Google I/O 2026: Gemini Spark Brings Agentic Experiences Across Google Docs, Slides and More Apps

Google announced Gemini Spark, a personal AI agent that can take actions on the user's behalf during the Google I/O 2026 developers conference at the Shoreline Amphitheatre in California on Tuesday. The agentic AI tool is powered by Gemini 3.5 and is built on the Mountain View-based tech giant's Antigravity integrated development environment (IDE). Gemini Spark runs on dedicated virtual machines on Google Cloud and is integrated with Google tools.

Gemini Spark as an AI Agent Workspace

In a blog post, Google said that Gemini Spark is built around planning and structuring tasks for AI agents and collaborative teams, and for repeatable use cases. The platform runs on Gemini 3.5 and uses the Antigravity IDE. It is integrated with Google Workspace apps such as Gmail, Google Docs, and Google Slides. The company explained that since it is cloud-based, it continues to process tasks in the background even after users close their laptops or lock their phones.

The platform can convert simple prompts or rough goals into structured task briefs that include roles, workflows, deliverables, constraints, handoff notes, and verification checkpoints. Google says the outputs can then be used across various AI runtimes, agent systems, and enterprise workflows.

The tech giant demonstrated several workflow examples during the keynote. For example, it can automatically scan monthly credit card statements to identify new subscription fees. Users can also teach the platform custom routines, including monitoring school emails, extracting important deadlines, and generating consolidated daily summaries. 

Lastly, it can automatically analyse meeting notes across emails and chats, create structured Google Docs reports from the findings, and draft follow-up emails for project kickoffs.

Google also detailed the workflow process behind Gemini Spark. Users begin by entering a broad objective or project requirement into the platform. The agentic AI platform then analyses the input and automatically structures the project into smaller tasks, expected outputs, assigned roles, workflow stages, and review checkpoints.

The platform, however, does not directly execute tasks. Instead, it focuses on designing and structuring the workflow logic. Meanwhile, the execution depends upon the AI platform, automation tool, or operational system that users integrate afterwards. Google added that Gemini Spark is integrated with Google's existing ecosystem and will soon support third-party integrations through MCP (Model Context Protocol).