Google has unveiled Gemini Spark, a 24/7 cloud-based AI assistant built to work continuously in the background – handling tasks, managing workflows and operating independently even when devices are switched off. The announcement came at Google I/O 2026 as the company doubled down on its vision of an “agentic” AI future.
Unlike conventional assistants that respond only when prompted, Spark is designed to receive a task once and keep working without repeated instructions. Powered by Gemini 3.5 and Google’s Antigravity technology, the assistant is built to support complex, long-running activities – from monitoring inboxes to compiling daily summaries.
Deep integration with Google Workspace sits at the core of Spark’s value proposition. The assistant connects with Gmail, Docs and Slides, enabling it to gather information across tools, organise it into structured documents and even draft related emails to kick off projects. Practical use cases include analysing monthly credit card statements, tracking school communications and flagging important dates.
Google also announced new MCP connections with third-party platforms including Canva, OpenTable and Instacart, signalling the assistant’s ambitions beyond the Workspace ecosystem.
On privacy and user control, Google confirmed that Spark will seek explicit permission before executing high-stakes actions such as sending emails or completing purchases.
Rollout begins this week with trusted testers, followed by a beta launch for Google AI Ultra subscribers in the United States next week. Broader capability expansions are expected in the months ahead.
