Short answer
Google Cloud has announced a Gemini agent that works from a single prompt box and goes beyond answering questions to plan and complete work. It connects to an organisation's documents, email and software tools and acts through its own account like a colleague. It launches for businesses first, with individual users to follow later.
Highlights
- The Gemini agent takes goals, plans the work, uses skills and tools, and returns finished results.
- The agent has its own Workspace account, is called by tagging or emailing it, and leaves an audit trail in its own name.
- Multi-model routing and real-time spend caps are promised to keep costs in check.
3 min readAuthor: UNIT Journal EditorEditor-in-chief: Uğur Deniz İlhan
At an event on Thursday, Google Cloud announced a unified agent that takes Gemini into the agentic age. Google's announcement calls it a universal agent for work, while TechCrunch reports that it will go to businesses first.
What does the Gemini agent do?
The agent stops being just a chat tool that answers questions and takes on the work a user gives it. According to Google, work now starts in the prompt window: the agent plans the work, uses skills and tools, connects to the organisation's business systems and brings back a finished result inside documents, the inbox and developer environments.
According to Google Cloud CEO Thomas Kurian, the agent can be given objectives, not just instructions. By default it picks the best model for the task; users can choose the model themselves, and Anthropic's Claude models can be selected from the start.
What does an account like a colleague's mean?
The agent has its own Workspace account and email address. It knows the teams in the company, their time zones, the approval chain and calendars, and can be called by tagging it, emailing it, sharing with it or adding it to a group chat. When it acts, it writes an audit trail in the agent's name rather than a person's. From a task inbox, users can follow the agent's thinking, its delegation to subagents and its progress.
Which systems does it connect to, and how will costs be controlled?
According to TechCrunch, the agent can connect to systems such as Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres and Snowflake, as well as MCP (Model Context Protocol) servers inside or outside the company network. Google says multi-model routing and real-time spend caps will keep enterprise AI costs under control.
At the event, Sundar Pichai said Gemini has more than 1 billion monthly active users and that nearly 90% of Fortune 100 companies use Gemini Enterprise.
What should teams in Turkey do?
This section is not in the sources; it is our own assessment. The sources give no country-by-country launch date or pricing, so preparation should focus on work that can be done without waiting for the product.
- Keep the agent's access to documents, email and data sources to the minimum; manage the agent like an employee account.
- Pick two or three measurable tasks, such as reporting, preparing briefs or compiling campaign data, as pilots.
- Decide in advance who will review the agent's audit trails and what the spending cap will be.
Frequently asked
- Will the Gemini agent come to individual users too?
- According to TechCrunch, Google is offering the agent to businesses first and plans to open it to consumers later. The sources give no date for this.
Sources
- Google ·
- TechCrunch ·
Follow UNIT Journal
What's new in search, AI and technology, in your feed every day.
