Gemini Agent: Google Gives AI Coworkers Their Own Workspace IDs

A new kind of colleague

Google Cloud used its Gemini at Work 2026 event to introduce what CEO Thomas Kurian calls a single, universal agent for work. The Gemini agent is meant to handle knowledge work, coding, content creation and data analysis from one prompt box, and it is natively wired into Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar 14. According to Google, it also reaches beyond its own ecosystem into Microsoft 365 and Slack 34.

The most consequential detail is identity. Google is giving "coworker agents" their own presence inside Workspace, including an email address, a calendar, Drive storage and a listing in the company directory 2. Kurian framed the agent as something that can work for one person, for a team as a kind of project manager, or for a specific role, such as an analyst in a finance department 4. That is a different proposition from a chatbot in a sidebar. It is software with a seat at the table.

How it works

Users set objectives rather than giving step-by-step instructions. The agent then plans the work, spins up temporary sub-agents that run in parallel or in sequence, and keeps going after the person who assigned the task has logged off 2. Progress appears in a tasks inbox that shows the agent's reasoning, the sub-tasks it delegated, the skills it loaded, and any code it wrote 2. The agent can write and run code, generate images and media, and return finished deliverables instead of drafts 4.

It also does not rely only on Google's models. The agent can route jobs to Anthropic's Claude when Claude suits the task better 23. Futurum argues that separating the agent from the underlying model matters more than any single feature in the launch, because it signals that Google wants to own the workflow layer rather than win purely on model quality 3.

Different readings of the same launch

The four outlets agree on the basic facts but stress different things. CBS News treats it as a competitive move in a crowded race among tech giants to build virtual office assistants 4. DEV Community's coverage centers on the persistent identity and on multi-day autonomy 2. Futurum focuses on strategy, specifically the model-agnostic routing and what enterprise usage data says about Gemini and Claude adoption 3. Wursta, a Google Cloud partner, writes for IT leaders and puts governance first, arguing that a universal agent needs a universal governance framework 1.

Wursta also names a tension the other coverage mostly skips. It describes an "AI Paradox": leadership wants to scale AI for efficiency, while employees quietly worry about what these tools mean for them 1. Google's adoption figures make the tension sharper. Nearly 80% of Google Cloud customers use its AI products, and almost 90% of the Fortune 100 use Gemini Enterprise 1. The tools are already widespread. How they will be used is still unsettled.

Why the identity piece matters

Giving an agent a directory entry and an inbox has practical benefits. It lets the agent receive assignments, hold calendar slots, own files and be audited through the same systems that track human employees. For IT, that may be easier to govern than an assistant acting invisibly through a person's credentials. The tasks inbox, with visible reasoning and code, points in the same direction toward transparency 2.

The symbolism is harder to dismiss. A teammate that appears in the org chart, keeps working overnight, and can be assigned to a role like "finance analyst" invites direct comparisons to headcount 4. Even if Google pitches it as augmentation, workers will judge it by how their employers deploy it. Kurian's language about the agent acting as a team member is meant to reassure, but it also makes the substitution plain 4.

The takeaway

This launch looks less like a feature update and more like an attempt to redefine how AI enters the workplace, as a managed participant in it rather than a tool that sits beside it. Futurum is likely right that the model-agnostic architecture is the long-term strategic bet 3. In the near term, though, adoption will depend on what Wursta flags: governance and trust 1. Organizations that hand agents logins without clear policies on permissions, accountability and job design risk turning quiet unease into open resistance. The technology appears ready to function as a coworker. Whether workplaces are ready to treat it as one is a separate question, and it is mostly one for managers to answer.