Google is moving Gemini beyond answering questions — the new agent takes an objective, plans the work, and returns a finished result in the tools employees already use.
October 9, 2026 — Google Cloud has unveiled the Gemini agent, described by the company as a single, universal agent for work, at its Gemini at Work 2026 event on October 8. In a post adapted from Google Cloud CEO Thomas Kurian's keynote, according to the company, it can answer questions, handle knowledge work, create media, and write and run code.
Image: A central AI agent hub connected to office documents, email, calendars and code in a modern workplace. Credit: AI-generated illustration / Red Eye Eagle News, original illustration.
Editor's Highlight: The shift is not a new chat window. Google says the agent is separate from the model underneath it, routes each job to the model that fits best — including Anthropic Claude today — and can work in the cloud for hours or days.
Table of Contents
- Key Facts
- One Agent Instead of Many Separate Tools
- How the Agent Works Across Workspace and Beyond
- Memory, Sub-Agents and Coworker Roles
- Model Choice, Routing and Cost Controls
- Industry Versions for Finance, Legal and More
- Security, Governance and Agent Identity
- What Availability Looks Like Now
- Bottom Line
- FAQ
- Sources
Key Facts
| Detail | Information |
|---|---|
| What was announced | Gemini agent, a single universal agent for work |
| Announced by | Google Cloud at Gemini at Work 2026 |
| Announcement date | October 8, 2026 |
| What it can do | Answer questions, knowledge work, create media, write and run code |
| Where it works | Gemini Enterprise app, Workspace, Slack, Microsoft 365 and third-party services |
| Models it can use | Gemini and Anthropic Claude today, according to Google; other models planned |
| Industry versions | Finance and legal in preview; government, healthcare and retail coming soon |
| Availability | Private preview now; wider availability soon for select Workspace plans |
One Agent Instead of Many Separate Tools
Google presents the Gemini agent as one place to assign work, rather than separate assistants per app.
According to the company, users give it objectives rather than step-by-step instructions. The agent plans the work, uses skills and tools, connects to business systems, and returns finished work in documents, inboxes and developer environments.
Newsroom Focus: This pitch is about completing tasks, not just generating text. The test is whether it reliably finishes multi-step jobs unsupervised.
How the Agent Works Across Workspace and Beyond
Inside Google Workspace, the agent works in Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, according to the company.
Reports describe a proactive mode: if a manager asks by email for a project update as a slide deck, the system can offer a one-click handoff to Gemini.
Access is not limited to Workspace. Google says the agent is available from any device, including web, mobile and desktop, the command line, Slack and Microsoft 365, and can run headless inside third-party apps via an API.
Memory, Sub-Agents and Coworker Roles
Because it runs in the cloud, Google says the agent keeps one set of memories and context across devices. Work can continue after a user closes a laptop.
It lists four memory types: session for the current task, semantic for knowledge built from work, procedural for how jobs get done, and episodic for past work.
For complex work, it can create sub-agents and coordinate parallel and sequential steps that, according to Google, can run for hours or days.
It can also act as a coworker, with its own Workspace identity, email address, calendar, Drive storage and directory place, reports say. It sees only what is shared with it.
Analysis Note: Its own identity matters for accountability — actions are logged under the agent, not a human employee.
Model Choice, Routing and Cost Controls
Google stresses that the agent and the model are separate choices. The agent runs each job on the model that fits best, according to the company.
Today that includes the Gemini family and Anthropic's Claude family, with other private and open models planned, according to reports.
Cost controls include multi-model orchestration, Smart Routing to a suitable lower-cost model, and real-time spend caps set per project.
Practical Takeaway: For IT teams, routing and spend caps may matter as much as model quality for keeping bills predictable.
Industry Versions for Finance, Legal and More
Alongside the universal agent, Google announced industry-specific agents for financial services, legal, government, healthcare and retail.
The financial services and legal versions are now in preview, according to the company, while the other three will be released soon.
Each version adds specific skills and tools. For financial services, Google says it automates investment research, credit analysis and risk modelling, using data including FactSet, LSEG and SEC filings.
Security, Governance and Agent Identity
Google says every agent has its own identity, governed with least-privilege, role-based permissions approved by administrators.
Reports describe that identity as cryptographically attested, with the agent's work tracked in logs. Tasks run inside an Agent Sandbox with its own network boundary.
Traffic passes through an Agent Gateway, described as a firewall enforcing organisation policies in real time.
What Availability Looks Like Now
This is not a consumer launch. Reports say the Gemini agent is currently in private preview for enterprise customers.
Wide availability is expected soon for select Workspace Business and Enterprise plans, according to 9to5Google. No date or pricing was announced in the reports read.
Google cited scale for its existing platform: according to the company, nearly 80% of Google Cloud customers use its AI products and nearly 90% of Fortune 100 companies use Gemini Enterprise — Google's own reported figures.
Bottom Line
The Gemini agent is Google Cloud's attempt to turn an assistant into a persistent worker: one agent, with memory, tools and a governed identity, that can be given an objective and left to finish the job.
What to watch: preview performance in real companies, wider availability and pricing, and whether routing to rival models such as Claude lasts.
FAQ
What is the Gemini agent?
A single, universal agent for work announced by Google Cloud on October 8, 2026, for questions, knowledge work, media creation, and writing and running code, according to Google.
How is it different from a normal chatbot?
Google says users give it objectives, not just prompts. It plans steps, uses tools and sub-agents, and can keep working in the cloud for hours or days.
Which models does it use?
According to Google, it routes jobs across the Gemini family and Anthropic's Claude family today. Other private and open models are planned.
Where can employees use it?
In the Gemini Enterprise app, across Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, plus Slack, Microsoft 365 and third-party apps, according to Google.
Can anyone sign up now?
Not generally. It is in private preview for enterprise customers, with wider availability expected soon for select Workspace plans. No pricing was announced.
Sources
- 9to5Google — Gemini agent announcement report: https://9to5google.com/2026/10/08/gemini-agent-google-cloud/
- PYMNTS — Universal agent and industry versions report: https://www.pymnts.com/news/artificial-intelligence/2026/google-cloud-targets-enterprise-market-with-universal-agent-work/
- Neowin — Gemini at Work 2026 unveiling report: https://www.neowin.net/news/google-cloud-unveils-universal-gemini-agent-at-gemini-at-work-2026/
- Unite.AI — Detailed announcement and architecture report: https://www.unite.ai/google-cloud-unveils-gemini-its-universal-agent-for-work/
- Hero image — Original AI-generated illustration created for Red Eye Eagle News for this article.
Join the conversation
Comments
Post a Comment