Google Unveils Unified Gemini Agent to Complete Tasks Across Workplace Apps

TL;DR

• Google announced a unified Gemini agent that can plan and complete work from a single interface.

• The rollout will initially target businesses, with consumer access planned later.

• The agent can connect to workplace apps, internal data and Model Context Protocol servers.

• Users can let Gemini select a model automatically or choose alternatives, initially including Anthropic’s Claude.

Google announced a unified Gemini AI agent on Thursday, expanding its workplace offering beyond answering questions to planning and completing tasks on users’ behalf.

Unveiled at a Google Cloud event, the agent can use tools, access business systems and coordinate work through a single interface. Google will introduce it to businesses first before expanding to consumers.

The launch reflects growing competition among AI providers to build assistants that can take responsibility for workflows rather than simply respond to prompts.

Google Puts Businesses First in Gemini Agent Rollout

Google CEO Sundar Pichai said Gemini now has more than 1 billion monthly active users, while nearly 90% of Fortune 100 companies use Gemini Enterprise.

That corporate adoption gives Google an established audience for its new agent. Pichai said starting with businesses would help the company address demanding security, scale and performance requirements.

Early testers included sportswear company On, Shopify and PayPal. Google also named BNP Paribas, Bradesco, Merck, Orange Spain, Santee Cooper, SOMPO, Ulta Beauty and Wesfarmers among Gemini Enterprise customers.

Gemini Can Work Toward Objectives Across Business Systems

Google Cloud CEO Thomas Kurian described the agent as capable of receiving “objectives, not just instructions.”

The agent can plan its work, load specialized skills and use tools connected to a company’s internal systems. Users can supply files, folders or projects containing materials and skills relevant to a particular workflow.

Supported connections include Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres and Snowflake.

Google said the agent can also securely work with Model Context Protocol, or MCP, servers both inside and outside a company’s network, extending its access to additional tools and services.

Model Selection Includes Anthropic’s Claude

By default, Gemini will select the model it considers best suited to a task. Users can also choose a model themselves.

The initial third-party options include Anthropic’s Claude models. Google plans to expand the selection to open-source models and other private models.

That approach allows businesses to use multiple model providers through the same agent interface, with automated routing available when users do not want to make the selection manually.

Dedicated Workspace Account Creates an Agent Audit Trail

The agent will have its own Google Workspace account, including an email address and organizational context.

Google said it can use information about teams, time zones, calendars and approval responsibilities to coordinate work. Employees can involve it by tagging it, sending an email, sharing materials or adding it to a group chat.

Actions will generate an audit trail attributed to the agent, helping distinguish its activity from work performed by employees.

A tasks inbox will let users monitor progress, including tool activity, code, specialized skills and work delegated to subagents.

Google Adds Broad Access and Spending Controls

The agent will be accessible through iOS and Android devices, Windows and Mac computers, and a command-line interface. It will also be available through Google Workspace, Microsoft 365, ServiceNow and Slack.

Google announced flexible spending options alongside the launch, including multi-model orchestration, smart routing and real-time spending caps.

Those controls aim to help businesses manage costs as agents perform increasingly complex tasks across their software and data systems.