The Pulse
Google Introduces Gemini as a Universal Agent for Work
Google Cloud announced Gemini, a single agent designed to handle business questions, create content and run code using company context and connected tools. The system can coordinate specialist agents, work across major workplace apps and apply enterprise identity, access and spending controls.

Google Cloud announced the Gemini agent on October 8, describing it as a single assistant for business research, content creation, questions and code. Rather than switching among separate AI tools, employees would give Gemini an outcome to pursue, with the agent using company context and connected systems to produce finished work.
“You give it objectives, not instructions.”
Thomas Kurian, CEO, Google Cloud
Kurian introduced the product in a post adapted from his keynote at Gemini at Work 2026. Google’s pitch goes beyond a chat interface: the agent can plan tasks, use tools and skills, and return results in documents, email and developer environments.
One agent, with access across work apps
Gemini can run on the web, iOS and Android, Windows and Mac computers, and through channels including the command line, Google Workspace, Microsoft 365 and Slack. It can also connect to third-party applications or run without a dedicated user interface. The company describes cloud-based execution that keeps memory and task context available across devices, including for work that continues for hours or days.
Within Google Workspace, the agent is designed to work directly in Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, carrying the same memory and controls between them. Google’s examples include coordinating a meeting using a team’s existing chat and calendar information, then researching a topic, building a financial model in Sheets and preparing a presentation. Gemini can also suggest tasks from an email and create team agents with their own Workspace accounts, calendars and Drive access.
Those team agents would act under their own identities rather than a colleague’s account. Their access follows the files and spaces shared with them. The distinction matters for organizations deciding whether an agent can take actions as an employee or operate as a separately governed team member.
Specialist agents, model choice and controls
For larger assignments, Gemini can create temporary sub-agents and coordinate their work in parallel or in sequence. Gemini can also act as a continuing “coworker” agent with a defined role, persistent storage and an email address under the company’s agent domain. The agent can select among Google’s Gemini models and Anthropic’s Claude models, according to the announcement.
Google pairs that autonomy with administrative controls. Each agent is assigned an identity and permissions, while its actions are recorded in an audit trail. Tasks run in an Agent Sandbox with its own network boundary, and traffic passes through Agent Gateway, which Google describes as a policy-enforcing firewall. Administrators can also set project-level AI spending limits; when a limit is reached, the agent pauses until work is resumed.
The announcement also extends Gemini’s role into data work. Technical teams can use natural-language requests to generate PySpark code, work with notebooks, train models and troubleshoot pipelines. Business users can request operational reports grounded in BigQuery and the company’s Knowledge Catalog, with saved queries intended to provide repeatable answers without further token costs.
Google cites enterprise use, but adoption claims come from the company
Nearly 80% of its Cloud customers use its AI products and nearly 90% of Fortune 100 companies use Gemini Enterprise, Google says. Those figures describe existing use of Google’s AI offerings; the post does not establish how many customers have adopted the newly announced agent.
The company points to customer deployments as evidence of demand. Brazil’s Bradesco cut document-review time from one hour to five minutes using Gemini Enterprise for fiscal, accounting and contract analysis, while Orange Spain has deployed more than 1,000 custom agents, Google says. Those examples are Google’s descriptions of customer results, not independent evaluations of the new Gemini agent.
Google presents the product as a way to move from separate AI tasks to delegated work that spans apps, data and coding environments. Its announcement lays out the intended reach—from Workspace and Slack to company databases—and the guardrails it says will govern that reach. The agent’s practical test is whether businesses can delegate that work while keeping its access, actions and costs within the limits they set.
Sources
- Original source · Blog post Google Cloud Blog cloud.google.com