Google Cloud announced on October 8, 2026 a universal Gemini agent for the enterprise workplace that can run for hours or days on a single assignment while keeping its memory, business context and identity intact across applications and devices. Each Gemini agent gets its own Gmail address, calendar, Drive storage and seat in the company directory, so colleagues can delegate work to it the way they would to a human teammate, according to VentureBeat, which first reported the launch from the company's Gemini at Work 2026 event.

The Gemini Agent runs persistently in the cloud and maintains four kinds of memory: session memory for the task in front of it even when it runs for days, plus semantic, procedural and episodic memory, according to AI Weekly's coverage of the announcement. The Gemini Agent works directly inside Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar, and reaches into Microsoft 365 and Slack from web, mobile and desktop, so employees can stay inside familiar apps instead of switching to a dedicated interface. In a launch post, Google Cloud CEO Thomas Kurian described the Gemini Agent as a single, universal agent for work, arguing that work now starts in the prompt window, as reported by AI Weekly.

Access is deliberately scoped. A coworker agent acts under its own identity rather than the user's, and it sees only what is shared with it, the Gemini Agent product team said. Agents can run for up to seven days with persistent memory and spawn temporary sub-agents, each carrying scoped cryptographic identities and permissions, according to Particle's summary of the launch. Private preview is available to a select group of customers, with wider availability promised soon but no specific date given. Google says the Gemini Agent will be included at no additional charge for customers where Gemini Enterprise is available, including through existing Google Workspace business and enterprise subscriptions, as reported by VentureBeat.

One agent, many models

The architecture is notably model-agnostic. Google routes jobs across more than 200 models, including its own Gemini family — Argon for frontier reasoning, Flash for speed, Omni for generative media and Gemma for lightweight edge workloads — plus Anthropic's Claude today, with other proprietary and open-weight models planned, according to AI Weekly. A Smart Routing tool picks which model handles each job, and administrators can cap a project's AI spend in the Cloud Billing Console. That flexibility positions Google's governance and tooling, rather than a single flagship model, as the differentiator against single-model lock-in.

For enterprise buyers, the pitch is that AI graduates from a collection of assistants into a platform that takes on entire assignments. Rather than managing separate agents for different tasks, an employee delegates an objective to one Gemini Agent that figures out how to complete it, accesses the necessary enterprise systems and returns finished work, VentureBeat reported. Industry-specific versions for financial services and legal work are available now, with other sector packs on the way, according to Particle. CBS News also covered the launch, noting the agent is designed to write and run code while automating everyday workplace tasks.

The enterprise agent race heats up

Google's move lands days after Microsoft expanded agentic AI across Dynamics 365, embedding autonomous sales and service agents into Teams, Outlook and Copilot — a rivalry that now spans both productivity suites. Both vendors are converging on the same idea: the agent that wins is the one living inside the workflow employees never leave, whether that is Google Workspace or Microsoft 365. Google's edge, for now, is the universal design: a single AI agent with its own directory identity rather than dozens of single-purpose bots.

The mailbox idea is the detail that could matter most. Giving an agent its own email address and calendar entry turns it into a visible, auditable participant in workflows instead of a background integration — a design choice that may do more for enterprise trust than any benchmark. Google likened eventual deployment of the Gemini Agent to a software version update, according to VentureBeat. Pricing and a general-availability date were not announced.

Sources: VentureBeat, AI Weekly, Particle