Most Alberta businesses aren’t buying an enterprise work agent this month, and Google isn’t selling one broadly yet. Still, the Gemini agent shows where office AI is heading: you state an objective, and software plans the steps, picks a model, works across your tools and reports back. The useful question for a small or mid-sized team isn’t whether it’s impressive. It’s what you would need to know before handing it your inbox.
What Google announced
“Today at Gemini at Work 2026, Google Cloud announced the Gemini agent, a universal agent for work.”Google, Google Cloud introduces the Gemini agent, October 8, 2026
In the announcement, Google says you give the agent objectives rather than instructions. It describes persistent execution in the cloud, so work that takes hours or days keeps running after a laptop closes, and sub-agents that handle multi-step jobs. The Google Cloud post says the agent can act as a personal assistant, as a team member such as a project manager, or in a role such as a finance analyst. It adds that a coworker agent gets its own identity, including an @agents.company.com email address and its own storage, and sees only the context that people share with it.
The same post lists channels including the command line, Google Workspace, Microsoft 365 and Slack, and tools it can connect to such as Confluence, Salesforce, ServiceNow, BigQuery and Snowflake, plus Model Context Protocol servers. It says the agent works inline in Gmail, Drive, Docs, Slides, Sheets, Chat and Calendar. These are Google’s descriptions of intended capabilities. We haven’t tested any of them.
The controls Google highlights
| Area | What Google says |
|---|---|
| Identity | Every agent gets its own identity, cryptographically attested and governed like an employee, with least-privilege permissions |
| Permissions | Fine-grained, role-based access approved by security administrators; identity propagated to external systems through standards such as OAuth |
| Auditing | Every action is written to an audit trail and attributed to the agent, not a person |
| Policy | Agents run in an Agent Sandbox; traffic passes through Agent Gateway, which enforces organization policy |
| Cost | Smart routing across models, and real-time spend caps per project that pause the agent when triggered |
What isn’t stated
Neither Google page we read gives a price, a general availability date or a list of regions. The Google Cloud post does say its industry versions for financial services and legal are in preview, with government, healthcare and retail coming soon, but that describes the industry packs. 9to5Google reports the Gemini agent itself is in private preview, with wider availability “soon” for Workspace customers on select Business and Enterprise plans. We are relying on that report for the status, and Google’s pages we read don’t confirm it. In short, there is nothing to buy yet.
What to ask Google before joining (our advice)
What follows is our advice, not Google’s. If you’re invited to a preview, or are choosing between agent platforms, ask four things in writing. First, identity and permissions: what exactly can an agent open, and who approves it? Second, data residency: where are prompts, files and memory stored and processed, and is Canada an option? Third, audit logs: can you export them, and for how long are they kept? Fourth, spend caps: are limits enforced per project, and what happens to work in progress when a cap is hit? Also ask which model providers see your data when the agent routes a job, since Google says it routes across Gemini and Claude.
Meanwhile, do the homework that doesn’t depend on any vendor. List the workflows you’d hand an agent: lead follow-up, meeting scheduling, weekly reports, quote preparation. For each, write down the data it touches, who should approve its output and what a bad result would cost. For a contrast, our earlier briefing on Cohere North 2 covered a private, air-gapped agent platform with cost caps, a very different deployment model from a cloud-run agent. Our AI opportunity scan is a quick way to find the first workflow, and AI consulting can help you write the questions for any vendor.