e-commerce purchases or book a cab using an app on the phone.

“Alexa, ask Ola to book a ride,” Sharma tells her Amazon Echo smart speaker. Her mother is amazed that a voice command can get a cab booked and is gobsmacked when told the cylindrical speaker can book air tickets, find recipes, listen to the latest news and astrology forecasts, and order food.

The 50-year old housewife from East Delhi speaks with Sharma every evening on a WhatsApp video call. “When you hear and see someone, it is so much real… chatting (on text) isn’t,” Sharma says. “Look at Alexa. You just need to speak to a device, just like talking to a person.”

Sharma and her mother are among millions today making a pivot towards voice, which they say is the most natural interface between humans and machines. For Sharma, it is the biggest revolution in the world to tech after smartphones.

New Delhi-based graphic designer Surabhi Sharma, seen here with her Amazon Echo, insists the voice interface is the most natural and scores over textual chat | Photo: Sunny Sen

French consultancy Wavestone estimates that between 2016 and 2017, the number of virtual or voice assistants globally went up to 710 from 504, and is expected to go up to 1,016 in 2018, and 1,831 by 2021. These voice assistants are led by the popular ones Google Assistant, Amazon’s Alexa, Siri of Apple, and Microsoft’s Cortana. By 2020, half of the searches on the internet are expected to done on voice.

In India, the shift to voice assistants is expected to be faster and more definite. For, most textual chatbots are in English, a language that under one-sixth of the country’s population is versed with. With voice assistants, combined with artificial intelligence (AI) and machine learning, machines are learning to understand regional languages – especially the popular ones such as Hindi, Tamil, Telugu, and Bengali.

Consultancy Accenture predicts 39% of India’s online population will use some kind of voice assistant by end of 2018. The country has some 325 million broadband users, as per September 2017 data from the Telecom Regulatory Authority of India, most of who access data and the internet from smartphones. Assuming a growth of 8% and combining it with the Accenture projection, India will have over 135 million users of voice bots at the turn of this year — making it the world’s hotspot for the new interface.

Voice bots or assistants are AI-powered voice applications, which collect data and answers from the web and also on-premise information and completes a task like booking a cab or ordering food — a role that chatbots were expected to dominate until about two years ago.

Hundreds of companies in India (many who never built a chatbot) are getting on to Alexa and using Google Assistant to have their own voice assistant (that makes it easier as the technology is available off-shelves from Amazon and Google).

Chatbots on the wane

In India, according to Google’s internal estimates, voice searches already make for about 28% of overall searches; the rest of the world averages around 20%. While Google did not respond to the queries of FactorDaily, data shows that there is a clear migration towards voice-based interaction with computing devices.

That wasn’t the case two years ago. Almost anyone who had a customer-facing interface was betting on chatbots — banks and financial services companies, e-commerce firms, media houses, among others, and even Facebook. Chatbots are automated text conversations using AI to perform tasks.

But chatbots seem to have had a limited lifespan – particularly in India. And, voice bots are fast taking their place. As millions of Indians access data on mobile phones for the first time, voice is increasingly becoming the natural interface with devices leading to the country becoming the hotspot for voice assistants in the world.

Besides, chatbots have a higher failure rate – a reason why Facebook is shutting down its chatbots. According to this report, Facebook’s chatbot faced 70% failure rate.

India is known to jump tech cycles. While the world was still accessing the internet on laptops and web browsers, India moved to become the No. 1 country with mobile broadband access. “A similar thing is happening from chatbots to voice assistants,” says Dilip R S, country manager, Alexa Skills, Amazon India.

Dilip R S, country manager, Alexa Skills, Amazon India, says in adopting voice assistants India is leapfrogging over chatbots. | Photo: Rajesh Subramanian

There are already about 100 voice assistants operational or being developed, and many more in the works. Compare that to chatbots, about 1,000 chatbots in India, and voice bots may seem behind the curve but their adoption is at a faster pace, says Sandeep Kumar Sharma, associate research manager – software and services at tech researcher IDC.

“Over the last four to five years we have seen the entry and gradual evolution of chat. But still many don’t use it. In the next 18 months voice (assistants) will take over chatbots in penetration. The penetration rate of chatbots is high right now, but in the past six months voice assistants have seen a steep growth,” he says.

Other analysts, too, predict that voice will become bigger than chatbots. “India market voice assistants will overtake chatbots. India has always been a voice-driven market,” says Neha Dharia, senior analyst with London consultancy Ovum.

Primary among the reasons for the decline of chatbots is the user experience. “When you design for chat, you restrict the experience with text, especially with touch and tap… voice assistant is like talking into a black box and it answers back,” says Dilip.

