Google launches Gemini agent as AI giants race to build autonomous workers

# Google Cloud Launches Gemini Agent to Automate Workplace Tasks Across Business Apps

Google Cloud has introduced a new Gemini agent designed to handle workplace tasks with greater independence, expanding the company’s efforts to move artificial intelligence beyond conversational assistance and toward completing complex assignments on behalf of employees.

Announced Thursday, the system is described by Google as a “single, universal agent” that can receive an objective, develop a plan and carry out the steps required to achieve it. The launch comes as major technology companies, including OpenAI and Meta, compete to develop AI agents capable of performing increasingly sophisticated tasks with limited human intervention.

Unlike conventional chatbots that primarily respond to individual prompts, Google’s Gemini agent is designed to manage multistep assignments. According to the company, it can answer questions, conduct knowledge-based work, generate content, write code and execute it to complete specific objectives.

The agent can also connect with enterprise systems and deliver finished work directly into the applications employees already use. Google aims to reduce the need for workers to switch between separate AI tools and business software, allowing more tasks to be managed through a unified system.

The service is designed to operate across Google’s workplace ecosystem, including Gmail, Google Sheets, Slides, Chat and Calendar. It can also connect with third-party platforms such as Slack and Microsoft 365, potentially allowing businesses to coordinate work across multiple services.

Employees can access Gemini through the web, iOS and Android devices, as well as Windows and Mac computers. Its cloud-based architecture also allows the agent to continue working after users leave their computers, retaining task context and carrying out assignments that may take hours or even days to complete.

For more complicated projects, the system can create temporary sub-agents to handle different parts of an assignment. These agents can work in parallel or sequentially, with their individual results combined into a final output. The approach is intended to help Gemini manage projects that involve multiple steps, information sources or specialized tasks.

Google is also promoting the concept of “coworker agents,” which businesses can configure to operate like digital members of their teams. According to a company blog post, these agents can have their own identities and email addresses and access organizational information needed to carry out assigned responsibilities.

Such capabilities could allow businesses to delegate recurring tasks and coordinate digital workflows more systematically. However, the extent to which these agents can operate independently will depend on their assigned permissions, available information and the requirements of each organization.

Another feature highlighted by Google is the agent’s ability to select an AI model suited to a particular task. The system can choose between Google’s Gemini models and Anthropic’s Claude models, according to the company. Google says this model-selection approach is intended to improve performance while helping businesses manage computing expenses.

Security and governance are central to the product’s enterprise positioning. Google said the Gemini agent includes administrative controls, authorization and permission management, secure sandboxing and other governance features designed to help organizations oversee how AI systems interact with their data and applications.

These controls are particularly important as companies consider giving AI agents access to business communications, internal documents, customer information and operational systems. The ability to complete tasks independently creates new opportunities for productivity but also increases the importance of limiting access and monitoring automated actions.

Google’s announcement points to growing adoption of AI across its cloud customer base. In its blog post, the company said nearly 500 Google Cloud customers processed more than one trillion tokens each during the previous year. It also reported that nearly 80% of its Google Cloud customers use its AI products, while nearly 90% of Fortune 100 companies use Gemini Enterprise.

The company is extending its strategy into specialized business applications as well. Versions intended for financial and legal work are already in preview, while Google said it is planning offerings for industries including government, healthcare and retail.

These sector-specific agents could help organizations adapt AI to particular workflows and regulatory requirements rather than relying solely on general-purpose assistants. Their practical value will depend on how reliably they perform specialized tasks and how effectively businesses integrate them into existing operations.

The launch comes amid intensifying competition to build AI systems that can do more than generate text, answer questions or summarize documents. Technology companies are increasingly seeking to develop agents that can plan work, use digital tools and execute tasks across multiple stages.

OpenAI and Meta are among the companies pursuing this direction, with their own agent-related initiatives adding to the competition. Google’s approach emphasizes integration with workplace software, access to multiple AI models and the ability to keep tasks running in the cloud.

For businesses, the promise of these systems lies in reducing repetitive work and helping employees manage complex assignments more efficiently. Yet adoption will require organizations to balance automation with oversight, ensuring that agents operate within defined boundaries and deliver dependable results.

With its new Gemini agent, Google Cloud is positioning AI as an active participant in workplace processes rather than simply a tool employees consult. The next test will be whether these capabilities translate into reliable productivity gains across real-world business operations.