Google wants its AI to stop behaving like a feature and start behaving like a colleague. At its Gemini at Work event on Thursday, the company introduced what it calls a “universal” Gemini agent, an enterprise-focused assistant designed to take on tasks across apps and devices from one place, and to keep working in the background while you get on with something else.
The agent sits inside the Gemini Enterprise app, where users can chat with it and hand off work from a single interface. It also operates directly within Google Workspace, including Gmail, Drive, Docs, Sheets and Calendar. Google is also extending it beyond its own ecosystem: the agent can be reached from phones, desktops, the web, and third-party platforms such as Slack and Microsoft 365. Because it runs in the cloud rather than on a specific device, Google says it carries the same context everywhere, so a task started on a laptop shouldn’t need re-explaining on a phone.
The more interesting details are about how the agent is structured. Google says Gemini can delegate to job-specific sub-agents, and that it automatically selects whichever AI model it judges best suited to a given task. It can also be deployed as a “coworker agent” with its own dedicated identity and an email address on an “@agents.company.com” domain. That last point shifts the idea of AI at work from a tool employees use toward something closer to a digital staff member.
That shift raises practical questions Google hasn’t fully answered yet. If an agent has its own inbox and identity, companies will want clear answers on permissions, audit trails, and who is accountable when it sends the wrong email or edits the wrong spreadsheet. Automatic model selection is convenient, but it also means less visibility into what is doing the work at any given moment. Pricing and a wider release timeline also remain unclear.
For now, access is limited. The agent is only available to enterprise customers through a private preview, so most businesses won’t be able to try it soon. Consumers already have something similar in Gemini Spark, the agent Google launched in May for subscribers, which handles tasks across Workspace and the web. Spark, however, lives in its own separate space within the Gemini app rather than serving as the main way to use it.
That split is the real tell. Google appears to be moving toward one window where you ask questions and assign work interchangeably, and the enterprise version is where it’s testing that idea first. Whether businesses will trust an AI coworker with its own email address is the question the preview will need to answer.
