Your AI Agent Just Got Its Own Email Address

How Google's Gemini Spark works while you sleep, and why the first AI product designed to operate without the user present changes everything about knowledge work

Author

Justin Wright
May 25, 2026 • Est. Reading Time: 5 minutes

Happy Monday!

Last week, we covered Google declaring that Android was transitioning from an operating system to an intelligence system. AI agents that work across apps, completing tasks on your behalf. The implication was that the app economy breaks when users stop opening apps.

This week, Google went one step further. At I/O on Tuesday, Sundar Pichai unveiled Gemini Spark: a personal AI agent that runs 24/7 on a dedicated Google Cloud virtual machine. It does not stop when you close your laptop or pause when you lock your phone. It has its own Gmail address. You can email it instructions, and it will execute them while you sleep.

"It runs on dedicated virtual machines on Google Cloud seamlessly," Pichai told reporters. "You don't need to keep your laptop open to make sure it's running."

This is not just an assistant that helps you while you work. It is an agent that works instead of you when you are not there. That is a completely different product category than anything that has existed in consumer technology before. While we have covered Open Claw in the past, this is a completely different step in launching a more secure, consumer-facing version of this technology.

Google unveiled Gemini Spark at I/O 2026, a 24/7 AI agent that runs on cloud VMs with its own Gmail address. It monitors inboxes, drafts emails, pulls data across Workspace apps, and executes multi-step tasks autonomously, even when all your devices are off. It launches next week for AI Ultra subscribers at $100/month. Spark is the first mainstream AI product designed to work without the user being present, and it arrives as McKinsey estimates AI agents could automate tasks occupying 44% of US work hours. Last week Google killed the app. This week it is coming for the 9-to-5.

TL;DR

How Spark Actually Works

Each Spark instance gets a dedicated cloud virtual machine that hosts its memory, tools, and execution sandbox. The agent is powered by Gemini 3.5 Flash and Google's Antigravity agent framework, and it integrates natively with Gmail, Google Docs, Sheets, and Slides out of the box.

The Gmail integration is the detail that matters most; Spark has its own email address. You can email it a task the same way you would email a colleague. It can monitor your inbox, flag important messages, draft responses, and pull data from across your Workspace to compile reports or status updates. Google VP Josh Woodward described the use case directly: "Small businesses are using Spark. They can watch over their inbox, so they never miss a question from a customer."

Spark also supports MCP (Model Context Protocol) for third-party integrations. At launch, it connects to Canva, OpenTable, and Instacart. Adobe, Samsung, Spotify, CapCut, and dozens more are coming over the summer. On mobile, a new Android Halo system lets you track the agent's progress in real time.

What Makes This Different

Anthropic has Claude Cowork. OpenAI has ChatGPT Agent. Both are capable agentic products, but Spark has a structural advantage neither can easily replicate: Google already has all your emails, documents, calendars, contacts, and payment methods. The setup friction that plagues every other agent disappears when the agent lives inside the ecosystem that already houses your digital life.

The always-on architecture is the other differentiator. Claude Cowork and ChatGPT Agent require an active session. You prompt, they execute, you review. Spark runs continuously on its own VM, closer to a background service than a chat interface. The mental model is not "I am using an AI tool." It is "I have a digital employee who never clocks out."

The 44% Number

Spark arrives in a context that makes it harder to dismiss as a novelty. McKinsey's latest research on AI and labor estimates that AI agents could automate tasks occupying 44% of US work hours today. Keep in mind this is not 44% of jobs, but 44% of the tasks within those jobs. The distinction matters because it means AI does not replace the worker wholesale. It removes the parts of the job that consume nearly half of the workday: document preparation, data gathering, email management, scheduling, basic research, status reporting.

The World Economic Forum projects that agentic AI systems could automate up to 70% of office-based tasks by 2030. McKinsey's midpoint scenario estimates AI agents and robots could generate $2.9 trillion in annual US economic value by the same year, but only if organizations redesign workflows around human-agent collaboration rather than simply layering AI onto existing processes.

Spark does not cause these trends, but it does give them a consumer interface. When a $100/month subscription provides a 24/7 agent that monitors your inbox and completes tasks while you sleep, the question every knowledge worker faces becomes uncomfortably concrete: which parts of my job can this do, and what do I do with the time that frees up?

The Two-Week Arc

In ten days, Google made two moves. Week one: Gemini Intelligence turns Android into an agent that moves across apps autonomously. eMarketer estimated it could disrupt two-thirds of the app economy. Week two: Gemini Spark runs 24/7 on a cloud VM while your devices are off.

Week one killed the app and week two is coming for the workday. Gemini Intelligence removes the need to open apps. Spark removes the need to be present. Together, they represent Google's vision of AI not as a tool you use but as an agent that operates your digital life on your behalf. The user becomes the manager while the AI becomes the worker.

What This Means for Practitioners

For knowledge workers, the honest assessment is this: if your daily work consists primarily of email management, document compilation, scheduling, and status updates, Spark can do most of it for $100 a month. That does not mean your job disappears tomorrow. It does mean the value of those tasks trends toward zero, and your value increasingly depends on the judgment, creativity, and relationship skills that Spark cannot replicate.

For engineering and product teams, Spark's MCP support is the actionable detail. If your product is not accessible through MCP, it risks being invisible to the agent layer that is increasingly mediating how users interact with services. Building MCP compatibility is no longer optional for products that depend on user engagement.

For business leaders, the pricing is the signal. Google cut AI Ultra from $250 to $100. That is a market-making move designed to put Spark in the hands of every freelancer and small business owner. At $100 per month, the comparison is not to other AI products but to the cost of a part-time assistant.

The Bottom Line

Google gave its AI agent an email address, a cloud VM, and a job that never ends. Spark monitors, drafts, compiles, and executes while you sleep. At $100 a month, it costs less than a single hour of professional services in most industries.

Last week, Google killed the app. This week, it is coming for the 9-to-5. The question is no longer whether AI will change your job. It is whether the parts of your job that Spark can handle are the parts you are being paid for.

In motion,
Justin Wright

If a $100/month AI agent can monitor your inbox, compile your reports, and execute your tasks at 3 AM, what exactly are you doing between 9 and 5 that it cannot, and how long before your employer asks the same question?

Food for Thought

If you haven’t listened to my podcast Mostly Humans: An AI and business podcast for everyone yet, new episodes drop every week!

Episodes can be found below - please like, subscribe, and comment!