Google Cloud launches Gemini agent as universal AI for work across Workspace and Microsoft 365

On October 8, 2026, at its Gemini at Work 2026 event, Google Cloud introduced the Gemini agent, a single, universal agent for work. Google Cloud CEO Thomas Kurian used his keynote to present the new enterprise AI agent that can answer questions, handle knowledge work, create images and media, and write and run code. The Gemini agent is designed to plan tasks and execute them across Google Workspace, Microsoft 365, Slack and the command line, where it operates as a headless agent embedded in third party applications.

The announcement marks a significant step in the evolution of enterprise AI, moving from chat based assistants to autonomous agents that can be delegated outcomes rather than step by step instructions. As Kurian put it, users can give it objectives, not instructions, and come back to finished work. The Gemini agent runs on web, iOS, Android, Windows, Mac and the command line, and it is intended to be omnipresent across the tools that knowledge workers already use every day.

Google Cloud's new agent is not a single model but an orchestration layer. It runs each job on the model that fits best, using Gemini Flash for simple tasks, a frontier model called Argon for long horizon work, and Anthropic Claude models that are available today. The company says other private and open models will be supported in the future. This multi model approach is paired with cost controls such as Smart Routing and real time spend caps that can pause an agent when a budget threshold is reached.

The Gemini agent is currently in private preview, with wide availability coming soon for Workspace customers on select Business and Enterprise plans. Google Cloud has not disclosed pricing or general availability dates for most of the announcements, according to analyst firm Futurum Group. The launch arrives as enterprises grapple with how to deploy AI agents safely and at scale, and as Google competes to define the operating layer for AI powered work.

Key Facts

Google Cloud describes the Gemini agent as a single, universal agent for work. It can chat with users, take on assigned objectives, be scheduled, or be triggered by events. It runs persistently in the cloud with one set of memories, context and a personalization graph, so jobs that take hours or days continue even after a laptop is closed. The agent supports four memory types: session, semantic, procedural and episodic.

The agent is omnipresent across channels. It works inside Google Workspace applications including Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, as well as Microsoft 365 and Slack. It can be invoked by mentioning @Gemini in Gmail, Docs, Sheets, Slides and Chat. It also has an API that lets it operate as a headless agent, meaning it can be embedded in third party apps without a dedicated interface.

Under the hood, the Gemini agent orchestrates multiple models. Simple tasks run on Gemini Flash, while long horizon or frontier work runs on Argon. Anthropic Claude models are available today, and Google says other private and open models will come later. The agent can spin up temporary sub agents for multi step tasks and supports persistent coworker agents with their own @agents.company.com email addresses, persistent storage and Workspace accounts.

Governance and security are central to the offering. Google provides cryptographically attested agent identity, OAuth propagated permissions, action level audit trails attributed to the agent, an Agent Sandbox and an Agent Gateway that the company calls an AI network firewall. Cost controls include multi model orchestration, Smart Routing and real time per project spend caps that pause an agent when triggered. Google also introduced Knowledge Catalog, Smart Storage and a Borderless Lakehouse that can query Amazon S3 and Azure Data Lake with no variable egress fees.

Adoption metrics shared by Kurian underscore Google's enterprise traction. Nearly 90% of the Fortune 100 use Gemini Enterprise, and nearly 500 Google Cloud customers each processed more than one trillion tokens in the last year. Nearly 80% of all Google Cloud customers now use its AI products. The Gemini agent connects to a tools registry that includes Salesforce, ServiceNow, Jira, Snowflake, Databricks, Microsoft Office, Teams and any MCP server, as well as collaboration tools like Confluence and Slack. Industry specializations for financial services and legal are in preview, with government, healthcare and retail coming later.

Analysis

What this really means is that Google is betting enterprises will prefer a single orchestration layer over a patchwork of specialized AI assistants. Instead of forcing workers to switch between separate tools for writing, coding, research and scheduling, the Gemini agent aims to be the one interface that delegates tasks to the right model and the right application. That is a bold vision, and it directly challenges the fragmented approach that has defined enterprise AI so far.

The multi model strategy is pragmatic. By offering both Gemini models and Anthropic Claude models today, Google acknowledges that no single model wins every task. It also hedges against the risk that a competitor's model might be better for certain jobs. The promise of future support for private and open models suggests Google wants the Gemini agent to be a neutral conductor rather than a walled garden, though the company will naturally benefit from routing more work to its own models.

Cost controls may prove to be the most underrated feature. SiliconANGLE reported on October 8, 2026, that Google noted token prices fell about 98% since 2024 while enterprise AI volume greatly increased. Real time spend caps and Smart Routing give finance teams a way to prevent runaway agent spending, which has been a concern for early enterprise deployments. Futurum Group reported on October 9, 2026, that Google did not disclose pricing or general availability dates for most announcements, leaving some questions unanswered.

Governance is another differentiator. The Agent Sandbox, Agent Gateway and cryptographically attested identity address the fear that autonomous agents could act unpredictably or access data they should not. By giving agents their own identities, audit trails and scoped permissions, Google is trying to make it safe for enterprises to delegate real work. The bigger picture here is that trust, not raw capability, will determine which agent platforms win in the enterprise.

Why It Matters

For enterprises, the Gemini agent promises to reduce tool sprawl. Instead of buying and integrating dozens of point solutions, companies can give employees one agent that works inside Gmail, Microsoft 365, Slack and the command line. That could simplify procurement, training and security reviews. It also means the agent can act as a personal assistant or as a team member, taking on roles like project manager or financial analyst, according to 9to5Google, which reported on October 8, 2026, that Google's pitch is to give it objectives, not instructions.

The integration with Microsoft 365 and Slack is particularly notable because it places Google's agent inside rival ecosystems. That could accelerate adoption among companies that are not fully committed to Google Workspace, while also increasing competitive pressure on Microsoft's own Copilot offerings. For workers, the shift from instructions to objectives changes the nature of delegation. Managers can forward an email asking for a project update and have the agent produce a slide deck, or ask the agent to set up a meeting with regional event leads without providing names.

The emphasis on persistent memory and a personalization graph means the agent gets better over time and across sessions. That raises the stakes for data governance, because memories and context must be protected and audited. Google's inclusion of four memory types and action level audit trails shows it understands that enterprises will not delegate work to an agent they cannot inspect or control. The outcome of this launch will shape how quickly other vendors adopt similar universal agent architectures.

Next Up

Wide availability of the Gemini agent is coming soon for Workspace customers on select Business and Enterprise plans, though it is currently in private preview. Google has said industry specializations for financial services and legal are in preview, with government, healthcare and retail to follow. The company also plans to support other private and open models beyond Gemini and Anthropic Claude.

Key unknowns remain. Google Cloud Blog reported on October 8, 2026, that cost controls include Smart Routing and real time spend caps, but pricing and general availability dates for most announcements were not disclosed, according to Futurum Group. Enterprises will be watching for details on how the agent's governance framework works in practice and whether the multi model orchestration delivers the promised quality and cost savings at scale.