Chatbots make a lot of noise but remain on the sidelines

Photo credit: Валерий Качаев.

We Indians love to talk. And the startup ecosystem is taking note. No wonder the chatbot is the latest buzzword in technology.

Chatbots are machine agents hidden in your messenger or app that can conjure up information from other programs using simple keywords. These agents act like humans, and you can use one to order food, pick up your laundry, pay your bills, or book tickets just by sending a text message.

The vision for a chatbot: get machines to respond to questions like a human being. Have chatbots achieved that vision? Not so much. The technology is still nascent, so a chatbot may have trouble understanding your query or direct you onto a wrong path.

Nevertheless, major tech companies worldwide are backing chatbots. Facebook launched bots for Messenger in April and WhatsApp will soon join the bandwagon. Amazon’s Alexa and Apple’s Siri are being developed further to add additional features. Cisco, Oracle, SAP, and Samsung are also moving in this direction.

In India, the early adopters are startups. As of April, over 30,000 chatbots were launched on Facebook Messenger’s chatbot platform alone.

Chatbots have been around for a fairly long time. The first ever coded chatterbot – Eliza – was invented as far back as in 1966. Eliza could imitate the language of a therapist.

Why are they suddenly popular? For starters, the technology behind chatbots – machine learning and natural language processing – has evolved to make chatbots friendlier and more intuitive.

Eliza’s program coding was too basic to go beyond a short conversation. As the legacy of technology has advanced, the romance of talking to a machine has compounded too. Today you can call on Siri without even touching your phone, and she will set reminders for you, call your mom for you, or place an order for you online.

Second, the cost of developing a chatbot is one-third of what is required in developing a mobile app. Industry sources estimate that building a chatbot from scratch would cost about US$5,000. But if you are using a readily available chatbot building platform like Facebook’s Messenger, the cost will drop to between US$250 and US$1,000. Mobile apps, on the other hand, can cost anywhere between US$100,000 and US$1 million.

Then, a chatbot uses very low bandwidth. So a user doesn’t have to own a smartphone – thus allowing startups to widen their reach to users of basic phones.

India is now the second-biggest smartphone market, replacing the US, with over 220 million users, according to a study by Counterpoint Research.

Despite the surge in smartphone usage in India, only about 17 percent of the country owns one, according to a study by the Pew Research Center. This figure plunges to 7 percent in the lower-income category.

But the popularity of chat services in towns and villages is high. According to a report released jointly by the Internet and Mobile Association of India and consultancy firm, IMRB, 37 percent of the rural population uses the internet for communication, and 33 percent for social media. In urban areas, over 70 percent use the internet for chat apps.

The foremost reason chatbots have gained traction is ease of use. “Adjusting to a machine does not come naturally to us. With every app you need to learn how to use it. We wanted to break those shackles and make commerce more interactive. Conversations come naturally to us,” Niki.ai co-founder Sachin Jaiswal says.

The excitement around chatbots has grown in a year. An increasing number of startups such as Jugnoo and Niki.ai have entered the market with an aim to simplify ecommerce.

See: So what if the Internet sucks in India? This AI chatbot gets things done

Humanizing the bots

By the looks of it, chatbots make a lot of business sense.

Facebook’s all-in-one virtual assistant, codenamed MoneyPenny or M, is one of the most awaited chatbot platforms. Facebook’s Messenger platform allows you to build chatbots over it, but with M goes one step further: it integrates all other apps available into one. So if you’re confused about buying a pair of shoes, just ask M instead of messaging different store bots. It’s still in the early testing stage.

“It’s an experimental stage for chatbots. Players are still trying to find ways to make it more interactive and useful,” says Harsh Shah, co-founder of fashion shopping app Fynd. Mumbai-based Shopsense Retail Technology is the startup behind Fynd. The company has launched a fashion shopping chatbot that the company hopes will contribute 25 per cent to its revenue in 2017.

Orahi, a carpooling app, initially developed a chatbot just to take feedback. It then revamped the chatbot to include Bollywood-style conversations.

Like Hike Messenger, Orahi has rotating Bollywood themes. You may have Gabbar ask you “kitne aadmi hain?” in that typical sardonic tone made popular by the iconic movie Sholay of the 80s. Or when you finish your payment, Crime Master Gogo from cult classic Andaz Apna Apna may express his excitement with a “Gogo khush hua” remark.

MindIQ, a bot builder platform, introduced quick selection buttons integrated with text to simplify the process of finding a product through the chatbot.

“With buttons, we were able to do two things. We reduced the amount of words the user had to type to get an answer, and the chatbot was more accurate in its recommendation,” MindIQ co-founder Mounish Kothapally says.

See:They developed Cortana for MS, now they help businesses build AI chatbots in minutes

Niki.ai’s Nikibot helps people in India hail a cab, order food, pay for laundry or the electricity bill, among other services. It’s betting more on the utility factor than entertainment.

“We want to give a wallet experience while making payment through bank accounts by linking it with UPI. Once that happens, the failure rate will be lower as the users don’t have to move out of the bot platform,” Sachin says.

