Google pushes agentic AI into products

Google spent its May 19 I/O keynote tying “agentic” AI to products people can use now, not just to research demos or concept videos. Chief Executive Sundar Pichai said the company was in a phase of the AI cycle where users “want to see the value in the products they use every day,” and Google’s announcements centered on Gemini app features, Search tools and developer infrastructure built to take actions on a user’s behalf. EDN, reviewing the conference on May 21, described the difference from prior years in one word: “Maturity.” Google’s own developer posts made the same case more concretely, announcing Managed Agents in the Gemini API, an updated Antigravity platform and Gemini 3.5 Flash as the model meant to power “real-world agentic workflows.” ### What did Google actually ship at I/O? Google’s May 20 roundup listed Gemini Omni, Gemini 3.5 Flash, Google Antigravity and “Universal Cart” among more than 100 announcements from I/O 2026. (blog.google) In product terms, the company pushed agent features into the Gemini app and developer stack rather than presenting agents as a separate lab project. The Gemini app now reaches more than 900 million monthly users across 230 countries and more than 70 languages, Google said in a May 19 product post. (edn.com) That same post said the app was becoming “more agentic,” with proactive and around-the-clock help, while Google’s developer materials said Managed Agents would be available through the Gemini API and Google AI Studio. (blog.google) ### Why does “agentic” matter in Google’s version? Sundar Pichai said on May 19 that Google wanted to help users “get more done with Gemini,” language that framed the company’s pitch around task completion rather than chatbot conversation. Google’s developer post was more explicit, saying it was “accelerating the shift from prompts to action” and describing dynamic subagents, scheduled tasks and integrations across Android, Firebase and Google Cloud. (blog.google) Google also tied that pitch to speed and deployment. The company said Gemini 3.5 Flash outperformed Gemini 3.1 Pro on most benchmarks while running four times faster than other frontier models, which Google said made it suitable for production agent workflows. ### How does this fit Google’s broader AI strategy? Axios reported on May 21 that Google is trying to expand aggressively in AI without undermining the search and advertising businesses that generate most of its profits. (blog.google) That tension was visible at I/O, where Google folded new AI behavior into existing consumer surfaces instead of replacing them outright. Axios also reported on April 24 that Google would invest $10 billion in Anthropic, with the potential for another $30 billion, a commitment that would bring the total to about $40 billion. (blog.google) That gives Google both an internal model-and-product push and a major outside stake in a leading AI lab. ### Where are the regulatory and market-structure questions? Regulatory pressure is building around the companies supplying chips, models and distribution. (axios.com) The source briefing for this story cited a report that regulators in the United States, European Union, China and France were examining Nvidia-related competition concerns, even as AI companies deepen infrastructure and partnership ties. OpenAI is facing a parallel legal and governance squeeze. Reuters reported on May 21 that the company was expanding its roster of law firms to handle major lawsuits and deals, while Crypto Briefing reported that Republican attorneys general and the House Oversight Committee were scrutinizing governance and Chief Executive Sam Altman’s potential conflicts as an IPO valuation approached $850 billion. (axios.com) ### What comes next after the keynote? Google said the next phase is rollout. Its May 19 and May 20 posts pointed users and developers to the Gemini app updates, Gemini API, Google AI Studio and Antigravity tooling announced at I/O, while the company’s keynote transcript framed the effort as continued “shipping” after a year of rapid releases. The near-term test will be whether those agent features move from demos into routine use inside Gemini, Search and enterprise development tools. (cryptobriefing.com) Google has already named the products and platforms involved: Gemini 3.5 Flash, Managed Agents, Antigravity 2.0 and the updated Gemini app. (blog.google 1) (blog.google 2)