Google launches Gemini as autonomous workplace agent with subagent delegation

The detail worth flagging is the 'distinct workplace identities' capability, which suggests Google is building persistent agent personas that operate with delegated authority inside enterprise systems, not just one-off task runners. That's an identity and access management question as much as an AI one, and it's largely absent from the coverage so far.

This move lands almost simultaneously with Google's Verge-covered universal work orchestrator announcement from October 8, and the two stories together reveal a coordinated push: Google is not iterating on Gemini as a chat product but repositioning it as enterprise infrastructure. That framing puts it in direct competition with what we covered in OpenAI's Dots launch from early October, where the same thesis played out: the durable advantage goes to whoever owns the enterprise integration layer, not the underlying model. The MIT Technology Review piece from October 5 on connecting agents to enterprise knowledge adds a useful constraint here, because multi-system orchestration is only as good as the contextual knowledge the agent can actually reason over, and neither Google nor OpenAI has fully answered that architecture question in public.

Watch whether Google publishes concrete integration counts (named enterprise software partners, not categories) within the next 60 days. If the list stays vague, the 'cross-system execution' claim is positioning rather than a shipping product.