Google is turning Gemini from a chatbot into a workplace agent designed to carry out assignments across business systems. TechCrunch reported on October 8 that the company unveiled the unified agent at a Google Cloud event, describing a single interface where users can ask questions or hand over work for Gemini to plan and execute.

The initial rollout is aimed at businesses rather than consumers. Google CEO Sundar Pichai said Gemini has more than 1 billion monthly active users and that nearly 90% of Fortune 100 companies use Gemini Enterprise, according to TechCrunch. Google said beginning in the corporate market will let it address security, performance and scale before bringing the agent to consumers.

An AI core delegates parts of a complex assignment across protected business modules.
Gemini can split work among subagents and connect to a range of corporate tools and data sources.

Google Cloud CEO Thomas Kurian said the system can be given objectives instead of only step-by-step instructions. The agent can plan a job, call specialized skills and tools, and connect to a company’s internal systems. By default, it selects a model for the task, though users can make the choice themselves. The initial third-party option is Anthropic’s Claude, with open-source and other private models planned for later.

The range of connections is central to Google’s pitch. TechCrunch reported that Gemini can work with data and tools including Google Workspace, Microsoft 365, Slack, Jira, Confluence, Git, BigQuery, Databricks, Postgres and Snowflake. It can also connect to Model Context Protocol servers located inside or outside a company network. Those integrations could give the agent broad reach, but the report does not provide independent testing of how it handles permissions or mistakes across them.

The agent can break work into smaller jobs and delegate them to subagents. A tasks inbox shows users its progress, the skills it loads, its code, and its delegation activity, TechCrunch reported. This design appears intended to give workers a place to supervise long-running tasks instead of treating each interaction as a single chat response.

A distinct AI identity travels through workplace communication and approval paths while leaving an audit trail.
The enterprise agent receives its own workplace identity, separating automated actions from human activity.

One of the most consequential design choices is that the agent receives its own Workspace account. Google said it can have an email address and organizational context, including team membership, time zones, calendars and approval relationships. Workers can tag it, email it, share material with it or add it to group chats. Actions are recorded in an audit trail under the agent’s identity rather than a person’s, creating a clearer distinction between human and automated work.

Google plans to make the agent available through iOS and Android devices, Windows and Mac computers, a command-line interface, Google Workspace, Microsoft 365, ServiceNow and Slack. TechCrunch named On, Shopify and PayPal among early testers, while Google listed several large Gemini Enterprise customers. The article did not report evaluation results from those testers, so the announcement establishes scope more clearly than real-world reliability.

Google also announced spending controls that include multi-model orchestration, smart routing and real-time caps. Those features reflect a practical issue for companies: an agent that can choose models, call tools and run multiple steps can also create variable costs. The launch therefore combines ambitious autonomy with administrative controls, but its broader impact will depend on whether companies can safely grant an AI account enough access to do useful work without losing oversight.