Since December 2025, Google has been rolling out a tool that lets any Workspace employee build and deploy AI agents on company data: no coding, no IT sign-off required. If your IT team hasn’t noticed, today is the point at which they probably should.
Google Workspace Studio reached all Rapid Release domains on February 27. Today, March 2, the rollout to Scheduled Release domains begins and will continue throughout the month. By the time it completes, every Workspace Business and Enterprise user in your organisation will have access to it.
It means any employee on a Business or Enterprise Workspace plan can go to studio.workspace.google.com, describe a workflow in plain English, and have a working AI agent live in minutes.
During Google’s Gemini Alpha programme, Workspace Studio agents already completed more than 20 million automated tasks in a single month. This week, that access extends to the wider user base.
What Does Google Workspace Studio Actually Do?
Google Workspace Studio is powered by Google Workspace Studio and lets employees build AI agents from pre-built templates or natural language prompts. Those agents run natively across Gmail, Drive, Docs, Sheets, and Chat, and connect to third-party tools including Salesforce, Jira, Asana, and Mailchimp via pre-built connectors. More technical users can extend further using Apps Script and Vertex AI. Each user can create up to 100 agents and share them with colleagues like a Drive file.
Google points to Kärcher as an early case study. The cleaning equipment manufacturer used a chain of Studio agents to automate its feature evaluation process, covering assessment, feasibility, UX review, and user story drafting, reducing drafting time by 90%. During the Gemini Alpha programme, users logged more than 20 million automated tasks in a single month.
Shadow Agents: The Governance Problem Nobody Has Solved Yet
IT teams have dealt with Shadow IT for years, then Shadow AI. Each time the pattern is roughly the same: employees find something useful, start using it without flagging it, and a compliance or security problem surfaces further down the line.
AI agents follow the same logic, with a meaningful difference. A shadow app stores data. A shadow agent acts on it, reading emails, sending messages, querying external systems, running on a schedule: often with permissions its creator set up quickly and didn’t think through.
Google’s own CISO team has laid out the scenarios: an email agent that forwards excerpts from a confidential client email to the wrong person; a marketing agent that scrapes health and financial data to enrich a CRM, falling foul of GDPR and CCPA; a financial services agent that misreads a news article and sends unsanctioned client communications at scale. The problem, in each case, was no governance framework around them.
A more routine version of this plays out when a sales manager builds a quick inbox agent, grants it Gmail access, and it ends up referencing sensitive prospect information in an outbound reply.
82% of enterprises plan to integrate AI agents within three years, and 10% are already running them, many without formal oversight. Futurum Group analyst Keith Kirkpatrick flags a gap most organisations haven’t thought through yet:
“Think of the scenario of a worker who sets up dozens of agents and then abruptly leaves the company. Proper governance and documentation must be set up to ensure that whoever takes on the role has a clear understanding of what agents have been set up.”
What Microsoft Is Doing Differently
Microsoft’s February 27 update is a useful reference point. It introduced Purview Data Loss Prevention for Copilot, blocking agent interactions when sensitive data appears in a prompt, alongside an Agent Dashboard giving IT teams visibility across every agent in their organisation, regardless of who built it.
80% of Fortune 500 companies are now running active AI agents, according to Microsoft’s Cyber Pulse report, most without clear visibility into what those agents are doing. Google doesn’t yet offer controls at the same level of granularity. For now, that falls to IT admins to manage manually.
Four Things to Do Before Studio Reaches Your Users
For organisations on Scheduled Release, there is still time to get ahead of this.
- Find out what’s already running. Rapid Release users and Gemini Alpha participants may already have agents active on live systems. An inventory is the starting point before the rollout widens further.
- Configure admin controls before employees do. Review which external apps agents can connect to, what data they can access, and who has permission to build them. The out-of-the-box settings are not configured with enterprise security in mind.
- Publish a library of approved agents. Vetted agents for common tasks, email triage, status updates, document drafting, give employees a sanctioned path and reduce the incentive to build their own ungoverned versions. Studio lets you share agents across the organisation like any Drive file.
- Assign ownership to every agent. Each one needs a named owner and documented permissions. When that person changes role or leaves, the agent should be reviewed rather than left running with no one responsible for it.
As Google’s own security team notes, blanket restrictions tend to push usage underground and make governance harder. A structured adoption approach is more effective than trying to block access outright.
Workspace Studio has real use cases, and Google’s adoption numbers suggest it’s landing well with employees. But a quiet rollout since December means many IT teams are catching up, not getting ahead. For those on Scheduled Release, this week is the window.
