Google, Accel India accelerator selects 5 startups and no ‘AI wrappers’

Most artificial intelligence startup ideas are still little more than “ears” built on top of existing models. But as AI developers add more, investors are wary of startups that may become redundant.

Case in point: in a review of 4,000 joint AI accelerator programs in India run by Google and the Accel firm, the “wrapper” concept came to the fore. But none of them were among the five founders of the latest batch, Accel partner Prayank Swaroop told TechCrunch (pictured above).

It was announced in November, the AI-oriented Atom program developed by Google and Accel aims to restore the initial AI infrastructure connected to India. Startups selected for the latest round will receive $2 million in funding from Accel and Google’s AI Futures Fundand up to $350,000 in cloud and AI computing loans from Google, the companies said.

About 70% of the rejected apps were “overlays” — startups that layered AI features like chatbots on top of existing apps but “didn’t think about new workflows using AI,” Swaroop said.

Most of the remaining projects that were rejected, Swaroop said, fell into crowded categories such as marketing automation and AI recruiting tools, areas that investors saw as unusual. Startups in those fields often struggle to differentiate themselves, he said.

This is not surprising, either. This year’s program received almost four times the number of previous batches of Accel’s Atoms – with more founders.

The natural growth of AI in India is focused on businesses and Swaroop said the projects reflect this. About 62% of the offers focused on product development tools and another 13% on software development and documentation, meaning that almost three-quarters of the applications were business software ideas rather than product purchases. (Swaroop expected to see more medical and educational ideas.)

Jonathan Silber, co-founder and director of Google’s AI Futures Fund, said the five startups selected are closely related to areas where Google hopes AI will see real-world adoption.

The program doesn’t require startups to use only Google brands, Silber said, noting that many companies combine multiple brands depending on how they work. The goal, he said, is to gather feedback from the ground up about how Google’s brands work in the real world.

Ideas from that startup can then be fed back to Google’s DeepMind teams to help refine future models, creating what Silber described as a “flywheel” between initial testing and AI development. “If a company is using a certain brand, then Google has a responsibility to create the best brand in the market,” he told TechCrunch.

This year’s nominees are:

• K-Densewhich is developing an AI “co-scientist” to accelerate research in areas such as life sciences and chemistry;
• Dodge.aiwhich creates independent providers of ERP systems;
• Persistence Labswhich focuses on voice AI for call center services;
• Zingrollwhich is building a platform for AI-powered movies and shows;
• Level Planewhich uses AI to design cars and airplanes.