Google Launches Gemini Spark as Always-On Personal AI Agent

Google has unveiled Gemini Spark, an always-on personal AI agent that represents the company's most ambitious attempt yet to transform its AI assistant from a reactive question-answerer into a proactive digital partner. Announced at Google I/O 2026 on May 19 and rolling out to US-based Google AI Ultra subscribers starting May 29, Spark marks a fundamental shift in how Google envisions the AI-human relationship.

Unlike traditional AI assistants that respond to explicit prompts, Gemini Spark operates continuously in the background, monitoring your connected apps and taking actions on your behalf. The system can reason across Gmail, Google Docs, Slides, Calendar, and a growing list of third-party services, executing multi-step workflows without requiring constant user input.

"Spark represents a big shift for Gemini, transforming it from an assistant that can answer your questions into an active partner that does real work on your behalf and under your direction," wrote Sissie Hsiao, Vice President and General Manager of Google Assistant and Gemini, in the official announcement.

How It Works

Gemini Spark is built on Google's Gemini base models combined with what the company calls an "agentic harness" from Google Antigravity, its internal agent development framework. The system runs as a cloud-based agent, meaning it continues working even when you close your laptop or lock your phone.

Users set up Spark by teaching it specific tasks through natural language instructions. Examples highlighted by Google include: creating weekly lists of critical deadlines from Gmail threads, summarizing ongoing email conversations, spotting hidden fees in credit card bills on a monthly schedule, and executing multi-step workflows that span several applications.

"You can program it to complete several interconnected tasks for complete workflows," the company explained. A demonstration showed Spark monitoring incoming emails, extracting action items, adding them to a task list, and scheduling follow-up reminders -- all without user intervention.

Third-Party Integration

At launch, Spark integrates with Google Workspace applications and three external partners: Canva, OpenTable, and Instacart. Google indicated that additional partner integrations will follow in the coming weeks, suggesting a platform approach that could eventually rival the breadth of Apple's Shortcuts ecosystem.

The third-party connections enable compound actions -- for example, having Spark identify a dinner invitation in your email, check your calendar for availability, and make a restaurant reservation through OpenTable, all in one automated flow.

Privacy and Control

Google emphasized that Spark operates strictly under user direction, with explicit permissions required for each type of action it can take. Users maintain a dashboard showing all actions Spark has performed, with the ability to pause, modify, or revoke any automated workflow.

However, privacy advocates have raised concerns about an AI system that continuously monitors email, documents, and calendar data. The Electronic Frontier Foundation noted that always-on monitoring creates a comprehensive record of a user's digital life that could be subpoenaed or breached.

The Competitive Context

Gemini Spark enters a rapidly forming market for agentic personal AI. Anthropic's Claude Managed Agents, OpenAI's Codex autonomous mode, and Apple's rumored Siri overhaul for iOS 27 all target similar use cases. Google's advantage lies in its deep integration with Workspace, which hundreds of millions of people use daily.

What to Watch

The initial rollout to AI Ultra subscribers is essentially a high-stakes beta test. Success will be measured not just by technical capability but by whether users trust an always-on AI agent enough to hand it meaningful responsibilities. Watch for the expansion beyond US subscribers and the pace of third-party integrations, which will determine whether Spark becomes a platform or remains a feature.