Google and venture capital firm Accel's "Atoms" AI accelerator program, jointly launched for Indian startups, recently selected winners from over 4000 applications.5 highly promising startupsIt is worth noting that none of these selected companies are simply "AI wrappers." These startups are all dedicated to delving into specific industries and using AI to completely reshape complex workflows.
Investors are prioritizing process reengineering over 70% of "AI shell" proposals.
The Atoms program, announced last November, is primarily supported by Accel and Google's AI Futures Fund to foster early-stage AI startups in India. Selected teams will receive up to $2 million in funding and up to $350,000 worth of Google Cloud and AI computing resources.
However, this investment will not be given away easily. Accel partner Prayank Swaroop said that among more than 4000 applications, about 70% of the rejected proposals were "AI wrappers", which are simply adding AI interface functions such as chatbots to existing software, without "using AI to rethink new workflows".
Furthermore, many of the rejected proposals focused on already crowded areas, such as "marketing automation" and "AI recruitment tools." Investors believe these areas lack innovation, making it difficult for startups to build a competitive advantage and stand out.
Enterprise-level applications are gaining popularity: solving real-world industry pain points.
This year's Atoms accelerator received nearly four times more applications than in previous years, many from first-time startup founders. Data shows that India's AI startup ecosystem remains highly focused on "B2B enterprise applications," with 62% of proposals focusing on productivity tools and 13% on software development and programming.
Overall, about three-quarters of the applications were for enterprise software, rather than products aimed at general consumers. Prayank Swaroop admitted that he had hoped to see more ideas related to healthcare and education.
The five startups that ultimately emerged all precisely targeted areas where Google anticipates deep adoption of AI in the real world:
• K-Dense:BuildAI "Joint Scientists" (co-scientist) will accelerate research in fields such as life sciences and chemistry.
• Dodge.ai:Develop for enterprise ERP systemsAutonomous agents.
• Persistence Labs:Focus on customer service center operationsVoice AI technology.
• Zingroll:Build aA platform specifically designed for AI-generated movies and shows..
• Level Plane:Applying AI to automotive and aerospace manufacturingIndustrial automation.
Google's open-minded approach: not being tied to its own models, thereby creating a "flywheel effect".
Another noteworthy highlight is the high degree of openness Google has demonstrated in this accelerator program.
Jonathan Silber, co-founder and director of the Google AI Futures Fund, stated that the program "does not require" startups to exclusively use Google's AI models. He is well aware that many companies combine multiple models depending on their different workflows.
Google's real goal is to gather feedback from these startups on the performance of Google models in real-world applications. These insights will be fed back to the Google DeepMind team to help improve future models, thereby creating a "flywheel effect" between startup experimentation and AI development. Jonathan Silber points out, "If a company chooses to use an alternative model, it means that Google still has room to work, and we must build the best model on the market."
