Google is developing a task automation capability for its Gemini AI, identified as “Scheduled Actions.” This feature, currently undergoing testing, aims to let users instruct Gemini to perform specific actions automatically at predetermined future times. The functionality draws comparisons to the “Scheduled Tasks” feature available within OpenAI’s competing ChatGPT platform, which already offers a similar beta capability allowing users to automate prompts like daily news briefings or language practice.
The feature was initially spotted within the AI’s web interface by X user @testingcatalog.
Google's "Scheduled Prompts" feature is now called "Scheduled Actions". The implementation seems to remain the same, but it seems that Google is slowly preparing for the release.
— TestingCatalog News 🗞 (@testingcatalog) April 17, 2025
Earlier, traces of this feature have been spotted by @legit_api on the Android app. pic.twitter.com/SLJWJBI5fT
How Gemini Will Execute Future Tasks
The core concept allows users to define an action and schedule it for a particular time, after which Gemini is expected to execute this instruction automatically, potentially functioning even when the user is not actively interacting with the AI. Early examples suggest uses like setting timed reminders for appointments or establishing regular prompts, such as a periodic suggestion to take a break.
Utilizing this on desktop systems will require users to grant notification permissions, as alerts are necessary for confirming task completion or signaling upcoming events. Analysis of Google App code (v16.14.39) revealed strings indicating controls like “Pause” and “Resume” for these actions, alongside status markers for “Completed” and “Paused,” hinting at user management capabilities; Android Authority uncovered interface elements and code strings related to the feature.
Early indications also suggest access might be integrated within the “More from Gemini” section of a user’s Google Account management page. While currently centered on time-based execution, the capability could evolve, noting “it will eventually get better when Google is able to integrate other Google services.” This suggests potential future linking to platforms like Google Tasks or offering more complex, prompt-driven automation flows similar to ChatGPT.
Context: Gemini’s Rapid Feature Expansion
The appearance of “Scheduled Actions,” reportedly renamed from “Scheduled Prompts,” aligns with a period of accelerated updates for the Gemini ecosystem. Google recently broadened access to its Gemini 2.5 Pro model, first offered to paying subscribers on March 25, before making the experimental version available to all free users.
On April 17, Google introduced a public preview of Gemini 2.5 Flash, another iteration optimized for speed and cost-efficiency. This quick succession demonstrates Google’s push to enhance Gemini’s features. Existing integrations, such as beta extensions allowing Workspace users to interact with Calendar, Keep, and Tasks via Gemini, also indicate ongoing efforts in task management.
Pace of Development and Transparency Questions
This swift deployment strategy, however, has drawn attention to the timing of associated technical and safety documentation. The preliminary safety “model card” for Gemini 2.5 Pro, for instance, surfaced around April 16 over two weeks after its public preview commenced.
AI governance specialists expressed concerns. Kevin Bankston, a senior advisor at the Center for Democracy and Technology, described the six-page document on X as “meager” documentation, adding it tells a “troubling story of a race to the bottom on AI safety and transparency as companies rush their models to market.”
Specific critiques highlighted the absence of detailed results from internal safety procedures like red-teaming – simulated attacks designed to uncover potential harms. Thomas Woodside of the Secure AI Project further pointed out to TechCrunch that Google’s last dedicated report on dangerous capability testing was from June 2024, and it remained unconfirmed if Gemini 2.5 Pro underwent external review by US or UK AI Safety Institutes before its preview.
Highlighting broader industry concerns about the pace of AI releases versus safety checks, Sandra Wachter, a professor at the Oxford Internet Institute, told Fortune earlier in April: “If this was a car or a plane, we wouldn’t say: let’s just bring this to market as quickly as possible and we will look into the safety aspects later. Whereas with generative AI there’s an attitude of putting this out there and worrying, investigating, and fixing the issues with it later.”
Google has indicated that more comprehensive technical and safety reports are planned following the model’s general availability. The “Scheduled Actions” feature is expected by Google to become available to all users in the upcoming weeks.
