Google’s new Gemini agent comes with an early-access catch

A campaign brief often lives across old proposals, product notes and a shared calendar. Google Cloud’s October 8, 2026 Gemini agent announcement describes an agent that can use connected business context to carry work beyond a single answer. For smaller businesses, access is still early: Google gives no date for its wider rollout. Announcement, smaller-business availability.

A job can continue after the laptop closes

Google describes cloud execution that persists across devices, with tasks, tools and reusable skills. It says the agent can produce documents or media and write or run code. Model selection is separate from the agent: Google describes orchestration across Gemini and Claude. These are vendor capability statements, rather than results we reproduced. How Google describes the agent.

That matters to a small creative team if gathering approved material is the bottleneck. Our proposed first job is a campaign draft with links back to the product notes it used, held for human review before publication. Check whether a missing fact stays visibly missing. A polished draft built on the wrong source is harder to correct than an honest unfinished section.

Access and spend need their own check

The smaller-business announcement says early access with SMB customers, with all-customer availability coming later. It describes connected tools and project-level spend caps in the Cloud Billing Console. The checked announcements do not establish a standalone agent price or personal-plan entitlement. Ask your workspace administrator which access, connected systems and charges apply before assigning a job. Availability and cost controls.

Try next: if your organization has access, scope one draft to approved sources and inspect its references before using it. For the separate reusable-instruction feature in Gemini, read our Skills coverage; a saved instruction and this business agent are different parts of a workflow.