Gemini Spark: What Google's 24/7 AI Agent Means for Small Businesses

By Eshaan Pawan, Head of Growth at Runable

Published July 30, 2026 · Last updated July 30, 2026

Key Takeaways

• Gemini Spark is Google's always-on personal AI agent, and it is best for recurring chores inside your Google account (inbox triage, invoice audits, scheduled digests), not for one-off projects that need a finished deliverable.
• Spark is not free. It requires a Google AI Pro or Ultra subscription, which is $20 per month for Pro in the US and Rs 1,950 per month in India as of July 2026, and it is excluded from both Google's free plan and the cheaper Google AI Plus tier.
• Google rolled Spark out to India on July 29, 2026, with availability arriving for Pro and Ultra subscribers "over the next few weeks," and to Hong Kong the same week, per Google's India blog and the South China Morning Post.
• Spark runs on Google's cloud, so it keeps working when your laptop is closed and your phone is locked, and it connects to Gmail, Docs, and Sheets with no setup.
• The real limitation is the walls, not the timing: Spark is strongest on data already inside your Google account, which means open-web research, document production, and app building still need a general-purpose agent alongside it. General-purpose agents now run scheduled and triggered tasks too, Runable included, so "does it work on a schedule" no longer separates the two.

Gemini Spark at a glance

What is Gemini Spark?

Gemini Spark is a personal AI agent that works in the background under your direction, on a schedule or trigger you define, instead of waiting for you to type a prompt. Google describes it as a shift "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 mechanic that matters is where it runs. Spark executes on Google's cloud infrastructure, not on your device, so a task you set up on your phone continues to run after you lock the phone or shut the laptop. Google's India announcement puts it plainly: Spark "works quietly in the background, even when your laptop is closed or your phone is locked."

Spark is driven by Gemini 3.6 Flash, Google's faster and cheaper model tier, which is a sensible engineering choice for something that may fire dozens of small background jobs a day rather than one large reasoning job.

Google first showed Spark at I/O 2026 in May. What changed in late July 2026 is access. Spark spent its early life locked behind Google AI Ultra, and Google has since brought it down to the $20 per month AI Pro tier in the US and started the India and Hong Kong rollouts. This is a distribution event more than a product launch, and distribution is what makes an agent matter.

How Spark differs from the Gemini app you already use

The Gemini app is a conversation. You ask, it answers, the session ends. Spark is a standing instruction. You describe a condition and an action once, and it keeps applying that rule until you turn it off.

That difference sounds small and is not. A conversational assistant saves you the minutes it takes to write something. A background agent saves you the overhead of remembering to do something at all, which is usually the more expensive of the two for anyone running a business.

What can Gemini Spark actually do?

Spark handles tasks that can be expressed as a trigger plus an action on data it can reach. Google published example prompts alongside the India launch, and they are worth reading closely because they reveal the shape of the product better than any feature list.

Grouped by job, the published examples cover:

Inbox rules that produce artifacts. "Whenever a new flight or hotel booking confirmation email arrives, automatically add the details to my 'My Bali Trip Itinerary' Google Sheet." The agent watches Gmail, extracts structured fields, and writes them into a Sheet.

Scheduled research digests. "Every Friday at 8AM, find local events that you think I'd be interested in and add them to my rolling 'Weekend plans' Google Doc." This one combines a recurring timer, an open-web lookup, and a write into Docs.

Third-party app checks with conditional alerts. "Every Tuesday at 8am, log into and check the PickleMe app for free indoor evening courts. If only outdoor courts are available, alert me only if the forecast is clear." Spark can reach outside Google properties, and it can branch on a condition.

Data extraction into tables. "Find my wedding RSVPs in Gmail, organise the dietary restrictions into a table, and draft a summary email to our caterer."

Recurring financial hygiene. "Every month on the 1st, audit my digital invoices in Gmail from the last month. Notify me about any subscription price hikes, or trials that are about to end and pre-draft cancellation emails."

Sales and support triage. "Scan my inbox for customer inquiries. Automatically draft an email response recommending my services and pricing packages based on their request, and save it to drafts for my review." Google labels this example the "hustling entrepreneur" use case, which tells you who Google thinks will pay for this.

Reusable skills. "Read through the last 50 emails that I wrote and turn it into a style guide for how I write emails. Turn that into a skill that gets used every time I ask you to draft emails for me. Call that skill 'ghostwriter'." Spark can define a named, persistent behavior and reuse it, which is closer to programming than prompting.

Read as a set, every one of those examples has the same structure: a trigger, a source Spark can already see, and a write into a Google surface. That is the product's center of gravity.

The consent model

Spark asks before it does anything expensive or public. Google states it "is designed to ask you first before performing high-stakes actions like spending money or sending emails," and that you choose whether to turn it on and which apps it connects to.

This is the correct design, and it also sets your expectations. Spark drafts, organizes, and notifies autonomously. It does not send, buy, or commit on your behalf without a check. In the customer-inquiry example above, note that the drafted reply is saved to drafts, not sent.