Dharia says that orders and billing enquiries will be among the early set of customer actions to move to voice. “Anything which is to do with fetching information will happen sooner… troubleshooting will be much later on and complex information will take time,” she says.

Early, enthusiastic movers

Companies using both chatbots and voice assistants say they are gravitating towards the latter. Restaurant search app Zomato is one such example. “Its very early to disclose numbers, but compared to chatbots, voice has seen a bigger uptake. Over time voice will be the way to engage with customers,” says Pramod Rao, head of content and marketing, Zomato.

Many other companies are building voice assistants or are already in the market, too: HDFC Bank, Ola Cabs, Uber, Saavn, Times of India, Aaj Tak, Byju’s, among others.

The rapid growth of voice assistants on Amazon’s Alexa platform is indicative of the demand. In the first quarter of 2016, Amazon had 135 Alexa skills or apps, which were using the voice assistant. That number has already surpassed 10,000 today.

On the other hand, Google’s big bet for the Assistant is on localisation. It is folding its India search intelligence into the Assistant. For example, one can search for Italian restaurants around you or parantha joints in Chandni Chowk in the Assistant app. It is also bringing the Assistant to a $23 phone. “Indians will soon have easy and fast access to all types of apps. Once an app works with the Assistant you can just tell your Assistant to connect you with the app of your choice with a simple voice command,” Brad Abrams, group product manager, Google Assistant, wrote in a blog in November.

Cortana, Microsoft’s voice assistant, can be downloaded through Android and iOS, but the tech major hasn’t been able to make a dent in India yet. Apple’s Siri can open apps, answer some questions, even type out a WhatsApp message, and understand English in an Indian accent. Both Microsoft and Siri haven’t disclosed any major plans for Cortana and Siri.

With a voice bot, orders can be placed without even touching a smartphone. FreshMenu, a food ordering site, was one of the first companies to integrate the Alexa voice assistant. Using the voice assistant, a FreshMenu customer can reorder the dish she had the last time or ask the assistant for a preparation with a desired ingredient.

A customer can pay on the FreshMenu assistant using the Amazon Pay wallet, which is integrated into Alexa. “As payment solutions get better, voice assistants will find a large number of use cases in e-commerce,” says Rashmi Daga, founder of FreshMenu.

Hemang Pandit, managing director of astrology platform, Ganeshaspeaks, believes AI will help understanding the emotions of a customer and help with complex predictions

The early shift towards voice assistants in India includes niche players. Hemang Pandit, managing director of online astrology platform Ganeshaspeaks, says innovation has shifted away from the interface of smartphones.

“When you look at astrological counselling, voice is a convenient tool… The early results are good. I am already running out of my free trial space on the cloud,” says Pandit, adding that he leapfrogged over chatbots to voice assistants. A lot of demand on voice comes for daily forecasts, he adds.

While Pandit sees limitations of a voice assistant – serious counselling requires complex permutations and psychological understanding, he insists – he sees promise in the underlying AI layer. “AI can help learn emotions of a human, and we will be able to do significant part of forecasting,” he says.

Efficiencies to the fore

Zomato’s Rao feels that 50% of total activity of the platform will shift to voice in the next couple of years. The reason for his confidence? A large set of actions that take time and effort on chat can be done by Zomato users through simple and short voice commands at almost 100% accuracy. “You can re-order, track your order, look for restaurants, with a one short command,” he says. To be sure, “when you put multiple layers of commands it becomes difficult, but it will improve over time,” he predicts.

Tickets aggregator Ixigo has had an unexpected experience, says co-founder and chief technology officer Rajnish Kumar. In the beta stage, its voice assistant, which uses Google’s voice recognition technology and underlying architecture, can perform 60% of tasks that involved a human in customer support. “For the remaining 40% we still need to go back to human being,” says he.

Ixigo had started work on its voice assistant, called Tara, some two-and-a-half years ago. Things were different then. Voice recognition systems weren’t as good as they are now. Google and Amazon were still figuring out how to crack voice. For every new voice search, technologists at Ixigo had to do manual coding for an answer.

As Google put its weight behind voice, Ixigo decided to double down on Tara. Today, Ixigo is training its voice bots with data the company has accumulated over years through call centre support, emails, and chat.

Ixigo’s Google Assistant-based voice bot, Tara, can perform 60% of tasks humans did in customer support, says Rajnish Kumar, co-founder and CTO

The accuracy of the Ixigo voice assistant is 90% for flights and customer support. “Every time a human intervention is needed or the assistant fails, it raises an alarm,” says Kumar. That prompts Ixigo to further improve the assistant, feed it with better answers and evolve. The target, says Kumar, is to have the assistant do sentiment analysis: anger, sadness, happiness, and humour.

