Google pushes Gemini agents

Google is widening Gemini from a chatbot into a built-in assistant across its apps, browser and, reportedly, an internal agent platform. (9to5google.com) On April 14, Google began rolling out Gemini’s “Personal Intelligence” globally, excluding the European Economic Area, Switzerland and the United Kingdom. The feature can pull context from Gmail, Calendar, Drive, Google Photos, YouTube, Search, Maps and other first-party apps. (9to5google.com) Google said in January that Personal Intelligence launched as a beta in the United States for Google AI Pro and Google AI Ultra subscribers, with users choosing which Google apps to connect. Google also said personal data from connected apps is not used to train Gemini directly. (blog.google) By March, Google had expanded Personal Intelligence in the United States across AI Mode in Search, the Gemini app and Gemini in Chrome, and started bringing it to free users there. Google said people were using it for tailored shopping suggestions and travel planning built from their own Google data. (blog.google) That shift puts Gemini closer to the old promise of a digital assistant: software that remembers your calendar, photos, messages and past requests instead of answering each prompt in isolation. Google has also been adding “Memory,” which lets Gemini use past chats to personalize later responses. (9to5google.com) The browser is becoming one of the main places Google is wiring that assistant in. Google says Gemini in Chrome can use the current tab by default, and on desktop users can share up to 10 open tabs so it can compare pages, summarize articles and draft messages. (support.google.com) On April 14, Google also introduced Chrome “Skills,” which turn saved prompts into one-click tools. Google’s examples include summarizing a YouTube video, pulling a recipe out of a cooking video and creating a quiz from a page you are reading. (blog.google) Google says some Chrome actions still require confirmation, including sending an email or adding a calendar event. Separate Chrome help pages describe “auto browse,” which lets Gemini complete multi-step tasks such as comparing products, finding deals, booking travel or making restaurant reservations for eligible users in the United States. (blog.google, support.google.com) Reports of a separate internal platform called “Agent” suggest Google is testing a more persistent, multi-agent setup inside Gemini and Gemini Enterprise ahead of Google I/O. That report came from Nokia Power User, which cited a leak; Google had not publicly announced such a product in the material reviewed here. (nokiapoweruser.com) Google has already signaled the direction in public. Chrome posts from late 2025 and April 2026 described “agentic” browsing and multi-step tasks, including ordering groceries and booking appointments, with Gemini working from the context of tabs and web pages. (blog.google, blog.google) The immediate change for users is not one new app but a wider spread of Gemini into the surfaces people already use: Search, Chrome and Google’s core productivity tools. The next test is whether people accept an assistant that knows more of their data and acts on more of their behalf. (blog.google, support.google.com)