IA2 MIN

Google’s work agent spans business apps, but key features remain in early access

The business agent brings documents, data and code into one workflow with permission controls and spending caps. Background delegation and third-party model choice are not yet generally available.

Gemini at Work 2026 logo on a black background
Image: Google · material oficial

Google is presenting Gemini as a unified business work agent across email, documents, code, data and enterprise applications. It is designed to take a multi-step assignment from one conversation. The product page, however, says several prominently demonstrated features remain in early access.

Those include background task delegation, mobile and desktop access, and third-party model choice. The October 9 presentation outlines the service’s direction rather than guaranteeing every capability to every customer immediately.

01

One assignment across several tools

Google describes a workflow in which the agent gathers information, analyzes it and prepares an output using enabled business sources. Connectors provide access to applications an organization chooses to connect, including Google, Microsoft and other vendors’ tools.

The company also describes task memory and long-running jobs on cloud machines, allowing work to continue after a laptop is closed. Availability restrictions still apply, and the announcement is not an independent test of reliable completion. Our AI agents explainer distinguishes generating an answer from taking actions through tools.

Official Gemini interface suggesting work tasks
Image: Google · material oficial
02

What administrators control

According to Google, source access follows the user’s existing permissions and security-team approval. An audit trail records the agent’s actions so administrators can inspect what it accessed or executed, rather than relying solely on its final response.

Project-level spending caps can pause the agent once the budget is reached. An administrator can then choose whether to resume. A budget-triggered pause does not mean the assignment has been completed.

Before connecting services, inspect the integration’s ability to read or change information. Our guide to reviewing and revoking agent connector permissions explains the checks that still matter when several workflows share one assistant.

Thumbnail of Google’s video about AI agents and Lloyds
Image: Google · material oficial
03

Early access needs a feature-level check

Google lists sales-assisted access and a 30-day Gemini Enterprise Plus trial. Neither removes the early-access notice. US businesses should verify the edition, actual feature availability, data administration and deployment cost they need before signing up. This business announcement does not establish that all of these capabilities are included in the free consumer Gemini app.

00

The conversation starts here

Sign in with a supporter account to comment. Sign in

Nobody has commented yet. Want to go first?

YOUR NEXT ROUTE

Keep following AI, work and education

If this story interests you, these three pieces are the best place to carry on.

OPEN THE FULL TOPIC
  1. 01Claude Frontier Academy requires nomination, four days of training and a 12-week projectIA · 2 MIN
  2. 02The new Copilot wants to do the work for you. Here is what launches now and what costs extraIA · 3 MIN
  3. 03Claude can use your private files in Slack. The team may still see its answerIA · 3 MIN

KEEP READING

You may also like

FRONT PAGE