For years, AI assistants waited for a question and then gave an answer. On Thursday, at a Google Cloud event, Google announced something different: a Gemini that takes on a goal, works through it on its own and shows up in your company directory like a new hire. The new unified agent even gets its own email address.
The short version
- Google launched a unified Gemini agent that plans tasks, uses tools and delegates to subagents
- It has its own Workspace account, with an email address and an audit trail under its own name
- It launches for businesses first, with a consumer version to follow later
- Users can pick the model, including Anthropic’s Claude, from a built-in picker
Objectives, not instructions
Thomas Kurian, CEO of Google Cloud, summed up the pitch by saying the agent can be given “objectives, not just instructions.” In practice that means you describe the outcome you want, and the agent plans the steps, loads the skills it needs and connects to your internal systems to get there.
By default it picks the best model for each job. You can override that and choose one yourself, and Google says third-party models are supported starting with Anthropic’s Claude models. Open source and private models are promised for the picker later. If you have been following how fast agent prices are moving, our breakdown of what it costs to run an agent all day on Gemini 3.6 Flash shows why model choice matters so much.
A coworker with a mailbox
The detail that stands out is identity. The agent gets its own Workspace account, as if it were just another colleague. It knows who sits on which team, what time zone they work in, who has to approve what and what is on people’s calendars. You can summon it by tagging it, emailing it, sharing a file with it or adding it to a group chat.
Just as important, every action it takes is written to an audit trail attributed to the agent, not to a person. That is a sensible answer to the question every security team asks first: who actually did this? It also matches a wider trend. Startups such as Manus now give agents their own phone numbers and wallets, as we covered when Manus raised $500 million.
Where it can work
Google says the agent can connect to the systems businesses already run. It can also talk to any Model Context Protocol (MCP) server, inside or outside the company network. You can reach it from iOS, Android, Windows and Mac desktops, the command line, and from inside several tools you may already use.
| Area | What Google named |
|---|---|
| Office suites | Google Workspace, Microsoft 365 |
| Team tools | Slack, Jira, Confluence, ServiceNow |
| Developer tools | Git, command line |
| Data platforms | BigQuery, Databricks, Postgres, Snowflake |
| Open standard | Any MCP server, inside or outside the network |
| Early testers | On, Shopify, PayPal |
Why businesses go first
Google has a big head start with companies. Sundar Pichai said Gemini has more than 1 billion monthly active users and that nearly 90% of Fortune 100 businesses use Gemini Enterprise at work. Starting with the corporate side lets Google, in Pichai’s words, solve the “harder problems around security, scale, and performance” before a consumer launch.
The catch: control and cost
Handing an agent a mailbox and access to your files raises the usual worries. Google addresses part of that with the audit trail and says new flexible spending options will help keep costs in check, including multi-model orchestration, smart routing and real-time spend caps. We have already seen how murky trust gets when agents get broad access, for example in our look at Meta’s agent that can email, shop and pay for you.
What is still unclear
- Pricing. Google talked about spend caps but the announcement did not give a per-seat price
- Consumer timing. A version for everyone is promised, with no date
- Real-world accuracy. Early testers are named, but no results were shared
What to watch next
- How admins limit it. Permissions for a non-human coworker will decide how far companies trust it
- Third-party models. Claude is first, and the promise of open and private models would make it far more flexible
- The consumer launch. Rivals such as Meta’s Muse and ChatGPT’s Dots are already chasing the same everyday users
The idea of an AI colleague with a name, a mailbox and a paper trail is neat, and it is easy to see why Google wants to make it feel ordinary. The test will be whether people hand it real work, and whether the audit trail holds up the first time something goes wrong.
Sources and further reading
- TechCrunch: Google brings agentic AI to Gemini, starting with businesses
- TechCrunch: Google ramps up its AI in the workplace ambitions with Gemini Enterprise (2025)
About this article: GeekBlog covers U.S. technology news, AI, phones, smartwatches and gaming. Every story is written and checked under our Editorial Policy. Spotted a mistake or have a story tip? Contact our editors.
