There’s a growing pipeline of high-caliber founders – who have “been there, done that” – taking the startup plunge with artificial intelligence (AI) and machine learning (ML) tech in India. On the other hand, startups with shallow applications of AI are quickly finding out that investors and clients alike have become more discerning.
For example, Zoogaad raised US$500,000 back in 2014 to provide personalized news powered by AI. But, unlike Toutiao in China, it did little to push the boundary beyond Google alerts, and had to shut down.
At the other end of the spectrum is Bangalore-based SigTuple, which ticks most of the boxes for the new wave of promising AI startups coming out of India. SigTuple applies AI-powered analytics to visual medical data, such as blood smear slides that go under a microscope. This improves the speed and accuracy of diagnosis.
This startup’s founders had worked together at American Express’s big data lab. One of them is SigTuple’s chief scientist officer Tathagato Rai Dastidar, a computer science PhD who was the director of the Amex lab. So they had the competence and experience to become deep-tech entrepreneurs.
A younger startup in Bangalore, AskSid, has chosen to focus on women’s fashion. Its AI-powered bot helps brands get into meaningful conversations with their customers while helping them shop online. After successful pilots, topline brands like Wolford have deployed the bot in multiple countries. AskSid’s founders come with rich corporate experience from India’s IT industry.
Bangalore-based incubator Excubator counts 2,312 AI and ML startups globally. According to its number-crunching engine Excube360, the count began in 2007 and reached a peak in 2015, when 456 startups were founded to leverage advances in the field. The number dips to 359 in 2016, and falls to 87 in 2017.
The pattern follows a similar trajectory in India, with nearly 186 AI startups being founded in 2015 and 2016, but only 42 last year, according to data from venture capital analytics firm Tracxn. What gives?
Excubator founder and CEO Guhesh Ramanathan offers an explanation. “There is definitely a trend toward support for more mature startups in this space,” he tells Tech in Asia.
In other words, the days of all-and-sundry startups trumpeting AI/ML in their pitch deck to catch a VC’s eye are gone. Investors are looking for differentiation and deeper applications of AI/ML in solving real-world problems in the enterprise as well as broader consumer markets.
VC funding in the Indian AI space rose from US$15 million in 2015 to US$67 million last year despite the steep fall in number of AI startups being founded, shows Tracxn data. The average deal size rose from US$0.8 million in 2015 to around US$1.5 million, as series A and B rounds began to kick in over the last couple of years.
The funding level in India is low compared to the leading AI ecosystems globally. Excube360 data puts the US at the top with over US$15 billion invested in the sector, followed by China with US$1.9 billion.
But the increased progression from the seed stage to the series A and B levels augurs well for the Indian ecosystem. A number of factors are contributing to this. One of the key elements is connecting startups with large organizations and even the government, where incubators like Excubator and accelerators are playing a part.
This assumes added importance for AI startups as success often depends on access to good sources of data. “Yes, data is the new oil, and given the relatively younger companies in India, we feel it will be a struggle in the beginning. But we also feel that given the value-addition that startups are building on raw data, this will quickly reach a level playing field,” says Ramanathan.
US tech giants sitting on mountains of user data – Google, Facebook, Amazon, Apple, and Microsoft – have started offering AI-as-a-service. From natural language processing to image analysis and sentiment analysis, developers everywhere can make API calls to these algorithms to come up with AI applications. But, as AskSid and Sigtuple have shown, the differentiation comes from access to unique data training sets – from large corporate clients in the case of AskSid and hospitals for SigTuple.
Domain expertise is the name of the game. For example, Lucep and Active.ai are helping banks and financial institutions know their customers better. Flutura combines AI with industrial IoT to bring predictive analytics to manufacturing. Edge Networks applies AI analytics to the HR domain in multiple ways, from uncovering the right talent to tracking the engagement level of employees.
You need “domain knowledge to address core problems in a certain vertical. To use the same bot to order a pizza and book a bus ticket – those days are gone,” AskSid co-founder Sanjoy Roy tells Tech in Asia.
Startups like AskSid, Lucep, and Flutura with experienced founders have also made encouraging forays into global market. Lucep partnered with a global bank to enter Mexico, and Flutura’s Cerebra product is being used by an oil and gas major in Houston, a German manufacturer of adhesives, and Hitachi in Japan. One-year-old AskSid already has top fashion brands in Europe as clients.
Where the Indian AI startup ecosystem is yet to see action is exits, even though most of the players gaining traction have a B2B (business-to-business) model. There were only two significant acquisitions of AI startups from India last year: Apus acquired Siftr, which used computer vision to curate images from user-generated content, and Google acquired Halli Labs, which was developing AI-powered speech and vision products.
Excube360 data on 2,312 AI startups around the world shows that 10 percent of them have had exits – 208 acquisitions and 28 IPOs. So there’s a lag in India.
“The number of AI startups that have been acquired globally is significantly higher than acquisitions in any other space, and a massive validation for AI as a technology,” says Ramanathan. “India, we feel, is still just getting into the game. Most of the acquisitions have been in the US and Europe.”
It may be early days yet for exits in India. But this could change soon, as the funding pattern shows a move away from infrastructure to applications. Tracxn data shows a big drop in startups providing cloud-based AI infrastructure, algorithms, and libraries. That’s not surprising because the tech giants have become active players in the game of AI-as-a-service and they’re better equipped for it.
Most of the AI startups being founded and funded last year in India were into AI-enabled applications for multiple use cases in different verticals. Now, if these applications can go deep enough, thanks to top-notch Indian tech talent, we could see a tipping point for AI startups from India this year or the next.
This post India’s AI startups are at a tipping point appeared first on Tech in Asia.