Google Cloud Unveils Gemini, Its Universal Agent for Work

Google Cloud announced the Gemini agent on October 8, 2026, at its Gemini at Work 2026 event, describing it as a single, universal agent for work that handles knowledge work, question answering, content creation, and coding from one prompt box.

In the announcement on The Keyword, Google said the agent holds an organization’s business context, plans the work, uses skills and tools, connects to Cloud customers’ business systems, and returns finished work inside the documents, inboxes, and developer environments people already use. The company said the agent selects the best model for each job, includes built-in cost controls, and carries the security, administration, and governance enterprise customers require. A detailed post on the Google Cloud Blog, adapted from Google Cloud CEO Thomas Kurian’s keynote address at the event, lays out the agent’s architecture, access surfaces, and enterprise controls.

Kurian’s post reports that nearly 500 Google Cloud customers each processed more than one trillion tokens over the preceding year, that nearly 80% of all Google Cloud customers are using its AI products, and that nearly 90% of the Fortune 100 use Gemini Enterprise.

The post names early testers and customers it says are already using the new capabilities. According to the post, sportswear brand On tested the agent’s dynamic model-selection capability, Shopify blends frontier models for millions of merchants, and PayPal routes 10 million multi-model requests every week. It also reports that Brazilian bank Bradesco cut document review time from one hour to five minutes and that telecommunications operator Orange Spain has deployed more than 1,000 custom Gemini Enterprise agents.

Agent Architecture

The post describes an agent built around a set of architectural principles. As a unified agent, Gemini answers questions in chat, works autonomously on objectives users assign, and generates code, all from a single interface and a single API; users can assign work, schedule tasks, or have the agent respond to events. Access spans the web, iOS and Android devices, Windows and Mac desktops, the command line, Google Workspace, Microsoft 365, and Slack, and the agent can also run headless inside third-party applications without a dedicated interface. Because it runs persistently in the cloud, the agent maintains one set of memories, context, and a single personalization graph across devices, and work that takes hours or days continues after a user closes a laptop.

For multi-step tasks, Gemini can dynamically create temporary sub-agents, each with its own identity, and coordinate parallel and sequential steps that can run for hours or days. A coworker mode gives the agent a persistent, defined role: coworker agents receive their own @agents.company.com email addresses and persistent storage, and they have access only to the context users or team members provide. Google says the agent is separate from the model underneath it, and that Gemini currently orchestrates across its Gemini model family and Anthropic’s Claude models, with other private and open models planned, matching the model to each task for quality and cost.

Tools, Skills, and Memory

According to the post, three capabilities give the agent its business context. A tools layer connects Gemini to collaboration software such as Confluence, Microsoft Office, Teams, Slack, and Workspace; development tools including Git and Jira; enterprise platforms such as Salesforce and ServiceNow; databases including BigQuery, Databricks, Postgres, and Snowflake; files on a user’s own desktop; and any Model Context Protocol server inside or outside the company network. An enterprise tools registry lets teams build and publish tools for the rest of the company.

Skills are reusable sets of instructions, knowledge, or workflows stored as modular prompts; Gemini ships with a global skills library, teams can publish custom skills to a shared company registry, and individuals can build personal skills. The post says the agent keeps four kinds of memory — session, semantic, procedural, and episodic — and that it onboards itself the way a new hire would, learning the user, the tools, and the team before starting work.

Gemini in Google Workspace

Gemini also works inline inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, carrying the same memory, skills, and controls it has elsewhere. The post describes three Workspace modes. In a personal-assistance mode, the agent arrives already briefed on a user’s calendar, team, and projects. In a proactive-delegation mode, Workspace Intelligence recognizes delegatable tasks and offers a single-click handoff. In the coworker mode, the agent receives its own Workspace account with an email address, calendar, Drive, and a presence in the company directory. A coworker agent acts under its own identity rather than the user’s, sees only what is shared with it, and follows the sharing and membership settings a team already uses.

Data Skills and Industry Specializations

New data skills let engineers describe an outcome in plain language while Gemini generates PySpark code, supplies notebooks to edit and test it, trains models, and troubleshoots pipeline issues, according to the post. Business users can generate real-time operational reports by asking, through skills integrated with BigQuery and the Knowledge Catalog, and saved reports can be rerun on demand without incurring token costs.

The post names three capabilities that keep those answers grounded: Knowledge Catalog, which maps business definitions once so all agents use them and reads metrics where they sit in Databricks, dbt, LookML, or SAP; Smart Storage, which enriches unstructured objects in place; and a Borderless Lakehouse that queries Amazon S3 and Azure Data Lake with no variable egress fees and reads Salesforce Data 360, SAP, ServiceNow, and Workday without copying data.

Industry specializations are in preview for Financial Services and Legal, with versions for Government, Healthcare, and Retail listed as coming soon. Google said the Financial Services version draws on data from FactSet, LSEG, SEC filings, and a customer’s own proprietary repositories, ships with more than 50 foundational skills, and is already in use at CME Group and Deutsche Bank. The Legal version inherits matter-level permissions and ethical walls from the NetDocuments and iManage document management platforms; Google names Harvey and Onit as partners and says Cooley is building a confidential-information redaction agent on the Legal version.

Security, Governance, and Cost Controls

Under the governance model, every agent receives its own cryptographically attested, least-privilege identity that is stamped into the logs capturing its work, along with fine-grained, role-based permissions approved by an organization’s security administrators; when the agent connects to an external system, its identity is propagated through OAuth. Every action the agent takes is written to an audit trail attributed to the agent rather than to a person. All Gemini agents execute inside an Agent Sandbox with its own network boundary, and all traffic passes through Agent Gateway, which the post describes as an AI network firewall that enforces organization-wide policies in real time.

Cost controls include multi-model orchestration; a Smart Routing tool that automatically triages enterprise workloads so each runs on the model that delivers maximum performance at the lowest possible cost; and real-time spend caps set per project in the Cloud Billing Console, which pause a project’s agent when triggered. Google said its latest TPU 8i system delivers 80% better price-performance than the prior generation, and it named its model lineup as Argon for frontier reasoning, Flash for speed and volume, Omni for generative media, and Gemma for lightweight, open-weights edge workloads.