How much does Gemini Spark cost, and is it free?

Gemini Spark is not free, and there is no free trial tier for it. It requires a paid Google AI subscription, and Google has explicitly kept it out of both the free plan and the lower-priced Google AI Plus subscription.

The price drop is the story. Chrome Unboxed reported that since its debut, Spark "has been locked behind the Google AI Ultra subscription, which runs anywhere from $100 to $200 a month," and that bringing it to the $20 AI Pro tier in the US "makes this more accessible for regular users." An agent that costs $200 a month is a demo for enthusiasts. An agent that costs $20 a month is a line item a freelancer will actually consider.

In India, Google AI Pro is priced at Rs 1,950 per month, with Ultra starting at Rs 6,500 per month and a higher tier at Rs 19,500 per month, according to BestMediaInfo's report on the India rollout. Ultra buys higher usage limits and earlier access to some features rather than a different Spark.

One caveat worth stating: Spark consumes your plan's usage allowance. A background agent that checks your inbox every morning and runs three scheduled digests a week is not free to operate against your quota, even though the subscription price is flat. Budget for the agent to use noticeably more of your allowance than chat did.

Where is Gemini Spark available?

As of July 30, 2026, Spark has reached the US, India, and Hong Kong.

The India rollout was announced on Google's India blog on July 29, 2026, and it is a staged rollout rather than a flip of a switch. Google says Spark is "officially expanding to Google AI Pro subscribers in India over the next few weeks," and BestMediaInfo notes the feature "may not appear immediately for every eligible user." If you subscribe today and do not see it, that is expected.

Hong Kong got Spark the same week, with the South China Morning Post framing the rollout as Google lowering its geofences on agentic features.

The staged, region-by-region pattern is worth watching if you operate outside these markets. Spark is a cloud service tied to a Google subscription, so expansion is a policy decision rather than a technical one, and the direction of travel is clearly outward.

What Gemini Spark means for small businesses

Here is the honest read for anyone running a small business, a freelance practice, or a two-person team.

The busywork floor just dropped. A meaningful share of small-business admin is rule-shaped: sort the inbox, chase the invoice, log the booking, draft the standard reply, produce the weekly summary. That work was too small to hire for and too frequent to enjoy. Spark handles exactly this class of task, on data you already keep in Gmail and Sheets, for $20 a month. If your day contains four or five recurring digital chores, this is the cheapest staff you will ever add.

Distribution beats capability. Spark is not the first background agent, and on raw capability it is not obviously the strongest. What it has is a billion-user surface, native access to the inbox where the work actually lives, and zero setup. Most small businesses never got around to wiring up an automation tool. They will not need to wire up Spark. That is the part competitors should worry about.

The $20 price point is now the anchor. Google has effectively set the reference price for "an AI agent that does things for you" at $20 per month. Every tool in this category, including ours, now gets compared against that number. For buyers this is straightforwardly good news.

Your Google account becomes more valuable and more concentrated. Spark's advantage is that it can already see your mail, your docs, and your sheets. That is also the tradeoff. The more of your operation Spark runs, the more of your operation depends on one vendor's account, quota, and regional availability. Worth a moment's thought if Google Workspace is already where your business lives.

It does not replace the projects. Spark is built for the recurring 15-minute task. It is not built for "research the 12 competitors in my category and give me a positioned comparison deck," or "build me a small internal tool for tracking inventory," or "write and format the 2,000-word case study." Those are one-off, open-ended jobs that produce an artifact, and they remain a different kind of work.

Where Gemini Spark falls short

No product review is useful without the drawbacks, so here are the real ones.

It is strongest inside Google's walls. Spark's zero-setup advantage comes from native Workspace access. If your business runs on Outlook, Notion, Airtable, or a CRM Google does not natively read, much of that advantage evaporates. Spark can reach some third-party apps, but "can log into an app and check a page" is a different capability from "is deeply integrated with your stack."

Paywalled at exactly the tier most people are not on. Excluding Spark from the free plan and from Google AI Plus is a deliberate monetization choice. The majority of Gemini users will read about Spark and not have it.

Staged availability makes it unreliable to plan around. "Over the next few weeks" for India, not visible to every eligible user immediately, and three markets so far. If you are building a workflow your business depends on, a feature you cannot confirm you have is a shaky foundation.

Background autonomy is hard to audit. An agent that acts while you sleep is an agent whose work you review after the fact. The consent gate on sending and spending helps a great deal, but you will still want to check its drafts and its Sheet writes for a while before you trust them unread.

Recurring rules are not the same as judgment. The published examples are triggers and transforms. Anything requiring taste, positioning, or a decision you would defend to a client is not what this product is for.

Gemini Spark vs a general-purpose AI agent

The clearest way to think about this category is by the shape of the work, not the brand.

Runable is best for marketing, research, and content work that ends in a deliverable. You describe the outcome you want, in a sentence, and it executes the whole task and hands back the finished artifact: the competitor analysis, the deck, the blog post, the small internal tool. That is a different job from Spark's, and it is the one we have built for.

