Google prepares Agent Mode on Gemini to tackle complex tasks

· 2 min read
Gemini

Google appears to be readying a substantial upgrade to its Gemini agent stack, with fresh build references pointing to a dedicated Agent Mode that will live as its own tab inside the Gemini app rather than sitting as a tool option inside the prompt bar. The new surface is built for tasks and workflows that go beyond a single chat turn, and the latest signals confirm earlier predictions that scheduled actions and skills would be folded directly into the agent experience.

The leaked screenshots include sample prompts that hint at the kinds of jobs Google wants this agent to own. Among them:

  1. Triaging and decluttering an inbox
  2. Preparing for meetings and generating slide decks
  3. Composing personalized news digests
  4. Working as a long-form ghostwriter
  5. Tracking recurring bills and more!
Gemini
Image by @Luke_litowitz

Skills, in this context, look closer to reusable procedural recipes the agent can pick up to handle specific multi-step jobs, while scheduled actions give it a way to run those routines on a cadence. Together, they point to a product aimed at knowledge workers, students, and anyone trying to offload repetitive Workspace and research chores to the cloud.

Gemini
Image by @Luke_litowitz

Strategically, this puts Gemini on a very similar trajectory to Microsoft's Copilot Tasks, which already separates Auto, Researcher, and Analyst-style modes from the chat thread and leans heavily on connectors, scheduling, and file generation. Google is clearly trying to close the gap on ambient, always-running assistance, and pairing that with its Workspace footprint, deep research stack, and Canvas could give the agent a credible foundation. The open question is whether Google plans to leapfrog rivals or simply match them; based on what is visible so far, the company looks set on parity rather than a category-defining move.

Gemini
Image by @Luke_litowitz

Where this lands in the broader product story should become clearer at Google I/O on May 19 and 20, where Agent Mode, skills, and the wider Gemini app refresh are all plausible candidates for the keynote.