Google has unveiled an AI agent called Gemini Spark, which helps users carry out tasks autonomously, even when they don’t explicitly instruct it to do so. Sparks do not return to the user when prompted, but continue to work even after the user provides prompts and conditions. The feature will be rolled out to users in India in the next few weeks, for Google AI Pro and Google AI Ultra subscribers.
Built and hosted on Google’s cloud infrastructure, Gemini Spark integrates with Google Workspace applications, such as Gmail, Google Docs, Google Sheets, and Google Calendar. It can be used to track incoming data and prepare responses and actions while a user is away from her phone or computer. According to Google, the feature relies on the user’s consent and is designed to utilise the Gemini 3.6 Flash. When taking important actions (such as sending an email or making a purchase), Spark requests permission.
Users need to be subscribed to Google AI Pro or Google AI Ultra before the feature is available in their account to use Gemini Spark. Once access is granted, users can open Gemini, click Gemini Spark, link Google apps, pick the services that the AI can use and establish instructions that explain to the AI what they want to have it perform automatically. Users also get notifications and are able to approve actions that need approval.
Spark is capable of complex natural language prompts and can be used to automate all kinds of routine tasks. It can watch your Gmail when you book a trip and update your trip itinerary spreadsheet, search for local events and add recommendations to a calendar, organize your wedding RSVP emails into a table, go through your monthly invoices for subscription price changes, write customer responses according to pre-made templates for pricing packages, or even learn a user’s email writing style to draft future emails in the same tone.
The unique thing about Gemini Spark is that it doesn’t stop working after it’s given instructions. It runs on Google’s cloud computers, which means users can view emails, edit documents, and draft them even if they are not connected to the Internet or directly interacting with the assistant. This enables it to run repetitive digital requests without having to repeat the same requests by the user each time.
Google states users will always have control over the feature. They are able to pick the Google Apps that will have access to Spark, and can modify or deny these permissions as needed. The AI is also only allowed to perform sensitive actions with the user’s consent, which helps to maintain human control over important decisions.
The Gemini Spark is designed for professionals, entrepreneurs, students and frequent travellers who engage in repetitive digital tasks. In business, it can be used to automate boring tasks like sorting emails, organizing documents and writing customer correspondence.