Runable now runs scheduled tasks too

One row of that table changed on our side. Runable now supports scheduled tasks, which means a task can run on a schedule or fire on a trigger you set instead of waiting for you to brief it again. The work you used to re-request every Monday morning now runs on its own and returns the finished output.

That matters for the comparison because "recurring" was the cleanest reason to reach for Spark over a general-purpose agent. It no longer is. A weekly competitor scan, a monthly content report, a Friday pipeline summary: those are recurring jobs whose output is a deliverable, and a scheduled Runable task produces the deliverable rather than a notification. The distinction that survives is not how often the work runs. It is what the agent can reach and what it hands you at the end.

To be direct about our own gaps: Runable does not have Google's zero-setup native hooks into Gmail, Docs, and Sheets, and it does not sit inside your mailbox watching what arrives. If what you want is "read my email every morning and file what is in it," Spark sits closer to that data and is the better instrument. Our pricing as of July 2026 is a free tier, Pro at $20 per month (our most popular plan), and Max at $100 per month.

The genuinely useful conclusion is that these are complements, not substitutes, and the split is by data and output rather than by timing. Spark keeps the rule-shaped machinery inside your Google account running. A general-purpose agent takes the projects off your desk, now on a schedule if you want them to keep coming. Most people running a small business need both, and both now start at $20 a month or less.

How to choose what you need

Ask what shape your problem is.

If the work is reading and writing inside your Google account, use Spark. Inbox triage, weekly digests, monthly invoice audits, booking confirmations into a Sheet. This is its home turf and $20 is a fair price for it. Note that repetition on its own no longer decides this question, because general-purpose agents run scheduled tasks now as well. The deciding factor is where the data lives.

If the task ends in a deliverable, use a general-purpose agent, whether you need it once or every week. Anything you would otherwise brief a freelancer on, where the output is a document, a deck, an app, or a piece of content, and where a schedule just means you want that output to keep arriving.

If your stack is not Google, discount Spark's convenience advantage substantially and evaluate on capability instead.

If you are on Google's free plan or AI Plus, you do not have Spark, so the practical comparison is between upgrading to AI Pro and using a tool with a free tier. Try the free option first and find out whether your real bottleneck is chores inside your inbox or projects that end in a deliverable. Most people guess wrong about which.

If you are in India, expect a wait even after subscribing, and note that Rs 1,950 per month is a real consideration against alternatives with free tiers.

What our own search data says about Spark's momentum

We pulled the demand picture on July 30, 2026, the day after the India announcement, and it is a useful snapshot of how fast a new agent brand forms.

In Ahrefs US data, "gemini spark" shows 50 monthly searches at a keyword difficulty of 4, with "what is gemini spark" at 150 per month and "gemini spark review" and "gemini spark cost" at 30 and 20 respectively. The broader category term "personal ai agent" sits at 250 per month with a difficulty of 22.

Those volumes look small, and there is a reason: Ahrefs volume is a trailing twelve-month average, and Spark has existed publicly since May 2026. The forward-looking signal is in autocomplete, where Google is already suggesting "gemini spark release date," "gemini spark india release," "gemini spark cost," "gemini spark reviews," "gemini spark agent," and comparison queries against other named agents. Autocomplete reflects current query behavior rather than a year of history, and it is already dense.

The practical read: this is a brand-new entity with near-zero ranking difficulty and rapidly forming demand, in a SERP currently occupied by Google's own pages, one Verge hands-on test, a Medium post, and a YouTube video. There is very little independent analysis. If you publish in this category, this is the week.

Frequently asked questions

Is Gemini Spark free?

No. Spark requires a paid Google AI Pro or Ultra subscription. Google has excluded it from both the free plan and the lower-priced Google AI Plus tier. AI Pro is $20 per month in the US and Rs 1,950 per month in India as of July 2026.

What model does Gemini Spark use?

Gemini Spark runs on Gemini 3.6 Flash, Google's faster model tier. That choice suits an agent designed to run many small background jobs rather than a single long reasoning task.

Does Gemini Spark work when my laptop is closed?

Yes. Spark executes on Google's cloud infrastructure rather than on your device, so scheduled and triggered tasks continue to run when your laptop is closed or your phone is locked.

Is Gemini Spark available in India?

Yes, as of the July 29, 2026 announcement, with a staged rollout. Google says availability is expanding to AI Pro subscribers "over the next few weeks," and eligible users may not see it immediately.

Will Gemini Spark send emails or spend money on its own?

No. Google states Spark is designed to ask you first before high-stakes actions such as spending money or sending emails. In Google's own customer-inquiry example, the agent drafts a reply and saves it to drafts for your review rather than sending it.

Is Gemini Spark the only AI agent that runs tasks on a schedule?

No. Scheduled and triggered tasks became standard across the category during 2026, with the major assistants and general-purpose agents all shipping some form of it. Runable supports scheduled tasks too, so a recurring job can return a finished deliverable rather than a notification. Choose on the data an agent can reach instead.