He’s referring to the Unified Payments Interface (UPI), which allows bank customers to send and receive money via a smartphone in real time. The biggest advantage of UPI is it that does not require an account number, bank details, or a bank-unique IFSC for payments. Paytm also adopted the model recently.

Niki.ai was one of the earliest movers in integrating a payments gateway in chatbots. The startup, since its inception in 2015, has seen a tenfold increase in chat-to-order conversion on Messenger, compared to the web. Jugnoo, a hyperlocal autorickshaw hailing app founded in November 2014, has seen about 5 percent of its 35,000-odd daily transactions being completed via chatbots. In a few more months, the startup expects this share to go up to 20 percent. Orahi expects to reduce human-led operations by 10-15 percent in 2017 on the back of its chatbot.

All this optimism is not without reason. Jugnoo is working on improving the customer experience on its chatbot by re-planting all its app features. The users won’t have to download the app to look up order history, or browse history. All of it will be done over the chatbot.

“We would spend months developing an app. Unlike apps, chatbots don’t need much time to develop or upgrade. We can act upon feedback faster and bring it to the users much sooner,” says Saurabh Wadhawan, co-founder of auto-hailing app Jugnoo.

See: Why India is primed for the chatbot revolution

Bots to the rescue? Not so much

The advantages of a machine-led system that can imitate – and thus replace – humans are many. At least in theory.

But the reality: chatbots need handholding by humans to be useful for consumers. Facebook predicts that there will always be a need for humans to jump in if M faces a question too difficult to answer.

“Chatbots are a parallel marketing opportunity. It’s an alternative at the best. It cannot replace human intervention, and it cannot become the sole platform for a company to do business. But it is a lucrative option to have,” Jugnoo’s Saurabh says.

A Niki.ai chatbot. Photo credit: Niki.ai.

Plus, there are some dumb bots that expose the imposter machine.

Take, for example, Microsoft’s Tay. A chatbot programmed to sound like a 14-year-old girl. If you tagged her in a conversation on Twitter or Snapchat, she would respond to what you said. But it seemed to have a loophole. A group of miscreants was able to exploit that loophole and trained the algorithm to dole out racist and inappropriate answers. So Microsoft shut Tay down for a while to “upgrade her.”

MindIQ thought their design was foolproof, but had to suffer certain setbacks. Their aim was simple: to help businesses create chatbots as easily as possible without requiring programming knowledge or AI jargon.

“But just as we thought we were ready, we received our first blow. One of the first users asked the bot this question: “I weigh 120 kg, my wife weighs 80 kg, we’re traveling to hilly terrain, which bike do you recommend?”” Mounish says.

The bot couldn’t answer.

HelpChat, one of the earliest chat-based providers of hyperlocal services, had to shut down its chat-feature in 2015. The startup, today known as Tapzo, was doing more than 70,000 chat sessions a day.

“The only problem with chat was that customers who tried it weren’t coming back. So we tried harder. That still didn’t work. It now seems foolhardy to even think like that but we really thought that chat UI will truly work,” founder Ankur Singla says.

In fact, the chatbot irritated the users to no end. “Coping with Hinglish (Hindi + English), SMS lingo, and vernacular languages to write a general purpose chatbot becomes a five-year science project very quickly,” Singla wrote in one of his blogs.

Does it really make sense to chat to get a cab when you can simply tap once and book? “No matter what you think, our user data was very obvious on this. The answer is no,” Ankur says.

Like software, chatbot language is built on a system of rules that develop and evolve over time; machine learning refines itself with data. But human conversations are so varied that the machine learning algorithms struggle to cope.

“For a chatbot to be mainstream, it needs to understand the cultural connotations of a language. A word like Gujjar can be fed into the computer, but if it doesn’t know it is offensive to some, it might create problems. Also, the personal connect is lost,” Orahi founder Sameer Khanna says.

But Niki.ai has found a way around it. Apart from using machine learning, the startup conducts mock interviews with staff and consumers once in three months to train the bot on new insinuations, words, phrases, and situations.

Other companies are innovating in their own ways. Payjo – a messenger bot that lets you top up your phone and set a reminder for prepaid phones – has introduced the bot in several non-English languages. It is available in Hindi, Tamil, Telugu, and Kannada. MindIQ’s CBPredictor is an algorithm that can predict the next user action and provide recommendations accordingly.

“Through our journey, we have understood that the best way to build bots for businesses is through a hybrid approach – use of buttons and quick replies along with text-based queries. This approach helps both businesses and customers,” Mounish says.

The chatbot arena is in a nascent stage, says Dinesh Tiwari, managing director at investment firm Broad Peak Capital Advisors. “Startups are innovating, and new features being added. The growth in chatbots has been very quick.” But they’re unlikely to become mainstream anytime soon, and a startup with chatbots as its core technology may not be that attractive to investors, he adds.

Our take: the number of players is growing; new features are being added constantly. But the technology is still raw. It’s going to take a lot of fine-tuning before chatbots become a viable option for ecommerce. At best, it can be a good assistive alternative to mobile apps.

This post Chatbots make a lot of noise but remain on the sidelines appeared first on Tech in Asia.

from Tech in Asia https://www.techinasia.com/chatbots-and-startups
via IFTTT