Google’s new Agent Mode might completely transform your web experience. Launching as part of Project Mariner, this digital worker bee handles complex tasks like apartment hunting while you do literally anything else. It juggles up to 10 tasks simultaneously, returning with neat summaries and actionable links. Currently exclusive to AI Ultra subscribers willing to drop $250 monthly, it represents Google’s big bet on AI delegation. The future of mindless web browsing may never be the same.
Revealed at I/O 2025, Agent Mode takes Google’s Project Mariner research and transforms it into something practical—a digital worker bee that can buzz around the web completing tasks while you kick back with a cup of coffee. Apartment hunting on Zillow? Gemini can handle that. Need sports tickets that won’t require a second mortgage? The AI’s got you covered.
What makes this different is that Gemini doesn’t need babysitting. It creates task lists and executes them sequentially, traversing websites and performing searches based on your criteria. And here’s the kicker—you don’t even need to keep the app open. The AI does its thing in the background, then returns with a neat summary and actionable links.
Technically speaking, Google’s implementation uses the Model Context Protocol introduced by Anthropic, allowing for reliable web interaction instead of the AI equivalent of “hunting and pecking” at interfaces. It’s like the difference between your tech-savvy friend using your computer versus watching your grandparent try to find the power button. This approach builds upon Google’s experience with AI Overviews that have already transformed how over a billion users interact with search results.
Of course, this digital freedom comes at a premium. Initially available only to Google AI Ultra subscribers—yes, the ones paying $250 monthly—Agent Mode is clearly positioned for early adopters and the “I’ll pay anything to be first” crowd. The rest of us peasants will have to wait.
The technology is gradually spreading across Google’s ecosystem, with Search and Chrome getting their own flavors of agentic capabilities. Project Mariner’s ability to manage 10 tasks simultaneously represents a significant advance in AI multitasking capabilities. Different implementations will cater to various user types, from power users to casual browsers. Google seems committed to this path despite growing concerns that bot traffic surpassed human traffic in April 2024.
Whether this becomes the next big thing or just another expensive tech toy remains to be seen. But one thing’s certain—Google is betting big that we’re all ready to delegate our digital lives to AI agents. And honestly, after the tenth time searching for “best pizza near me,” maybe they’re right.
