The first day in a new city and you’re there to attend a party. After a styling session that you booked earlier using the chatbot for the best salon in town, you walk into a clothing retail store to get yourself some new threads for a party. There’s no-one there except you and a couple of other customers. The moment you enter, you are identified, you’ve already specified what you need to one of the innumerable chat kiosks and a convenient in-store navigation screen pops up on your smartphone to guide you to the party wear section. The mechanised gallery highlights the ones that will suit you the best, according to size and past preferences, and once you make a choice, it suggests a couple of other outfits and accessories for you to try on as well. Once you’ve decided, you walk out wearing your new outfit. You have a self-driven cab ready and waiting, and on the way, your digital assistant suggests that you get some flowers or wine for the occasion. You agree and ask for the flowers and wine to be delivered to you at the venue itself. They’re waiting for you when your cab pulls up to the address.
This isn’t sci-fi or the future; the technology to do this already exists. You could literally do half of this right now if you walked into the Amazon Go store in Seattle, Washington (which is not open to the public yet), and get a ride on one of the many driverless vehicles being tested. AI and online shopping are already inseparable. When was the last time when you saw a completely suggestion-free shopping experience online? According to Gartner, by 2020, 85 percent of customer interactions will be managed without humans. And it’s predicted that by 2018, AI technology will advance to the point where service bots can recognise the voices and faces of customers.
Chatting with bots
While AI and robotics still have a long way to go, chatbots are pretty advanced right now. Talking to a bot is not new, but the improved chat AI has made it a customer service default. It’s why so many sites have chat options in their service menus. Chat is also the preferred way of interacting with real customer service people as well. Haptik is a platform that offers this but has shifted to a ‘human augmented intelligence’ model recently.
One of the highest rated apps in India, Haptik provides a lot of services via chat
While dedicated services such as Haptik are doing it independently, it is startups like Niki.ai and Talespin who are truly bringing chat to you by implementing it for any service. While Niki.ai also has a mobile app similar to Haptik, it also has a chatbot SDK. Explaining the direction in which the platform is headed, Sachin Jaiswal, Co-Founder, Niki.ai, says “In the very near future, Niki will also ‘learn’ about the user so that the recommendations are much more accurate. For example, if you frequently travel on Fridays and by night buses, just a ‘show me some buses to Mumbai’ will get you responses tailored to your needs.”
Why do chatbots work?
The growing popularity of chatbots might have a couple of local factors behind it. We all seek shopping tips to help us, be it from the local grocery shop owner who suggests the fruit, or the tailor who tells you about trending fashion. Another factor is that Indians especially, are already used to offloading work onto maids, drivers, and other blue-collar job workers. However, chatbots are becoming popular among millennials, who have neither gone to the kind of stores mentioned earlier nor utilised dedicated blue-collar services as much. So what factors influence chatbots?
eBay’s shopping bot ‘Shopbot’ lets you shop while you chat
First, instant messaging platforms have become ubiquitous and people have mastered texting as a platform long ago. Even language isn’t a barrier anymore, with multilingual keyboards and support for multiple languages on virtual assistants becoming quite common. Second, and more importantly, it removes the paradox of choice. We’re so spoiled for choice, that we often can’t choose anything!
Their effectiveness cannot be denied either. There are things that AI can do (that too in a short span of time) that humans will take ages to finish. “AI analytics and non-AI based methods can be compared through a small example of a lifestyle disorder that starts small but can be very dangerous”, explains Aayushi Kishore, CMO and Co-Founder, Boltt Sports Technologies Pvt. Ltd. “Think of any real life scenario void of any AI- enabled services – to monitor and improve your health, you’d be required to consult a sleep specialist, nutritionist, hire a fitness trainer or even take a gym membership. Each of these would work in isolation to improve and treat your health. In order for any of the above-mentioned service providers to deliver results, one would be required to go through the following process: sharing health history, gathering data, analysis, forming a hypothesis, carrying out multiple tests and repeating the process till the hypothesis is proven true, essentially following a hit and trial method. This process is inefficient in more ways than one ranging from the money wasted to geographic constraints and time consumed in the process. The simple inability of a human expert to comprehend the 1100 TB of data that we all produce in a lifetime is best catered to by AI intervention where huge chunks of data can be stored, a real-time analysis is carried out and concrete insight is shared with the end consumer“.
