18:36Machine-translated from Azerbaijani by AIRead in Azerbaijani
Google Cloud Unveils Gemini AI Assistant That Can Work for Days
Google Cloud has unveiled a universal work assistant for businesses called Gemini agent. Announced at the Gemini at Work 2026 event, the tool accepts tasks through a single window and carries them out across various apps and devices. Users do not need to explain each step separately. They simply need to specify their desired result. The agent can work in the background for hours or even days.
The agent works in the Gemini Enterprise app, Google Workspace services and other platforms. Users can assign it tasks by typing @Gemini in Gmail, Drive, Docs, Sheets, Slides, Calendar and Chat. The tool can also be accessed through Slack, Microsoft 365, the command line, mobile devices, computers and the web. The API makes it possible to connect the agent to other apps without the need for a separate management screen. Gemini agent answers questions, works with documents, creates media materials, writes code and runs that code. For example, it can recognize that a manager wants the latest project report as a presentation and take on the task with a single tap. The agent identifies meeting participants from previous correspondence, checks their calendars and starts an email discussion to select a suitable time. External participants can also be included in this process. In addition to working as a personal assistant, the tool can serve as a team member or as an agent performing a specific role. Such an agent can be given a separate identity and an @agents.company.com address. The system selects the appropriate model for the task on its own and currently uses the Gemini and Claude model families. The company plans to add proprietary company models and open models in the future. For complex work, the agent uses specialized sub-agents that work on individual components. It coordinates the work they perform simultaneously or in sequence. Because the system runs in the cloud, data, the content of previous conversations and personal settings are retained when switching devices. The memory system retains the current session, information gathered from documents and conversations, work rules and previous steps. This allows tasks lasting for days to be completed without losing previous information. Gemini agent is currently in closed testing for enterprise customers. Google said broader availability will begin soon on selected Workspace Business and Enterprise plans. The company also launched the Gemini Spark agent for regular users in May. Subscribers can use the tool to perform tasks in Workspace and on the web. However, Gemini Spark opens in a separate window within the Gemini app.
The agent works in the Gemini Enterprise app, Google Workspace services and other platforms. Users can assign it tasks by typing @Gemini in Gmail, Drive, Docs, Sheets, Slides, Calendar and Chat. The tool can also be accessed through Slack, Microsoft 365, the command line, mobile devices, computers and the web. The API makes it possible to connect the agent to other apps without the need for a separate management screen. Gemini agent answers questions, works with documents, creates media materials, writes code and runs that code. For example, it can recognize that a manager wants the latest project report as a presentation and take on the task with a single tap. The agent identifies meeting participants from previous correspondence, checks their calendars and starts an email discussion to select a suitable time. External participants can also be included in this process. In addition to working as a personal assistant, the tool can serve as a team member or as an agent performing a specific role. Such an agent can be given a separate identity and an @agents.company.com address. The system selects the appropriate model for the task on its own and currently uses the Gemini and Claude model families. The company plans to add proprietary company models and open models in the future. For complex work, the agent uses specialized sub-agents that work on individual components. It coordinates the work they perform simultaneously or in sequence. Because the system runs in the cloud, data, the content of previous conversations and personal settings are retained when switching devices. The memory system retains the current session, information gathered from documents and conversations, work rules and previous steps. This allows tasks lasting for days to be completed without losing previous information. Gemini agent is currently in closed testing for enterprise customers. Google said broader availability will begin soon on selected Workspace Business and Enterprise plans. The company also launched the Gemini Spark agent for regular users in May. Subscribers can use the tool to perform tasks in Workspace and on the web. However, Gemini Spark opens in a separate window within the Gemini app.