Google Releases Gemini Agent: AI with Its Own Email and Calendar, Ready to Be a "Coworker"

Google Cloud officially launched a universal artificial intelligence (AI) agent for the workplace on Thursday (13/2). The product, called Gemini Agent, is claimed to be able to receive instructions, execute them autonomously, and hand over the final results without the need for constant human supervision. Google Cloud CEO Thomas Kurian, at the Gemini at Work event, stated that the work paradigm now begins from a command window.

Unlike conventional AI assistants that only respond to conversation, Gemini Agent is designed to integrate directly into Google's productivity ecosystem—Gmail, Docs, Sheets, Slides, Drive, Chat, and Calendar—and to reach third-party applications such as Microsoft 365 and Slack. The agent keeps running on cloud servers even when the user closes the laptop, enabling the completion of tasks that take hours to days.

The most notable feature is what it calls coworker agents. A manager can describe a specific role, for example an event coordinator, and Gemini will create an agent with a complete Workspace account: email address, calendar, Drive storage, and even an entry in the company directory. The concept mimics moves by competitors such as Grok Bot from xAI, GPT dots from OpenAI, and the bots feature from Hermes, which introduced agents with their own identities earlier.

On the security side, Google applies an unusual approach: every agent is treated like an employee. Agents get a verified cryptographic identity, and every action is recorded in an audit trail attributed to the agent, not to a human. All agents operate inside the Agent Sandbox, an isolated environment, and their traffic passes through the Agent Gateway, described as an AI network firewall.

"We treat every agent like an employee who has clear credentials and responsibilities," Kurian said, as quoted in his presentation.

Even so, the risk of misuse remains. Last July, researchers from Accomplish AI reported that the local mode of Claude Cowork could escape the Linux virtual machine on a Mac and access files on the host computer. The incident is a reminder that containment systems for autonomous agents are not yet fully immune.

Operating costs are also a concern. AI agents consume tokens in large quantities, which form the basis for billing the service. Google claims to have added real-time spending limits; when a project reaches the limit, the agent will pause until there is manual confirmation to continue. However, the company did not detail the per-token rate or the pricing packages that will be applied.

For the Indonesian market, the presence of Gemini Agent opens up both opportunities and challenges. Local technology companies and startups that have already adopted Google Cloud services can use this agent to automate repetitive tasks, such as scheduling meetings, managing documents, or responding to customers. However, the readiness of security infrastructure and human resources capable of supervising autonomous agents remains a challenge. Personal data protection regulations also need to ensure that agent audit trails do not violate privacy.

So far, Google has not announced a general availability date or official pricing for Gemini Agent. Versions for the financial and legal sectors are still in trial, while the government, health, and retail sectors will follow later. The question is, are companies ready to give credentials and full trust to non-human entities within their organizations?