Google I/O agentic AI push

Google spent its May 19-20 I/O conference describing a shift from AI that answers questions to AI that completes tasks. Across Search, Workspace, Gemini and developer tooling, the company presented “agentic” systems as the next layer of its product strategy, with Gemini 3.5 Flash as the model meant to power them. (blog.google) Google said the new model combines higher-end reasoning with lower latency, while outside coverage tied the push to changes in how people search and work online. (blog.google) ### What did Google actually put at the center of I/O? Google’s official I/O roundup said the company announced new models, agents and tools across search, creation and software development, including Gemini 3.5 and Google Antigravity. (blog.google) In Sundar Pichai’s keynote post, Google said Search will use Gemini 3.5 Flash and Antigravity to build more customized experiences for individual questions, including dynamic layouts and interactive visuals. CNBC reported that Google paired those product updates with new “agentic” tools aimed at consumers and developers as it competes with OpenAI and Anthropic. CNN’s account of the broader search overhaul said Google was giving the search bar its biggest update in 25 years, reflecting a move toward longer, more conversational and multimodal queries. (blog.google) ### Why is Gemini 3.5 Flash so important in this rollout? Google’s developer highlights post called Gemini 3.5 Flash the engine for “real-world agentic workflows.” The company said the model outperforms Gemini 3.1 Pro across almost all benchmarks while running four times faster than other frontier models. (blog.google) Google’s Gemini 3.5 product post said the model is designed for “long-horizon agentic tasks” and that Gemini Spark, described there as a personal AI agent, runs on 3.5 Flash. That makes Flash less a chatbot upgrade than the default layer underneath products that are supposed to plan, retrieve and act. (cnbc.com) ### Where do Spark and Antigravity fit? Google’s I/O collection page described Antigravity as part of its new developer stack, while the developer highlights post called it an “agent-first development platform.” Google said the platform is intended to move developers beyond tools that only help write code toward agents that can take actions in workflows. (blog.google) Google also presented Spark as a consumer-facing agent. In the Gemini 3.5 post, the company said Spark is a “personal AI agent,” tying it directly to the same model family Google is pushing into Search and enterprise products. That links the company’s consumer and developer story: one model, multiple surfaces, more delegated work. (blog.google) ### How does this change the engineering work around AI products? Google Cloud’s I/O post said the company was “doubling down” on what it called the “Agentic Enterprise,” with Gemini Enterprise, an Agent Platform and Workspace integrations. That framing puts the operational questions around AI systems closer to the center of product design. (blog.google) Outside reporting points in the same direction. CNN tied Google’s search changes to a broader shift in how users find information online, while Google’s own materials emphasize action-taking systems rather than one-shot answers. Taken together, that means engineering teams will have to define permissions, fallbacks, verification and monitoring for systems that do more than generate text; that is an inference from Google’s product descriptions and deployment model, not a direct company quote. (blog.google) ### What should readers watch next? Google said many of the I/O announcements are now entering rollout across Search, Gemini, Cloud and developer products. (cloud.google.com) The clearest checkpoints will be product availability for Gemini Spark, wider use of Gemini 3.5 Flash in Google services, and further Antigravity releases on Google’s developer channels. (blog.google) (keyt.com)