Google has officially launched Gemini Spark in India, a cloud-based AI agent that operates continuously in the background — even when your phone is locked or your laptop is closed. Initially previewed at Google I/O 2026, the rollout begins for Google AI Ultra subscribers, with Google AI Pro subscribers gaining access in the coming weeks.
An Agent That Works While You Sleep
Unlike traditional AI assistants that respond to individual prompts, Gemini Spark is designed to autonomously complete multi-step tasks over time. It runs on Google’s cloud infrastructure and requires no active user involvement once a task is set. Built on the upgraded Gemini 3.6 Flash model, the agent integrates natively with Google Workspace apps including Gmail, Google Docs, Google Sheets, and Calendar — with no manual setup required.
Tasks range from organizing emails in a personalized style and summarizing wedding RSVPs to planning travel itineraries across multiple services and managing recurring subscriptions.
How Tasks, Schedules, and Skills Work
Gemini Spark operates using three core components:
- Task – The outcome the user wants the AI to achieve, from simple inbox clearing to complex multi-step workflows like coordinating a business trip.
- Schedule – Recurring automations, such as a weekly summary of unread emails every Monday morning.
- Skill – Specific capabilities Spark can apply, which will expand over time through updates. The agent can also interact with websites on the user’s behalf via a browser-based environment.
Google has not yet confirmed which web capabilities will be available immediately in India. The launch comes shortly after OpenAI released ChatGPT Work, targeting a similar productivity and enterprise niche.
User Control and High-Stakes Approval
Users maintain full control over what Gemini Spark can access and do. They can limit connected apps or disable the agent entirely. For high-stakes, irreversible actions — such as sending an email without prior review, making a purchase, or any difficult-to-reverse action — Spark is designed to ask for explicit approval before proceeding.
Google describes this human-in-the-loop design as a deliberate safeguard, positioning Spark as an agent that operates under user supervision. The India launch marks the company’s clearest step yet toward agentic AI that handles repetitive digital work autonomously.
