Google for Startups has launched a new Startup School focused on agentic AI, a training series aimed at helping founders and developers build AI systems that can act more autonomously and deliver practical business value.
The program, called Startup School: Agentic AI, is positioned as an immersive curriculum for people building live, multimodal products. According to the event page, the series will run from June 9 to June 18, 2026, and will be offered in live sessions with two time options for each class to accommodate participants across regions.
Google says the school is designed to move builders beyond simple chatbot experiences and into production-ready AI agents. The curriculum centers on a single example application, described as an autonomous customer intelligence and productivity platform, which participants will use as a guide for learning workflows that combine voice, vision and data grounding.
The program is organized around three main areas. The first is real-time voice AI using the Gemini Live API. The second is multimodal retrieval-augmented generation, or RAG, which is meant to help agents reason over complex information sources. The third is bidirectional vision agents for faster, human-in-the-loop data extraction.
The opening session is set to cover the shift from text-based chat to real-time agent workflows. Google says attendees will learn how to design prompts for low-latency voice interactions, build interruptible conversations, and move from rapid prototypes to stateful applications using Firebase Auth and Firestore.
A second session is focused on grounding conversational AI in larger datasets. That part of the curriculum will look at how to map text, audio transcripts, PDFs and other media into a shared semantic space, with tools including Gemini Embeddings 2, Gemini Context Caching and Enterprise Vector Search. Google says the goal is to help agents answer more accurately while keeping latency and token usage lower.
The school also highlights the use of Google Search Grounding and hybrid search methods as ways to reduce hallucinations in business-critical settings.
Google says the program is intended for decision makers such as founders and CxOs, as well as product leaders and startup practitioners including developers and data engineers. It is framed as a global training series, with class times set up for both AMER and EMEA participants, or EMEA and APAC participants, depending on the session.
Each class is live, but Google notes that the content is the same across the regional time options. Participants can select whichever time works best for them and add it directly to their calendars. If they want to attend a different session time later, they do not need to unregister from the first one.
The launch reflects Google’s continued effort to support startups building on its AI stack, especially those interested in moving from demos to more production-oriented systems. The sessions reference Google AI Studio, Firebase and several Gemini models, suggesting the company is using the Startup School to showcase how its tools fit together in agentic AI workflows.
The event page also includes on-demand watch options, calendar integrations and standard event-sharing links, indicating that Google is treating the school as both a live educational series and a broader outreach effort for its cloud and AI developer ecosystem.