Google launches Gemini AI agent built for workplace teams

By Splitfeed EditorialLast updated Oct 8, 2026, 16:30 UTC

On October 8, 2026, Google launched what it describes as a "universal" Gemini AI agent built for workplace teams. Unlike earlier AI assistants designed for individual users, this agent can be assigned to a group of people, is given its own email address, and is capable of writing code and executing tasks autonomously. The launch was framed as a direct move into enterprise AI, where Google now competes openly with OpenAI and Microsoft. A separate but related development noted the same day is a Gemini setting that could grant the AI broad access to users' Mac systems, suggesting that device-level permissions are being expanded alongside the workplace product.

The group-assignment model is the detail that sets this apart from most prior AI agent announcements. Rather than augmenting a single worker, the agent is designed to sit inside a team's workflow, receiving instructions from multiple people and acting with meaningful autonomy. That raises questions about accountability, since tasks run by a shared agent are harder to attribute to any one decision-maker. The Mac access setting compounds this, pointing to a trajectory where AI tools are not just answering questions but operating across file systems and applications in ways that go well beyond a chat interface. The competitive framing against OpenAI and Microsoft also signals that enterprise contracts, not consumer subscriptions, are now the main battleground.

No reactions from the Pro-AI, Anti-AI, or Middle Ground camps have been published yet. Typically, Pro-AI voices would emphasize the productivity gains from an agent that can handle coding and task execution at the team level, arguing this reduces friction and lets workers focus on higher-order decisions. Anti-AI voices would likely raise concerns about data access, especially given the Mac permissions angle, and question whether autonomous agents operating across shared team environments introduce unacceptable risks around privacy and error propagation. Middle Ground observers would probably call for clearer governance rules around shared agents, including audit trails and explicit controls over what the agent can access and on whose authority.

The immediate things to watch are how Google prices and restricts the agent within Google Workspace tiers, what specific permissions the Mac access setting actually requires, and whether OpenAI or Microsoft respond with competing announcements in the near term. Any enterprise security review or early customer case study would also help clarify whether the autonomous task execution holds up in regulated industries.