At its I/O 2026 conference, Google announced a suite of AI agents designed to run continuously in the background, perform tasks across its own services and external platforms, and integrate directly with tools like Gmail, Drive, and Search. The announcements, made in May 2026, represent Google’s most direct push yet into the AI agent space.
The centerpiece is Gemini Spark, a cloud-based AI agent that can operate 24/7 without requiring a laptop to stay open and syncs across the web, Android, and iOS. Google says Spark will support more than 30 external partners, including Dropbox, Uber, and Spotify. It is rolling out to trusted testers this week, with a US beta arriving next week on Google’s Ultra plan. Google’s Gemini app lead Josh Woodward said he has been using Spark to coordinate a neighborhood block party — tracking RSVPs, sending reminders, and checking homeowners’ association rules.
Google also announced AI-powered “information agents” for Search, intended to perform continuous background research on topics like stock market shifts or weather patterns. A new standalone desktop app for Antigravity, Google’s agentic development platform, will serve as a central hub for building and managing autonomous agents. A morning briefing feature called Daily Brief was also introduced. Underpinning all of this is Gemini 3.5, a new model series whose first entry, Gemini 3.5 Flash, is expected next month. Google claims it is four times faster than other frontier models and less than half the price of comparable options.
The moves come in direct response to OpenClaw, the open-source AI agent platform that gained millions of users after launching last November. OpenAI acquired OpenClaw in February 2026, but Google’s existing ecosystem of services — used by more than 900 million people monthly across 230 countries — may give it a structural advantage in deploying agents at scale.
“Before this, I think AI agents were more of an idea in research,” Koray Kavukcuoglu, CTO of Google DeepMind and Google’s chief AI architect, told The Verge. This year, he said, they’ll be “really in our lives.”
Whether Google’s agents perform reliably in real-world conditions remains to be seen. Earlier Google agent experiments were slow and browser-dependent, and last year’s Gemini 3-based agents succeeded at some tasks while failing at others. If Gemini Spark delivers on its stated capabilities, it could mark a significant step forward for consumer AI agents — and for Google’s position in an increasingly competitive field.
Source: The Verge