Google will let an AI agent work inside Indian users’ Gmail while their devices are off. The company said Gemini Spark reaches Google AI Pro subscribers in India over the next few weeks, with the agent running on Google’s own cloud infrastructure.
Quick Summary – TLDR:
- Google is rolling out Gemini Spark, a background AI agent, to Google AI Pro subscribers in India over the coming weeks.
- Gemini Spark runs on Gemini 3.6 Flash and keeps working when a laptop is closed or a phone is locked.
- Google AI Pro was listed at $19.99 a month and AI Ultra at $249.99 when Google published its Gemini tier caps.
- Gmail reaches over 82% of internet users in Indonesia and India, giving Spark a large base with no setup.
- Google says Spark asks permission before high-stakes actions such as spending money or sending emails.
What Happened?
Google announced Gemini Spark, a personal AI agent that runs on Google’s servers and acts on a user’s Workspace data under standing instructions. Google said the tool “fundamentally shifts AI from a reactive assistant that answers your questions into an active partner that gets real work done for you, even while you’re sleeping.”
The agent arrives with no connector setup. “It operates on Google’s secure cloud infrastructure, which means it connects natively with your Google Workspace tools like Gmail, Docs, and Sheets with zero setup required,” the company stated in its announcement. Gemini 3.6 Flash drives it.
Google’s post carries two eligibility lines. One says Spark is “expanding to Google AI Pro subscribers in India over the next few weeks.” A second says the tool is “Available to Google AI Ultra and Pro subscribers.” Google has not published an India price for Spark or for either tier alongside the announcement.
Google Rolls Out Gemini Spark In India, Bringing Autonomous AI Agent To Gmail, Docs & Sheetshttps://t.co/fylnBxy9lG
— Free Press Journal (@fpjindia) July 29, 2026
Why India Gets the Agent on the Cheaper Paid Tier?
Google put an autonomous agent on the lower of its two paid consumer tiers. AI Pro carried a $19.99 monthly listing and AI Ultra $249.99 when Google published per-tier caps for the Gemini app in September 2025. That places the headline capability on the tier most subscribers would actually buy.
India rewards that math. Gmail is used by over 82% of internet users in Indonesia and India, among the highest penetration rates worldwide, so Spark lands with its data source already populated and its permissions already granted. Zero setup is a distribution decision as much as an engineering one.
OpenAI reached the same market with price. ChatGPT drew over 29 million downloads in India across a 90-day window while generating $3.6 million in revenue over that same period, and OpenAI answered by launching ChatGPT Go at 399 rupees a month with UPI payment support. Google is answering the same gap with capability at its existing price.
What Spark Can Actually Touch?
Google published seven example prompts with the announcement. The most revealing has Spark building a reusable capability for itself: the user asks it to read their last 50 sent emails, then “Turn that into a skill that gets used every time I ask you to draft emails for me. Call that skill ‘ghostwriter‘.” Another sends the agent outside Google’s own products, instructing it to “log into and check the PickleMe app for free indoor evening courts.“
Google set a consent gate around the riskiest actions. Spark is “designed to ask you first before performing high-stakes actions like spending money or sending emails,” and users choose whether to turn it on and which apps it connects to.
The announcement establishes the permission model without the operating detail around it. Four things Google has not stated:
- What Gemini Spark costs in India, and whether the India rollout covers AI Ultra alongside AI Pro.
- How many actions Spark can chain before it checks back with the user.
- Which third-party services it can sign into beyond the PickleMe example.
- What happens to its standing access when a subscription lapses.
Users who switch Spark on can start by reviewing the connected app list and the Gmail drafts folder, since several of Google’s own examples end in an unsent draft waiting on approval. Auditing that list helps reduce the risk of the agent reaching data the user never meant to expose.
SQ Magazine’s Takeaway
The trust model matters more here than the model number. A chat assistant fails visibly, in a window the user is watching. An agent that works while a phone is locked fails quietly, inside a Sheet the user opens next week. Google’s answer is a permission prompt on money and outbound email, which covers the two failures a user would notice fastest and leaves the quieter ones to be caught on review.
What comes next is the pricing and the tier answer, neither of which Google has given for India. The other thing to watch is how Spark handles third party sign-ins, since the PickleMe example points to stored credentials for services Google does not run. Subscribers who get the rollout can treat the first week as a trial run: hand it one low-stakes standing task, check what it produced, and widen the brief from there.
