Google launches Gemini agent with its own email, calendar and Workspace account
Google's new workplace agent can be given its own Workspace account, calendar and seat in the company directory. The same day, the UK's ICO opened a consultation on agent risks — and the hard governance questions remain unanswered: no price, no date for general availability, and no explanation of how to revoke an agent's identity.
The concrete news
Google Cloud launched the new Gemini agent on October 8, 2026 at its Gemini at Work 2026 event. The company describes it as "a single, universal agent for work" intended to handle knowledge work, question answering, content creation and coding — all from one prompt box, according to Unite.AI.
What sets the launch apart from competitors' is not primarily what the agent can do, but who it is. In what Google calls "coworker mode", the agent can be provisioned as a persistent colleague with its own Workspace account, its own email address on the format @agents.company.com, its own calendar, its own Drive storage and its own place in the company directory. VentureBeat gives a concrete example from Google's material: a manager can create an "Event Planner Agent" that gets all of this as a single configuration — not as mandatory equipment for every agent instance.
The agent is also persistent: it can carry out tasks that run for hours or days, coordinate other AI agents, and retain the same memory, business context and identity across apps and devices, VentureBeat writes.
How the delegation model works
Google Cloud CEO Thomas Kurian summarised the thinking to SiliconANGLE: "You give it objectives, not instructions. You delegate an outcome and come back to finished work." And further: "For an agent to do that, it has to be connected to your personal workflows, your systems of record and your enterprise controls."
In the launch post, Kurian wrote, according to AI Weekly: "Today, we're excited to announce the Gemini agent: your new single, universal agent for work … Work now starts in the prompt box."
In coworker mode, the agent only sees the context that users or team members grant it access to, according to Unite.AI. Under the hood, the agent orchestrates, according to Google, across the Gemini model family and Anthropic's Claude models, with more private and open models planned — the model is matched to each task for quality and cost.
Identity and controls: machines as traceable actors
The governance mechanisms are the most consequential part of the launch. Each coworker agent gets a cryptographically attested identity, "stamped into its logs and any virtual machine it starts", Kurian wrote in the announcement, as TNW reports. Administrators approve role-based access, and audit trails are kept on the agent itself — not on the person who delegates the job.
That means when an agent sends an email, edits a document or starts a virtual machine, it is the agent's identity that appears in the log, not an employee's. It is a concrete change in how attribution of responsibility is conceived in enterprise systems: the machine becomes its own traceable directory entity.
There is also an economic gatekeeper. Users can set hard spending caps for AI consumption in the Cloud Billing Console, according to SiliconANGLE. Gemini monitors token consumption, and when the cap is triggered, the agent stops. The user can resume the work with approval — but must then accept exceeding the budget.
A packed fortnight of competitors
The launch does not come in a vacuum. VentureBeat's timeline shows a wave of persistent agent launches: Microsoft introduced an updated Copilot on September 25 that combines productivity work, software development and persistent automation. Four days later came OpenAI's always-on "dots" agents. And on October 6, Anthropic announced native Claude integration with Google Docs, Sheets and Slides, allowing users to edit Workspace files directly through Claude or use Claude inside Google's productivity apps.
Google's answer in this field, then, is to bet on identity: not just an agent that does things, but an agent that is someone in the company's systems.
Regulatory timing: the ICO right alongside
The timing is almost comically apt. Just hours before Google's announcement, the UK Information Commissioner's Office (ICO) had named Google as one of ten developers from which it has secured changes in data protection, and opened a consultation on agent risks with a response deadline of November 20, TNW reports.
The ICO's director of technology regulation, Richard Nevinson, was clear: that agents act with autonomy "is not an excuse for poor compliance".
That places Google's announcement in a specific context. The company presents attested identity and admin-approved access as a governance answer — but the UK regulator is simultaneously signalling that autonomy shifts responsibility, it does not remove it.
What is still missing
Despite the technical details, the commercial and governance foundations are unresolved:
- No price and no date for general availability. According to AI Weekly, neither was announced.
- No regions for the sovereignty claim. Google's post mentions Deutsche Telekom as a European customer for data sovereignty, but gives no regions, TNW writes.
- No explanation of revocation. TNW's analysis points out that the post does not describe how to revoke an agent's access or identity — a gap that becomes more visible the more the agent resembles an employee with its own account and calendar. This is TNW's analysis, not a confirmed deficiency, but the question remains unanswered in the material.
There are also no details on provisioning mechanics or licensing: whether administrators, for example, can switch individual components of an agent's Workspace presence on and off is not specified in the published material.
Kurian's numbers — the company's own claims
Kurian's post is rich with numbers, but the figures are not independently verified: according to Unite.AI, the post states that nearly 500 Google Cloud customers have each processed over a trillion tokens in the past year, that nearly 80 percent of all Google Cloud customers use the company's AI products, and that nearly 90 percent of the Fortune 100 use Gemini Enterprise. These are Google's own claims and should be read as marketing until independent verification is available.
What to watch
Two things will determine whether coworker agents become a real, governed category or a demonstration. The first is the ICO consultation: the responses coming in by November 20, and the ICO's subsequent line, will show whether Google's attested-identity model holds up as a compliance answer — or whether the regulator demands more, particularly on revocation and data flows.
The second is whether Google fills in the gaps: price, general availability date, regions for the Deutsche Telekom sovereignty arrangement, and concrete documentation on provisioning and revoking agent identity. Until then, the "workplace colleague" is a technical possibility with an unresolved governance regime — precisely where a UK regulator has just said that autonomy does not exempt from accountability.