Most artificial intelligence implementation ideas are still little more than “wrappers” built on top of existing models. But as AI model makers add more features, investors are wary of startups that may simply not be needed.
Case in point: when reviewing more than 4,000 applications for a joint AI accelerator for Indian startups run by Google and venture capital firm Accel, “wrapper” ideas dominated. But none of them were among the five startups in the latest batch, Accel partner Prayank Swaroop told TechCrunch (pictured above).
Announced in November, Google and Accel’s AI-focused Atoms program aims to bring back AI-connected product startups to India. Startups selected for the latter group will receive up to $2 million in funding from Accel and Google’s AI Futures Fund, as well as up to $350,000 in cloud and AI computing credits from Google, the firms said.
About 70% of rejected apps were “wrappers” — startups that added AI features like chatbots on top of existing software but weren’t “reimagining new workflows using AI,” Swaroop said.
Most of the remaining rejected applications, Swaroop said, fell into crowded categories like marketing automation and AI recruiting tools, areas where investors see something new. Beginners in those fields often struggle to distinguish themselves, he said.
This, perhaps, is not surprising. This year’s program received almost four times the previous Accel Atoms applications – with many first-time founders.
India’s growing AI ecosystem is still very much focused on business applications and Swaroop said the applications reflect that. About 62% of the applications submitted focused on productivity tools and another 13% focused on software development and coding, meaning three-quarters of the applications were business software ideas rather than consumer products. (Swaroop hoped to see more ideas for health care and education.)
Techcrunch event
San Francisco, CA
|
October 13-15, 2026
Jonathan Silber, founder and director of Google’s AI Futures Fund, said the five selected startups closely align with areas where Google expects AI to see deep real-world adoption.
The system doesn’t need to start using only Google models, Silber said, noting that many companies integrate multiple models depending on the workflow. The goal, he said, is to gather feedback from scratch on how Google’s models work in real-world applications.
Data from those startups could be fed back to Google’s DeepMind teams to help develop future models, creating what Silber described as a “flywheel” between startup testing and AI development. “If a company uses another model, that means Google has work to do to build the best model on the market,” he told TechCrunch.
This year’s selected startups are:
- Dense, which is building an AI “co-scientist” to accelerate research in fields such as life sciences and chemistry;
- Dodge.ai, which develops autonomous agents for enterprise ERP systems;
- Persistence Labs, which focuses on voice AI for call center operations;
- Zingroll, which creates a platform for AI-generated films and shows;
- Level Plane, which applies AI to industrial automation in automotive and aerospace manufacturing.
