Google has announced the arrival of the “Gemini agent.”
And no, it is not a character from the new James Bond movie.
Which is a shame, all things considered.
According to Google, it is a “new universal agent for work.” Google plays with the idea that this agent can “have all of the business context” and, according to the company’s not-at-all-modest announcement, “it can be used for everything from knowledge management to answering questions, and from content creation to coding, all from a single command window.” In addition, the agent selects the most suitable model for each task.
The key to the proposition is being able to delegate a goal and receive the finished work, without having to explain every step. According to Google, the Gemini agent can plan tasks, use tools, and coordinate other specialized agents to handle complex assignments. It also allows users to schedule jobs or trigger actions when certain events occur.
Because it runs in the cloud, it can keep working for hours or days even if you close your laptop. Its memory and context persist across devices and channels, including Workspace, Microsoft 365, and Slack. The idea is to avoid having to explain the project all over again every time you switch applications.
All right, and now for an important clarification: Gemini is the agent, but the model it uses may vary. Google envisions combining its own models with Anthropic’s Claude models, and expanding that selection in the future. Why they did not rename it to avoid confusion? Your guess is as good as ours. Google focuses on explaining that this model versatility is meant to balance quality and cost: a simple task does not necessarily require the most powerful model.
To understand the business, it can access, with the appropriate permissions, platforms used by the company, such as Salesforce, Jira, or ServiceNow. Companies can also create shared instructions to teach it how to carry out their usual processes.
In Workspace, the Gemini agent works inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, as you would expect. In fact, Google gives an example: you can research market trends, build a financial model in Sheets, and prepare a presentation with the results, without having to re-explain the task between applications.
In addition, it can suggest tasks that are worth delegating to it. If you receive an email asking for a presentation on the status of a project, the system can offer the option of assigning it to the agent. It can also propose prioritizing messages based on relevance and explain why they deserve attention.
Another possibility is to create agents with a defined role within the team, with their own email, calendar, and Drive. Yes, you read that right, with their own email account. Because, of course, it is an agent and it works on its own. Or almost.
And since this is probably the most controversial part of the announcement, let us use the exact quote:
“You can also use Gemini to create a collaborator agent that works with your entire team. Simply describe the role you need and Gemini creates it. The agent gets its own Workspace account, which includes an email address, calendar, Drive, and a presence in your company directory.
Your coworkers work with it like they would with anyone else, adding it to a chat space or mentioning it with @. For example, a marketing manager can ask an event coordinator agent in a group chat to draft a prep document for a launch, which it will post in the group when it is ready.
The marketing manager could also tag the agent in a document comment, where it can suggest an edit in the document and reply in the comment thread, appearing under its own name in the version history. A collaborator agent acts with its own identity, not yours, and only sees what you share with it. Access follows the sharing and membership system your team already uses, and no external connector stores your data.”
Wow.
[I do not even want to imagine how this news will be received by teams at major tech companies where layoffs/replacements have recently taken place, something that is generating enormous controversy in the United States, for example]
To control that activity, Google is adding dedicated identities, role-based permissions, and audit logs. Agents run tasks in isolated environments, while Agent Gateway applies security policies to traffic and communications. This allows the company to define what information they are allowed to access.
There will also be spending limits by project. When the set budget is reached, the agent will pause its activity. This measure is combined with automatic model selection to tailor cost to each job.
The announcement does not specify a general release date for all of these features, nor does it explain which plans will include them. So it remains to be seen when companies will be able to use them broadly and under what commercial terms.
Image: ChatGPT
Your email address will not be published. Required fields are marked *
Δ