Gemini Spark is finally available on Google AI Pro in Japan. Announced for Japan on July 29, 2026 (Google Japan Blog), it rolled out over several weeks, and the tab appeared on my account in August.
You can run up to 15 tasks simultaneously, and they won't stop even if you close your laptop. This is because the processing happens on Google's cloud side, not your local device.
It's now within reach for a monthly plan of 2,900 yen.
However, there is something I should tell you first.
It cannot be used with company or school Google accounts.
If you read this and try to use it but the tab doesn't appear, this is usually the reason.
"I asked the AI to do it, but I ended up stuck in front of the monitor giving instructions like 'next is this.'" "I know it's convenient, but nothing progresses while I'm sleeping."
About six months after starting to use AI, most people hit this wall. The intelligence is there. What's missing is the ability to work while we aren't watching. Spark is a feature designed to fill that gap head-on.
I'm tatsuki (@nobel_824). I support AI implementation for small and medium-sized enterprises, helping with Claude/Codex business integration while running Claude Code all day myself.
I've explored the tool, read the official help from start to finish, and cross-referenced it with three Japanese and three English verification articles. I've broken down what it can do into 40 items. I've also summarized the conditions and limitations scattered throughout the official help into one sheet.
In the latter half, I'll provide 8 copy-pasteable prompts and dive into intermediate topics like "what is the true nature of Skills" and "how it differs from Claude or ChatGPT agents." By the way, Spark is powered by Gemini 3.5 Flash, not Pro. I'll explain why in Chapter 6.
This information is as of August 8, 2026. Spark is in beta and updated weekly, so honestly, now is the time to read this. The latest changes are listed line-by-line on the official update history page.
I'm giving away a bonus at the end of the article.

Chapter 1: Understanding Gemini Spark via 4 Components
There are many feature names, which can be confusing, but the structure is simple. It has Tasks for the work, Schedules for when to run, and Skills as templates for how to do it. On top of that are Connected Apps for which apps it can touch. That's it.
Tasks: The work itself
Like "Check this week's inbox and extract only those needing a reply," this is the unit of work completed across multiple apps. The difference from normal chat is that you don't have to wait for a reply. You can close it after sending, and you'll get a notification in the Gemini app when it's done. It runs in the background on a remote computer in Google's cloud.
Schedules: When to run
Once, hourly, daily, weekly, monthly, or yearly. In addition to time, there are two types of triggers:
- When an email matching specific conditions arrives
- When topics like news, finance, sports, or local events move
Skills: The template for how to do it
The official definition is "a set of reusable instructions and additional context." It's a feature to save procedures you don't want to explain every time. For example: "Read the last 50 emails I wrote and use them as a style guide." You can upload skills as SKILL.md files.
Connected Apps: What it can touch
Google apps include Gmail, Calendar, Drive, Docs, Sheets, Slides, Keep, Tasks, YouTube, and Maps. Third-party apps like Canva, Instacart, OpenTable, Dropbox, and Zillow are already added. You can also connect apps via MCP (Model Context Protocol) server URLs.
Chapter 2: Getting it Ready (Where most people get stuck)
Requirements:
- Must be 18+ years old
- Must be a personal Google account
- Must have a Google AI Pro or Ultra subscription
- "Gemini Apps Activity" must be turned on
- Must be in a supported region (Japan is supported; EEA, UK, Switzerland, and Nigeria are not)
The biggest pitfall: Not for work accounts. Workspace accounts are currently excluded. A workaround is to run Spark on a personal account and pass the results to the company side via shared drives.
Pricing:

Limits:

Chapter 3: Catalog of 40 Capabilities

A. Email (Gmail)
- Scan inbox for items needing action.
- Summarize/archive newsletters.
- Unsubscribe from mailing lists.
- Create reply drafts.
- Extract items from emails to Sheets.
- Create a style guide from sent emails.
- Summarize entire threads.
B. Calendar & Appointments
- Find schedule conflicts.
- Build itineraries from confirmation emails.
- Plan several days across emails, tasks, and calendar.
C. Drive / Docs / Sheets / Slides
- Scan Drive to list important files.
- Organize files into folders.
- Create and edit Docs, Sheets, and Slides.
- Append data to Sheets periodically.
- Read document comments.
D. Keep / Tasks
- Read/write Keep notes.
- Create/organize Tasks.
E. Web Browsing
- Research across multiple sites with citations.
- Compare products/services.
- Navigate to login screens via remote browser.
- Manage/delete saved cookies.
- Operate your local Chrome (Chrome auto browse - 20/200 limit).
- Proceed up to the point of booking/application.
F. Monitoring (Schedules)
- Run by time.
- Run by conditions (emails/topics).
- Monitor news, stocks, sports, weather.
- Manual creation/stopping from the schedule screen.
- View progress/interrupt running tasks.
G. Custom Templates (Skills)
- Create skills via dialogue with Gemini.
- Edit templates or write from scratch.
- Upload
SKILL.mdor.zip(up to 100MB). - Turn a successful task into a skill.
H. Local Computer
- Read/analyze/organize Mac folders.
- Remote control Mac from mobile/tablet.
- Execute code on a remote computer.
- Delete remote code execution data.
I. External Services
- Connect with Canva, Instacart, OpenTable, Dropbox, Zillow.
- Connect any app via MCP server URL.
- Toggle connections per app.
- Use YouTube and Maps info.
Chapter 4: 8 Copy-Paste Prompt Recipes
- Morning Briefing: Summarize calendar, urgent emails, news, and weather at 7 AM.
- Weekly Inbox Cleanup: Archive unnecessary emails and draft replies for others.
- Multi-day Planning: Coordinate calendar, tasks, and emails for the next 5 days.
- Style Guide Skill: Analyze 50 sent emails to create a writing style guide.
- Periodic Research: Weekly AI news research appended to a Google Doc.
- Expense/Subscription Audit: Extract subscription info from the last 3 months of emails to a Sheet.
- Inquiry Handling: Automatically log "Inquiry" emails to a Sheet and draft replies.
- Travel Planning: Compare hotels and create an itinerary in a Doc.
Chapter 5: 6 Pitfalls
- "Designed to confirm" vs "Always confirms": Don't trust it with irreversible actions like sending or buying without manual check.
- Inbox as an Input: Risk of prompt injection via incoming emails.
- Accuracy: It's good at summarizing but can struggle with specific numerical data.
- Beta Status: Features may disappear and reappear.
- Silent Stops: If you hit the 15-task limit, new schedules won't run without warning.
- Mac File Operations: Changes are hard to undo.
Chapter 6: Deep Dive for Intermediate Users
Spark uses Gemini 3.5 Flash for speed and parallelism. While it supports SKILL.md like Claude, it is currently more of a "prompt-based task memory" than a full execution environment. The choice between Spark, Claude, and ChatGPT often depends more on your data environment (Google ecosystem) and company regulations than raw intelligence.
Summary
We are moving from being AI directors to AI judges. Spark moves the waiting time outside your calendar, but the work of judging the output remains. Start with one task that wouldn't hurt if it failed, and grow from there.
3 Steps to Try Today
- Check for the Spark tab in the Gemini app.
- Register the Morning Briefing recipe tonight.
- After a week, manually verify 3 numbers from the output to understand its accuracy.
If you want to learn more about AI utilization, Claude Code, and X operations, join my anonymous LINE OpenChat to receive 7 free Claude utilization bonuses.


Membership for serious learners:





