Google just hosted its Google I/O 2024, and as expected, it was all about AI and there were no hardware announcements this time. There's plenty of stuff to unpack, from updates to Gemini AI, new models, and AI features in Android. However, the event heavily focused on Gemini AI and its integration into Workspace and your daily lives.

While there's not much that is new with Gemini since most of the talk was centered around Gemini 1.5 Pro which the company unveiled a few months ago with a 1 million context window, we finally saw some real-life applications of the upgrade that would be possible.

One of these that caught our eyes was the opportunity for Gemini to become an AI agent for users and undertake tasks on their behalf. The plan for Gemini from Google is to turn it into an "intelligent system" that can show reasoning, planning, and memory. Simply put, it would be able to think multiple steps ahead and work across multiple services and systems to get things done on your behalf, as a real assistant can.

In a simple demo, we saw a possible use case in the form of the user enlisting Gemini to help return their online orders, like shoes that don't fit. In this use case, you can simply snap a picture of the shoes you received and ask Gemini, through the Gemini app, that you would like to return them.

Source: Google I/O Keynote

Then, Gemini gets to work: It can search your Gmail inbox for the order receipt, locate the order number from it, and fill out the return form on the website on your behalf. Not only that, it can also schedule a pickup for the return order with the delivery service. This would all be made possible because of Gemini's huge context window, multimodal capabilities, and its deeper integration with Google Workspace, including Gmail.

Source: Google I/O Keynote

Another complex example showed how Gemini, in conjunction with Google Chrome, could help you if you've moved to a new city. It can help you explore the city, find services nearby, and even ask for your new address to update it across dozens of websites, like online food ordering and shopping websites.

There's a lot to be explored with Gemini and plenty of ways it could help you by acting on your behalf. However, there are also security and privacy concerns around such autonomy for Gemini. This is why the experiences demoed above are only prototypes, showcasing the possibilities for Gemini as an AI agent while Google explores ways to implement this that are private, secure, and work for everyone.

Already, with Gemini's integration with Gmail, there are concerns from people who are worried about how Google will use their data if they allow Gemini access to their emails.

So, while Gemini is already capable and ready to assist you with tasks such as returning your online orders, it could be a while before Gemini will actually be able to return your orders as this was just a prototype, trying to get users excited for the future. This seemed to be the theme of Google I/O this year, as most of the updates and demos were for products that would be available later this year or are in early prototypes, like Project Astra.