If you’ve been paying attention to the startup world lately, you know that “AI wrapper” has become basically the most insulting thing an investor can call your company. And honestly? The criticism is fair. Slapping a chatbot on top of existing software doesn’t make you innovative, it just makes you lazy.

So when Google and venture firm Accel reviewed over 4,000 applications for their joint Atoms accelerator program focused on Indian startups, it’s not surprising that roughly 70% of them were exactly that kind of wrapper. What’s actually wild is that none of those made the cut for the latest cohort. Out of thousands of pitches, only five startups earned spots in the program.

According to Accel partner Prayank Swaroop, the rejected wrapper startups “were not reimagining new workflows using AI”. They were just adding AI features to existing products without actually thinking about how to do things differently. Beyond the wrappers, a ton of applications fell into oversaturated categories like marketing automation and AI recruitment tools, areas where literally everyone and their cousin is trying to build a startup.

What’s telling is that this year’s program got nearly four times more applications than previous cohorts, with a lot of first-time founders throwing their ideas at the wall. The breakdown of submissions shows why investors are frustrated: about 62% focused on productivity tools and another 13% on software development. That means roughly three-quarters of applications were just enterprise software ideas. Swaroop had hoped to see more submissions focused on healthcare and education, but the Indian startup ecosystem is still pretty enterprise-heavy.

The five startups that actually made it through the selection process are doing something different. K-Dense is building an AI “co-scientist” for research in life sciences and chemistry. Dodge.ai is developing autonomous agents for enterprise ERP systems. Persistence Labs is tackling voice AI for call centers. Zingroll is creating a platform for AI-generated films and shows. And Level Plane is applying AI to industrial automation in automotive and aerospace manufacturing.

These startups will each receive up to $2 million in funding from Accel and Google’s AI Futures Fund, plus up to $350,000 in cloud and AI compute credits from Google. Jonathan Silber, co-founder and director of Google’s AI Futures Fund, explained that these companies aligned with areas where Google expects AI to see deeper real-world adoption.

Interestingly, Google isn’t forcing these startups to use only Google’s AI models. The goal is to see how different models perform in actual applications, then use that feedback to improve future models. As Silber put it, if a startup uses a competitor’s model, that’s essentially Google’s signal to build something better. That kind of competitive drive is exactly what the AI startup space needs right now.