All this will allow Ixigo to significantly cut down cost. “No one wants to have a call centre with 100-200 people,” Kumar says.
A reason why voice is evolving faster than chat is better natural language understanding and conversion technology, which never happened in chatbots. Voice assistants have started understanding a mix of English and the local dialect.

Still, the hurdle with voice, says C Mohan Ram, managing director of Lattice Bridge, which has been working in speech recognition for some 15 years, is tonality and language. “In a country like India, the tonality, dialect and language change every few kilometers,” says Ram, adding the way out is by deploying AI.

Companies like Google and Amazon are trying to solve that problem with natural language processing where a computer program understands what human beings are saying. This requires the machine to understand the dialect, associate words colloquially used with what is meant, consider the context of the question, and respond back in an appropriate manner.

Power of AI on handsfree

The AI edge helps make compelling economics given customer support is the biggest area of deployment of voice assistants. “Our clients tell us that customer satisfaction is much higher on a voice assistant compared to chatbots,” says Preeti Anand, director and head of practice at consultancy firm Zinnov.

Other areas are sales training, education and entertainment – none of which is possible through chatbots. “Chat is a static engagement model, which can’t sense emotions. You can understand the tone and emotions and respond accordingly with voice assistants,” adds Anand.

Also, Zinnov’s data shows that building voice assistants are getting cheaper with APIs like those offered by Google Assistant. (APIs are short for application programming interfaces or code that allows apps or services work with an operating system, service or app.) Voice assistants can be built using APIs in a matter of few days at a reduced cost of up to 40%, Anand says.

In the US, 61% of respondents in a survey by venture capital firm Kleiner Perkins Caufield & Byers said that they use virtual assistants when their hands and eyes are busy like when driving. Another 30% said voice assistants produce results faster.

HDFC Bank, India’s second-largest lender, realises this. Going through a website or waiting on an IVR (interactive voice response) calls to go through the menu is tedious. “You can easily ask the voice assistant a query related to anything that is in the public domain on HDFC bank… especially things that are there on the website or in brochures,” says Nitin Chugh, who heads digital banking at the bank.

The bank is working on both Alexa and Google Assistant to build voice assistants. HDFC Bank’s first design on Alexa was ready in October and it started working with Google after the internet giant decided to open up Assistant to developers for third-party applications.

HDFC Bank looks at voice assistants as medium with unlimited options. Once the voice assistant is ready, it can be implemented anywhere and access will a lot easier. Right now the voice assistant is only available on a Google Home or an Alexa device. “The same thing can be simulated on the dashboard of a car, Echo at home, wherever you want to have a voice interface… Soon (as more people use it), the assistant will start asking what do you want to do today and perform the tasks,” says Chugh.

Bharat, the next stop

Many like Ixigo’s Kumar and FreshMenu’s Daga believe the real power of voice assistants will play out in hinterland India. “It’s a novelty factor right now… two years later we will be able to target a new customer base, which is not savvy with text and tech,” says Daga.

When phone services company Reliance Jio launched its voice assistant ‘Hello Jio’ on the JioPhone, it came as a surprise to many. “But why should anyone be surprised… voice and video are really big in India,” asks a Jio official, who requested to stay anonymous.

Add to that the next billion internet users, most of them will not be versed with typing on the small keypad or in English. “Voice assistants will understand the local language,” says the Jio official said.

As payment solutions mature, voice assistants will fulfil customer needs in many different situations, says Rashmi Daga, FreshMenu’s founder. | Photo: Rajesh Subramanian

Kumar says that in his conversations with Google, the search giant has said that voice searches are as high as 80% in some villages. That itself is a huge opportunity that is untapped by both voice assistant and consumer companies.

Amazon’s Dilip jokes that there is one thing that could still beat voice assistants. “The only thing that is better than voice is imagination… just imagine and the work gets done,” he laughs. “But, that’s not happening soon.”

(A version of this story was published in Mint newspaper on January 15.)

Visuals: Rajesh Subramanian


Updated at 10:42 am on January 15, 2018 to correct a typo in a table.

Disclosure: FactorDaily is owned by SourceCode Media, which counts Accel Partners, Blume Ventures and Vijay Shekhar Sharma among its investors. Accel Partners is an early investor in Flipkart. Vijay Shekhar Sharma is the founder of Paytm. None of FactorDaily’s investors have any influence on its reporting about India’s technology and startup ecosystem.