Every day we see news about this company or the other acquiring an AI startup, or tacking on ‘AI-driven’ (along with a .ai URL) to describe its model. Media coverage similarly reinforces the narrative of AI as either the solution to all the world’s problems or signalling an imminent techno-apocalypse. We are still far from nearly-human machines that can pass the Turing test (watch Ex Machina, if you haven’t already), let alone fall in love (in the case of Her). It’s only when you look beyond the hype that you start to see the very real and subtle ways that artificial intelligence is already part of life as we know it. Within our portfolio, companies are finding ways to improve efficiencies and create hyperlocal, personalised products and experiences for consumers, and we’ve only scratched the surface.
Understand ourselves better
Indians are producing huge amounts of data as we consume content and interact online, and a technology called Natural Language Processing (NLP) is helping brands figure out how to make sense of all that information and by extension, consumer behaviour. Through lexical analysis, HaikuJam sifts through thousands of user-generated haiku, looking at repeated words and ideas. Essentially, the lengthy process of conducting focus groups, surveys and sentiment analysis is condensed into a span of a couple of hours or minutes. Some of the insights have unearthed connections between completely disparate ideas. While talking about holidays, HaikuJam users mentioned Sleep 1.5x more often than Exploration, contradicting assumptions about wanderlust and jetsetting lifestyles. It’s a real-time glimpse into what people are actually feeling, without having to ask them.
Swiggy has integrated AI into many layers of the business to enhance the entire experience from discovery, ordering food, payment to delivery. By leveraging the terabytes of data generated from over 40 billion user interactions as of January 2019, it analyses past orders, preferences, price points and more to connect customers with restaurants they’ll love.
Artificial Intelligence is only as good as the quantity and quality of data you’re able to feed it. Not only does a system need a lot of data to be able to ‘learn’, it also needs to be properly structured and codified before it’s useable. This was a herculean task for Embibe, that built its business around understanding its users?—?students preparing for entrance exams?—?extremely well so that it could personalise questions and improve test scores. The company adopted a freemium model as a way to get more people onto the platform and gather a critical mass of data. By doing this, they could train AI to map data across a variety of concepts. That was key to creating a system that could automatically absorb content, tag it and then create customised question papers that helped people improve on the areas where they were weakest. A lecture hall could never talk to each person individually and say ‘this is what you need’. But that’s exactly what a system based on AI can do. We’re only beginning to recognise the potential of customised, on-demand products.
In fashion retail, companies like Stylumia use Computer Vision (a technology that analyses and extracts information from images) to forecast trends, create customised recommendations and even AI-powered stylists. What happens when you take this offline? Current retail trends point towards reinventing brick-and-mortar stores into experience and education destinations. Brands are already employing AI to track footfalls and demographics and measure in-store conversions. This kind of data could be used to recognise returning customers and create personalised in-store experiences?—?think of a knowledgeable concierge who greets you by name as you enter and curates products for you based on your interests and purchase history.
Cut out the middleman, literally
As a solution to relying on gut feeling and past experience, Google Maps APIs helped Rebel Foods bring estimated delivery times down to a science as well as identify underserved markets and pinpoint marketing spend. Data also guides how they identify product-market fit while launching new restaurant brands in new localities. Taking humans out of some of the equations removes subjectivity and improves operational efficiency.
The typical loan application process for used cars is tedious and heavily biased towards new vehicles. With its existing ecosystem of tools that track a vehicle’s history, ownership and valuation, Droom applied data science to its latest product Droom Credit, a marketplace for auto loans. In stark contrast to the existing, archaic process, right from signing up for a loan application, until you start paying your monthly instalments, everything is fully automated and data-driven. By taking out most of the paperwork and extremely high rejection rate, Droom implemented an objective decision-making process that was previously governed by the lenders’ whims. In analysing the data and repayment behaviour, it’s also become apparent that these borrowers are reliable, credit-worthy and aspirational. As a result, creating a new class of owners for whom financing their first car used to be prohibitive.
Discover new markets and segments
Google estimates that we will have 532 million Indian language internet users by 2021. A good portion of them will be coming online for the first time, and need some handholding as they navigate their way around the web. While an English internet creates friction, we don’t see companies translating their sites, apps and communications into the 6,500 spoken languages recognised today.
Source: Economic Times
Conversational AI?—?intelligent chat and voice bots?—?could play a major role in ‘onboarding’ the next billion users, which is why companies like Haptik as well as younger startups like Liv.ai and Slang Labs are focusing their energies on supporting regional languages. More than mastering vocabulary, the proof will be in how well they are able to train software to understand the cultural nuances of languages and solve problems beyond simply answering FAQs. If they can crack this, it could spell a whole new way of experiencing the internet.
We’re reaching a point in India where AI is starting to become useful and its impact measurable. Beyond made on the internet brands, we’re starting to see non-internet companies also explore what Rohit Pandharkar, Head of Data Science at Mahindra Group calls the “market of one” route to achieving profits as opposed to the time-tested ‘economies of scale’ model.
Instead of gadgets and gizmos, the conversation is shifting towards how it can transform healthcare, agriculture, education, mobility and the future of our cities. NITI Aayog envisions the country as an ‘AI Garage for 40% of the world’, where ‘Made for India’ solutions will trickle outwards to other places with similar complexities and challenges, such as countries in Africa, Middle East and Southeast Asia. There is going to come a time when AI will be business-as-usual, when it will no longer be a differentiator, but what’s critical is that AI also needs the ‘messy’ reality of India before we can get there.