Google has launched a Gemini agent for business that takes a goal, plans the work and carries it out, instead of just answering questions. It was announced on October 8, 2026 at a Google Cloud event, and companies get it first. Consumers will have to wait.
Here’s what it does, where it runs and what Google hasn’t said yet.
Quick facts first:
- Announced: October 8, 2026, at Google’s Gemini at Work event.
- Who gets it first: businesses. Google says consumers come later.
- Price: not stated in launch coverage.
- Runs on iOS, Android, Windows, Mac, the command line, Google Workspace, Microsoft 365, ServiceNow and Slack.
- Early testers include On, Shopify and PayPal.
Those five lines are the whole headline. The rest of this post explains what they mean for you.
What the Gemini agent for business actually does
Google Cloud CEO Thomas Kurian says you can give it “objectives, not just instructions.” So you don’t write step-by-step prompts. You say what you want done, and it works out the steps.
According to TechCrunch, the agent can use custom skills and tools, accept files, folders and project bundles, and hand pieces of a job to subagents. That’s the same pattern we covered in ChatGPT’s parallel agents for Pro users, now aimed at companies.
It also picks a model for you by default. You can choose one yourself, and Claude is on the list. Google says open-source and other private models are coming later.
Where it plugs in
The agent’s value depends on what it can reach. Google lists a wide range of business systems, plus any MCP server, which is the common standard for connecting AI tools to outside apps.
| Type | What Google says it connects to |
|---|---|
| Work apps | Google Workspace, Microsoft 365, Slack, ServiceNow |
| Dev tools | Jira, Confluence, Git, command-line interface |
| Data | BigQuery, Databricks, Postgres, Snowflake |
| Anything else | Any Model Context Protocol (MCP) server |
The MCP support matters most here. It means a company’s own internal tools can be wired in without waiting for Google to build a custom connector.
It gets its own account and a paper trail
Each agent can have its own Workspace account and email address. You can tag it, email it, share a file with it or add it to a group chat, just like a coworker.
Its actions are logged under the agent’s name, not yours. If something goes wrong, an admin can see that the agent did it. That’s a sensible choice, and it’s what security teams will ask about first.
You can follow along in a “tasks inbox.” It shows the agent’s reasoning, which subagents it called, which skills it loaded, the code it wrote and its progress.
What Google hasn’t told us yet
Launch coverage leaves some big gaps. Here’s what’s missing:
- Pricing. Google mentions spend caps, smart routing and multi-model orchestration, but no numbers.
- A general availability date. One report from Tech Insider says it starts as a private preview for select customers. TechCrunch doesn’t say that, so treat it as unconfirmed.
- Benchmarks. No performance numbers were published, so any coding or accuracy claims are unverified.
Until those arrive, you can’t judge cost or reliability. That matters, because agents that run for a long time can rack up bills fast.
Google also says it chose a business-first launch to solve “harder problems around security, scale, and performance,” per Sundar Pichai. In plain terms, powerful agents are risky, and Google wants to learn on managed customers before opening the doors wider. Our look at the Gmail AI inbox rollout to paid Workspace accounts shows the same cautious, paid-first pattern.
Frequently Asked Questions
Can I use the Gemini agent for business as an individual?
Not yet. Google says it’s starting with businesses and plans to bring the agent to consumers later, with no date given.
How much does the Gemini agent cost?
Google hasn’t said. It only mentions cost controls such as real-time spend caps for companies.
Does it only use Google’s models?
No. It picks the best model by default, and you can choose another, including Anthropic’s Claude. Google plans to add open-source and other private models later.
Is it the same as the regular Gemini app?
It’s built on Gemini, but it’s an agent that does tasks and tracks them in a tasks inbox. The consumer Gemini app will get the agent later, according to Google.
Our take
This is a serious move, and the MCP support and audit trail are the right ideas. But without pricing, a clear availability date or benchmarks, there’s nothing to buy yet. If you run IT for a company, ask your Google rep about spend caps and logging now, and test with low-risk tasks before giving any agent access to real data.
Leave a Reply