Google is testing a universal Gemini agent designed to carry work across applications, devices and specialized AI helpers from a single conversational hub. The Verge reported on October 8 that the company announced the system during its Gemini at Work event and will make it available inside the Gemini Enterprise app. For now, access is limited to enterprise customers in a private preview, so the announcement describes an early product direction rather than a broadly available workplace service.

The central promise is continuity. According to The Verge, employees can assign tasks to the agent through one interface while Gemini works inside Google Workspace products including Gmail, Drive, Docs, Sheets and Calendar. Google also says the agent can be reached from mobile devices, desktop computers and the web, as well as through third-party services such as Slack and Microsoft 365.

A glowing thread carries work context between desktop, phone and cloud.
Google says the cloud-based agent can maintain context as employees move among connected devices.

Because the agent runs in the cloud, Google says it can retain the same context as a person moves among connected devices. That architecture is meant to turn Gemini from an assistant attached to one application into a persistent layer spanning several work surfaces. The report does not describe how much context the preview retains, what administrators can restrict, or how organizations will evaluate the system before wider deployment.

The enterprise agent can also coordinate with job-specific sub-agents, The Verge reported. Google says those specialized systems can participate in completing assigned work while Gemini automatically chooses the AI model it considers best suited to each task. The announcement therefore combines a general front door with more narrowly focused agents behind it, rather than requiring employees to select every model or specialized tool themselves.

A digital coworker badge and email envelope sit inside permission gates.
The preview gives the agent a dedicated corporate identity, while detailed controls were not described in the report.

Google is also giving the system the trappings of a digital colleague. The company says Gemini can operate as a coworker agent with a dedicated identity and its own address using an @agents.company.com email domain. That identity could make the agent easier to include in established workplace processes, although the report does not specify what permissions, review steps or accountability controls accompany the preview.

The new system resembles a corporate counterpart to Gemini Spark, the consumer-focused agent Google launched in May. Spark can perform tasks across Workspace and the web, but The Verge noted that it lives in a separate interface within the Gemini app. By placing the enterprise version inside Gemini Enterprise and connecting it to both Google and third-party software, Google is presenting a broader vision: one agent window that can answer questions, preserve context and carry out work across the tools an organization already uses.

That vision remains constrained by the product's private-preview status. Google has not, in the report, announced general availability or shown how the universal agent will perform across the many combinations of corporate apps and permissions it is intended to navigate. What the launch does establish is the company's preferred shape for workplace agents: persistent across devices, able to delegate to specialists and recognizable inside an organization as an actor with its own identity.