Google’s new Gemini system changes the understanding of digital assistants

The new “Gemini Agent” experience prepared by Google for the artificial intelligence platform Gemini has come to the fore with more comprehensive personal assistant functions. The experimental feature, which the company introduced last year, was announced as a system that can perform multi-step tasks. The latest information shows that Google wants to transform this structure from a chat tool that only answers questions into a constantly working digital assistant that can take action on behalf of the user. The new details that emerged were obtained both from internal company documents and from lines of code in the Android application.

According to Business Insider, Google employees are currently testing the new system codenamed “Remy”. Internal documents describe Remy as a “24/7 personal agent for work, school and daily life.” In addition, it is stated that Gemini will no longer be a tool that only produces content or answers questions, but will be able to take direct action on behalf of the user. According to the explanations in the documents, the system will be able to learn user preferences over time, follow important developments and proactively perform complex tasks.

New lines discovered in the 17.20 beta version of the Google application also contain details confirming this transformation. In the codes, for Gemini Agent, “What can I do for you today?” While there is a welcome message saying, the system is described as a “24/7 digital partner”. In addition, it is stated that Gemini Agent can perform transactions on the web, work with connected applications, and perform tasks such as sharing documents, communicating or shopping on behalf of the user.

Google Gemini Agent will be able to learn user habits

According to Google’s statements, the new system uses different data sources to perform tasks. These include Gemini chat history, information from connected applications, personal context data, uploaded files and location information. In addition, it is stated that Gemini can provide more personalized results by learning user habits over time, thanks to the structure called “Personal Intelligence”. However, Google emphasizes that the system is still in the experimental stage.

Warnings shared by the company include information that incorrect transactions may be made or some data may be unintentionally disclosed. Therefore, users are asked to be careful in sensitive tasks and regularly audit the operations performed. Google also makes it clear that Gemini is not designed for transactions that require legal, financial or medical expertise. This approach reflects the influence of the recent increasing security and accuracy debates regarding generative artificial intelligence systems.

On the interface side, it seems that Gemini Agent will maintain its chat-based structure. However, it is stated that a special task panel may be included in the Gemini application. Completed tasks, ongoing processes, processes awaiting additional information from the user, and scheduled tasks can be listed under separate categories. Users will be able to pin these tasks, rename them, and then continue where they left off.

It seems that Google has recently turned to more autonomous systems on the artificial intelligence side. It is known that companies such as OpenAI, Microsoft and Anthropic are also working on similar “AI agent” solutions. However, Google’s extensive app ecosystem and Android integration may allow Gemini Agent to take on a more comprehensive role in daily usage scenarios. However, it seems that how reliable such systems will work and how they will manage user data will continue to be discussed in the coming period.