With Workspace Studio, Google wants workers to build their own AI agents

Google this week launched Workspace Studio, promising to let a wide range of employees build and use their own AI agents. 

Workspace Studio — it was called Workspace Flows during preview earlier this year — is a no-code application that lets users create and customize agents with natural language descriptions and muti-step actions. 

“Studio puts the full potential of agentic AI into the hands of everyone, not just specialists, by removing the friction of coding and making it easy for anyone to design agents that automate their unique business processes in minutes,” Farhaz Karmali, product director at Google Workspace Ecosystem, said in a blog post.

Agents can perform tasks such as drafting weekly project updates, Google said, or notifying a user in Google Chat when they receive an email on a certain topic. 

Studio combines Google’s Gemini 3 language model with rules-based automation to enable agents to reason and adapt to new information. This could mean flagging emails where the agent identifies a negative tone from the sender, for instance.

Studio agents can access data in Google Workspace apps such as Gmail, Drive and Sheets, as well as from the web and third-party apps including Asana, Jira, and Salesforce.

A catalogue of prebuilt agents is available in the Workspace Studio app that can be customized to a user’s needs. The tool will roll out to customers “over the next few weeks,” the company said. 

Google places a limit on the number of agents customers can create (up to 100) and how many can run each day. Agents are allowed a maximum of 20 steps. Workspace customers will get “promotional access” to higher usage limits during the initial launch period, with additional details coming with a future update in 2026.

Google is not alone in developing simple agent-builder tools; Microsoft’s Copilot Actions provides similar functionality.

As AI adoption continues to grow in the workplace, organizations are investigating the potential of agents to solve common tasks in systematic ways, said J. P. Gownder, Forrester vice president and principal analyst. “With an agent, an employee doesn’t have to write yet another prompt to give the AI directions. And, in theory, agents are more action-oriented than what you get from prompting,” he said.

While there’s interest in agentic AI among tech-savvy employees, “creating agents is beyond the skill level of most employees today,” said Gownder. He cited to a 2024 Forrester survey that showed only 26% of employees said they know what prompt engineering is and how to use it. 

“In 2025, that number didn’t budge — it was again 26%,” he said.

The survey also found that most (58%) employees have received no formal training on how to use AI at work. “Given this lack of skills, the prospect that most employees will be able to create AI agents is premature,” he said.

Assuming adoption does grow, “agentic sprawl” could be become a headache for IT teams tasked with managing them. “Google attempts to deal with this by giving IT management tools for agents,” Gownder said. “It remains to be seen how much time and staffing will be required to use these tools, however.

“Ultimately, Google is moving in a smart direction by introducing agentic tools into Workspace. It’s just that we can expect a need for hand-holding and curation from IT teams for the next few years.”