Google is taking its Gemini AI platform into a new phase by introducing an agent designed to do more than answer questions. The company says its new system can take objectives from users, plan the required steps, interact with business tools, and complete tasks on their behalf.
The announcement was made during a Google Cloud event as the technology industry continues shifting from conversational AI toward systems capable of taking action.
Instead of simply responding to a prompt, agentic AI is designed to understand a goal and determine how to accomplish it using available tools and information.
Businesses First
Google plans to introduce the new Gemini agent to businesses before expanding the technology to consumers.
That strategy reflects Gemini’s growing presence in the enterprise market. Google CEO Sundar Pichai said Gemini now has more than 1 billion monthly active users, while nearly 90% of Fortune 100 companies are using Gemini Enterprise at work.
Google sees the enterprise environment as an opportunity to address challenges around security, reliability, performance, and large-scale AI deployment before bringing similar capabilities to a broader consumer audience.
From Instructions to Objectives
One of the biggest changes is how users interact with the system.
Rather than providing a detailed list of individual instructions, users can give Gemini an objective. The agent can then determine the steps required to complete the assignment.
It can use specialized skills, connect to external tools, access company information, and coordinate multiple parts of a workflow.
Google Cloud CEO Thomas Kurian described this as a shift from simply telling AI what to do toward giving it a desired outcome.
This could allow employees to delegate more complex projects instead of using AI only for individual questions or content-generation tasks.
Connecting AI to Workplace Systems
The new agent is designed to work with a broad range of business applications and data sources.
Google says Gemini can connect with platforms including Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, PostgreSQL, and Snowflake.
It can also work with Model Context Protocol servers, allowing businesses to connect additional tools and systems inside or outside their corporate networks.
Attachments and project resources can also be included in tasks, giving the agent access to relevant files, folders, skills, and other information needed for a particular workflow.
AI Gets Its Own Workplace Identity
Perhaps one of the more notable features is that the Gemini agent can operate with its own Workspace identity.
Instead of appearing to act directly as an employee, the AI can have its own email address and organizational context. It can understand information such as teams, time zones, approval requirements, and calendars.
Employees can interact with it through email, group conversations, mentions, or shared resources.
Actions performed by the agent can also be recorded in an audit trail under the agent’s identity. This could help organizations understand what the AI did and distinguish automated actions from actions taken by employees.
Businesses Can Monitor Its Work
Google is also providing a task-focused interface where employees can follow the agent’s progress.
Users can see information about tasks being performed, specialized skills being used, subagents being delegated work, code execution, and overall progress.
The system will be available across several environments, including mobile devices, desktop computers, Google Workspace, Microsoft 365, Slack, ServiceNow, and the command line.
A Multi-Model Approach
Google is not limiting the agent to Google’s own AI models.
The system can select an appropriate model automatically, while users can also choose from available models. Anthropic’s Claude models will be among the initial third-party options, with Google planning to expand support for open-source and private models.
Google is also introducing spending controls designed to help organizations manage enterprise AI costs. Features such as model orchestration, intelligent routing, and real-time spending limits could give companies greater control over how much they spend on AI workloads.
The Next Stage of Enterprise AI
Google’s move highlights how quickly workplace AI is evolving.
The focus is moving from chatbots that provide information toward digital agents capable of interacting with software, coordinating workflows, and completing objectives.
Early testers reportedly include companies such as On, Shopify, and PayPal. If businesses adopt these systems at scale, AI could increasingly become another participant in everyday workplace operations rather than simply another software tool employees use.
For Google, putting Gemini agents into enterprise environments first could provide a crucial testing ground for building reliable and secure AI systems before bringing more autonomous capabilities to consumers.
