It seems that Google is preparing to integrate Gemini into the Gmail inbox, allowing the AI to handle email replies on behalf of users. The clue comes from the disassembly of the latest version of the Gmail app (version number v2026.09.28)’s APK – developers uncovered several relevant pieces of code from the installation package, indicating that this email client may be developing a smart agent powered by Gemini that can automatically process email replies.
From the comments hidden in the code, it seems that once the feature is live, Gmail’s AI inbox will display a “Use Gemini” button next to the to-do items. When users click on it, Gemini will actively browse the inbox. If there are emails that need to be replied to, an option to “Draft a reply” will appear on the interface, and the AI will first prepare a draft message.
Google did not fully hand over control to machines this time. The code shows that when Gemini is unsure about what to do next, it turns to the user for help instead; some emails also require manual approval before they are actually sent. In the corresponding strings, these human-interaction steps are marked as “requires your input” status. When the user clicks the “provide input” button and supplies the necessary details to complete the task, the AI then proceeds with the work. In other words, it is more like an assistant who needs to ask for permission before writing, rather than an agent who sends messages on your behalf without authorization.
At present, all of this is still in the evidence stage regarding the APK decomposition, and Google has not made an official announcement. The final form may also differ from what the code suggests. However, the email application has begun to integrate intelligent agents that can read messages, draft them, and know how to stop at key points to ask you questions—this has outlined a clear direction for the future: the most laborious manual operations in daily communication are gradually being taken over by AI.