Gmail’s AI Inbox could soon let Gemini finish your to-dos for you

Gmail’s AI Inbox could soon do more than tell you what needs doing. Code in a recent Gmail build suggests Google is preparing to let Gemini pick up those suggested to-dos and work through them itself, while still asking you to sign off before it sends anything or ticks anything off.

The evidence comes from an Android Authority teardown of Gmail for Android version 2026.09.28, which followed a post on X by Abhishek Yadav flagging the work. Google hasn’t announced the feature, and strings in an app’s code indicate intent rather than a release date. Features spotted this way can change or never ship.

How Gemini tasks in AI Inbox would work

AI Inbox already scans your mail and surfaces suggested actions, such as a bill to pay or a form to return. According to the strings, each of those items could gain a “Start with Gemini” button. Tap it, and the task gets an “In progress” label while Gemini handles it in the background. Items that only need a response may show a “Draft Reply” option instead.

The more interesting part is how much control stays with the user. Tasks where Gemini lacks information or needs approval would be marked “Needs your input”, with a “Give input” button to fill the gaps. Finished work lands under “Ready for Review”, where you can check what Gemini did and approve or reject its changes. Based on the code, Gemini won’t send an email or mark a task as complete without a human confirming it first.

That approval step is a sensible guardrail for an assistant with access to your inbox, where one confidently wrong reply to your landlord or your boss is a real cost rather than a funny screenshot.

Where AI Inbox stands now

Google introduced AI Inbox in January as part of a broader Gemini push in Gmail, initially for trusted testers in the US. At I/O in May, it expanded to Google AI Plus and Pro subscribers in the US, having already been available to AI Ultra subscribers and in preview for Workspace Enterprise Plus customers. That update added personalised draft replies, links to relevant Docs, Sheets and Slides files next to each to-do, and the option to mark tasks as done or dismiss unhelpful suggestions.

Agentic tasks would be the next logical step. Drafting a reply is helpful; having Gemini carry a task through to the point where you simply approve it is a different level of delegation. It also fits Google’s wider direction. The same I/O event introduced Gemini Spark, a 24/7 agent in the Gemini app that can take actions across Workspace apps, which has since picked up Google Photos workflows.

Availability in the UAE

This is where expectations need managing. As of mid-September, AI Inbox was still limited to paying Google AI subscribers in the US. Google has been widening other Gmail AI features: AI Overviews in Gmail search went global in September for paid plans with Gmail set to English, which includes the UAE. AI Inbox itself hasn’t followed yet, so any Gemini task feature would likely reach US subscribers first.

Google has also been adding conversational voice features to Gmail, Docs, and Keep, so It’s clearly building out Gmail’s AI features quickly. Whether agentic tasks arrive this year, and how widely, is still unknown.