What AI cannot do
If you were waiting for the section with the cons, congrats, you finally got there. However, if you’re looking for a shoulder to cry on while you say AI will take our jobs, look elsewhere.
In their current state, AI chatbots are not truly creative. “AI still cannot perform any task that requires creativity, or coming up with new ideas that are not resolved using a set pattern”, says Jaiswal. “For example, using the previously stated bus booking case, if the user was a 60-year old man making the booking for himself and his wife, a human being may not recommend a night bus to them due to various concerns. An AI bot, on the other hand, cannot make this differentiation unless specifically told to do so by the creators.”
There is another issue with using chatbots in India – the language diversity. A majority of the Indian retail market operates in vernacular language. As Mr. Anand Natarajan, Head of Strategy and Business Execution, Fullerton India Credit Company Limited says, “The biggest challenge using chatbots in India, with multiple conversational languages, is the ability to decipher bilingual words which are not always captured by artificial intelligence. However most algorithms are self-learning and adapt over a period of time with increased conversations.”
The future of shopping with AI
Developments in deep learning, AI, and natural language processing are showing a lot of promise when it comes to enhancing the average shopping experience. Thinking of chatbots as the only possible AI implementation when it comes to shopping is sheer short-sightedness. There are quite a few new applications that are gradually gaining popularity.
If properly implemented post the currently ongoing beta, Pinterest lens can change shopping as we know it
Visual search is one such area. Machine learning has taken a step forward by being able to identify objects and patterns within images. Take the example of Pinterest Lens – a technology based on the previously implemented visual search, related pins, and guided search on Pinterest – which uses your smartphone camera to identify real world objects in your images and serve up similar images from websites, making it easier for you to find and purchase the object online.
Beyond direct retail, AI will also affect purchase decisions indirectly with deep learning. As Gaurav Gandhi, COO, Viacom18 Digital Ventures explains about Voot, an OTT content platform, “We now better understand levels of ad fatigue, ideal places to insert advertising between content, making user/audience groups for targeting , better targeted marketing plans for our shows, levels of cross promotion and cross usage between genres, best time to publish new shows, how many notifications to send out and at what time – among others – all thanks to data insights.”
Overall, there is a lot of interesting work being done in this area in India. Take Fullerton India’s chatbot Asha, for instance. “We have seen several startups and established companies introduce chatbot services.”, says Anand, “The more successful ones have been those who have taken the approach of being available on already popular chat platforms like FB messenger etc. and have taken up a persona of a friend / assistant. Fullerton India’s chatbot ASHA is available on FB messenger and does not require the customer to get into a new app or adopt any new technology. This is a robotic salesperson who makes itself available to the customer on a platform which he is familiar and comfortable with.” This approach ensures availability along with approachability.
AI based chatbots are a work in progress
As with other aspects of AI, chatbots aren’t perfect yet. Recent debacles like that of Tay prove how machine learning could easily go wrong if not monitored. On the other hand, even with proven results, traditional stores aren’t really eager to jump onto the AI bandwagon. As Tanay Dixit, Co-Founder and Product Head, Talespin puts it, “While investors are generally excited about AI, deep learning, and chatbots, many remain cautious and the same can even be said about customers. Having said that, the lack of access to reams of data to build solutions for particular use cases is a challenge. In the search of data, companies may move into uncharted territories leading to privacy and infringement issues.”
That being said, there is no doubt that in the coming years, you’ll most likely be shopping, and living, aided by bots – on your phone and in your stores. Make sure to keep those credit cards handy, and those bank accounts filled – because the last time we checked, bots don’t maintain tabs.
This article was first published in March 2017 issue of Digit magazine. To read Digit’s articles first, subscribe here or download the Digit app for Android and iOS. You can also buy Digit’s previous issues here.