Your New Coworker Is Not a Person.

The company directory has a new entry, and the photograph is a cryptographic seal.

On the eighth of October, at its Gemini at Work event, Google Cloud announced the Gemini agent: a single, universal agent for work that answers questions, runs knowledge work, makes media and writes code from one prompt box. The detail that matters is smaller than the keynote. You can give it a role, and it becomes a coworker. It gets a name, an email address at agents.company.com, and a place in the company directory. Your colleagues @mention it in a chat the way they would @mention anyone.

It is easy to read that as a feature. It is more useful to read it as a hire.

The name in the directory

Google is careful about the distinction. A sub-agent is temporary and job-specific. A coworker agent is persistent: a defined role and an operational presence across days and sessions, its own address, its own storage, and access only to the context you share with it. The primary source puts the line plainly.

A coworker agent acts under its own identity rather than yours, and it sees only what you share with it.

Why an identity, not an account

This is the part worth slowing down for. An account is something a person logs into. An identity is something the system can hold responsible. Google gives every agent a cryptographically attested identity with least-privilege permissions. When it touches an outside system, that identity is propagated through OAuth. Every action is written to an audit trail and attributed to the agent, not the person who set it off. Traffic in, out and between agents passes through Agent Gateway, which Google describes as an AI network firewall: write the policy once and it applies to every agent. If you have read about agents going where they should not, this is the answer. Accountability is the product.

What is actually new

Less than the keynote implies. Google shipped agent identity, Agent Gateway, a persistent runtime and sub-agents on the twenty-second of April, inside the Gemini Enterprise Agent Platform, a rebrand of Vertex AI. Today mostly unifies those pieces behind one front-end and adds Workspace plus data and industry skills. The one sharp move is model choice: the agent runs each job on the model that fits best, across the Gemini family and Claude from Anthropic. Its agent will route work to a rival's model, and Google says so out loud.

The honest counterpoint

Google published no general-availability date and no standalone pricing, so I used the primary sources rather than the product. Broad availability is promised soon for select Workspace plans. The momentum numbers are Google's and unaudited: about 500 cloud customers who each processed more than a trillion tokens, roughly 80 per cent of Google Cloud customers using its AI, and about 90 per cent of the Fortune 100 on Gemini Enterprise. Treat those as marketing until someone independent checks them. What would change my mind: a published price, an independent audit of the agent trail, or a demo where the agent's identity is revoked mid-task and the work stops.

If you are deciding how much authority to hand an agent, and want a second pair of eyes on the permissions, email brandon@kreostudio.co.uk.