Google, Accel India accelerator picks 5 non-AI wrapper startups
▼ Summary
– Many AI startup applications were “wrappers” that added AI features to existing software without creating new workflows, and none were selected for the accelerator.
– The Google and Accel accelerator program selected five early-stage Indian AI startups, offering them funding and cloud credits.
– Most rejected applications were either wrappers or fell into crowded, unnovel categories like marketing automation and recruitment tools.
– The majority of applications focused on enterprise software, particularly productivity tools and software development, rather than consumer products or sectors like healthcare.
– The program aims to gather startup feedback on AI model performance to improve Google’s models, even if startups use competing models.
The recent selection process for a prominent AI accelerator in India reveals a clear shift in investor priorities, moving beyond simple applications to seek out truly transformative technology. When Google and venture capital firm Accel reviewed over four thousand applications for their joint program, a significant majority were dismissed as superficial “wrappers.” These are startups that merely add an AI chatbot or similar feature to existing software without fundamentally redesigning how work gets done. According to Accel partner Prayank Swaroop, roughly 70% of rejected proposals fell into this category, highlighting a growing wariness toward ideas that could be easily rendered obsolete as core AI models become more capable.
The Atoms accelerator program, announced last November, is designed to support early-stage Indian startups developing substantive AI products. The five chosen companies will receive up to $2 million in funding from Accel and Google’s AI Futures Fund, alongside substantial cloud and AI computing credits. The selection underscores a demand for genuine innovation. Beyond the wrapper ideas, many other applications were turned down because they crowded into well-trodden sectors like marketing automation and recruitment tools, where investors perceived little novelty or differentiation.
This year’s applicant pool was nearly four times larger than previous cohorts, attracting many first-time founders. The proposals largely mirrored India’s current AI landscape, which remains heavily enterprise-focused. Data shows about 62% of submissions were for productivity tools, with another 13% targeting software development, meaning a dominant three-quarters of all ideas were for business software rather than consumer applications. Swaroop noted he had hoped to see more ventures tackling sectors like healthcare and education.
The chosen startups aligned closely with areas where Google anticipates deeper, real-world AI adoption, according to Jonathan Silber, co-founder and director of Google’s AI Futures Fund. He emphasized that the program does not mandate the exclusive use of Google’s AI models. Many companies, he noted, intelligently combine multiple models depending on the specific task. A key objective for Google is to gather direct feedback from these startups on how its models perform in practical applications.
This feedback creates a valuable cycle, or “flywheel,” where insights from real-world startup experimentation are fed back to Google’s DeepMind teams to guide future model improvements. Silber pointed out that if a company opts for a competing model, it simply signals where Google needs to focus its efforts to build the superior product. The selected cohort for this year represents this philosophy in action, focusing on startups that are building foundational AI solutions rather than temporary layers on top of existing technology.
(Source: TechCrunch)
