Products

Google's Gemini agent gets its own work email. The price and release date are still missing.

Google's new enterprise agent works like a coworker, with its own inbox and a log of what it did. Google hasn't said what it costs or when most companies can use it.

·4 min read
Its own inbox: cover for the Google Gemini agent story

Google announced its Gemini agent on October 8 at an event called Gemini at Work. The company describes it as one agent for the whole job: answering questions, handling knowledge work, making images and writing code, all from one prompt box. In Google’s blog post, Thomas Kurian, Google Cloud’s CEO, says you give it objectives, not instructions.

That phrase is the pitch. It’s also the part I’d test first.

Objectives, not instructions

Most AI tools still wait for a prompt and then answer it. Google says this agent plans the work, picks the skills and tools it needs, and comes back with something finished. Tasks can keep running after you close your laptop, and the agent can respond to events in your systems, not only to messages you type into a box.

TechCrunch reports that you can follow its progress in a “tasks inbox,” which shows its reasoning, any subagents it hands work to, and the skills it loads. Google says it can set up temporary subagents for multi-step jobs. That’s where watching it gets harder.

An employee with its own email address

The detail I keep coming back to is identity. In Google’s description, a coworker agent gets its own @agents.company.com address, its own persistent storage, and access only to what a person shares with it. You can tag it in a document comment, email it, or add it to a group chat. When it acts, the audit trail is attributed to the agent, not to the person who asked.

That’s the right design. If a bot edits a shared sheet, people need to tell it apart from the colleague who asked for the edit. The catch is that a company now has a new kind of account to manage. Google’s post mentions “identity and policy management” and permission controls but doesn’t show what an admin can actually set. I’d want to see that admin screen before giving the agent access to a finance folder.

The agent also picks a model for each job, and it can run on Anthropic’s Claude models alongside Gemini. TechCrunch says Claude is the first third-party option. If you read the Claude add-on story, this is the same trend from the other side: the big vendors now share the same work files.

The part Google hasn’t priced

Google’s post gives one concrete cost control. You can set a hard spend limit on a project in the Cloud Billing Console, and Gemini enforces it. That’s a real safeguard. It isn’t a price.

Release status is thinner than the launch suggests. The AI LLM News write-up says the agent is in private preview for selected customers. Google says broad availability is coming “soon” for some Workspace plans, with no date. On price, reports disagree. One says the agent is included at no extra charge where Gemini Enterprise is sold. Another says Google hasn’t announced a separate price. I couldn’t find a Google pricing page that settles it, so treat any budget number as a guess until Google publishes one. The same article also warns that the name overlaps with a consumer Gemini Agent feature from November 2025, which needs a Google AI Ultra subscription.

Watch the customer list too. Google names large banks, retailers and government users. Those examples are about Gemini Enterprise, which AI LLM News notes is a separate product from this agent. Some of those companies may be testing the agent, but the post doesn’t say which ones.

Where I land

The design is sensible, and the open questions are the ones that matter for money and risk. I like an agent that signs its own work. I don’t yet trust “it chooses the best model for the job” to mean cheaper, because nobody outside Google has seen a bill. Until a company publishes a pilot with real costs and real mistakes, I’d treat this as a product preview, not a finished coworker.

I haven’t used the agent, so I can’t check how it behaves. What I could check is Google’s own post and the reporting around it.

Sources

  1. Thomas Kurian, “Welcome to Gemini at Work 2026: Introducing the Gemini agent,” Google Cloud Blog, October 9, 2026: cloud.google.com/blog
  2. Sarah Perez, “Google brings agentic AI to Gemini, starting with businesses,” TechCrunch, October 8, 2026: techcrunch.com
  3. “Google Cloud Gemini Agent: One Agent for Work, Code & Content,” AI LLM News, October 9, 2026 (independent coverage, used for release status and the conflicting pricing reports): aillmnews.co.uk