Google Unveils 24/7 AI Agent That Buys Items For Users

Google is preparing a new always-on AI agent for Gemini that can carry out tasks around the clock, including making purchases on a user’s behalf, according to recent reports.

The development has been described across multiple tech outlets as a “24/7” personal agent designed to act more like a persistent digital partner than a traditional on-demand assistant. Reports suggest the agent is being readied as part of Google’s Gemini efforts, with some coverage referring to the project as “Gemini Agent” and others using the reported name “Remy.”

Details reported so far indicate the agent concept centers on ongoing assistance rather than single, isolated requests. In addition to making purchases, the agent is described as being able to take actions for users, implying a shift from answering questions to completing tasks.

This matters because purchase capability is a major step up in authority and risk for consumer AI tools. A system that can buy items introduces new questions about user controls, approvals, and boundaries, especially when the software is positioned as always available. If an agent can operate continuously, it also raises the stakes for how it verifies intent, manages payments, and prevents unintended orders.

The move also reflects broader competition around AI agents—tools meant to navigate apps and services, coordinate tasks, and handle routine chores with minimal prompts. For Google, building a purchase-capable agent into Gemini could expand the product beyond chat and search-style interactions and into day-to-day transactions that typically happen across retail apps, browsers, and saved payment methods.

For consumers, the practical impact would depend on how much autonomy Google ultimately gives the agent. A purchase feature could be limited to tightly controlled checkouts and explicit approvals, or it could be designed to operate with higher independence. The reporting indicates the direction is toward more persistent help, but specifics about safeguards and final functionality have not been confirmed in the provided information.

What happens next is likely more clarity from Google as the product is tested and prepared. The reports point to Google “prepping” and “readying” the agent, suggesting the work is active and that additional details—such as how purchasing is authorized, where the agent lives inside Gemini, and what devices or accounts it can use—could emerge as the company advances the project.

Until Google provides official documentation, key questions remain unresolved, including the scope of transactions, what counts as user permission, and whether the agent is tied to a single Google account profile or can operate across multiple services under one identity.

Still, the central takeaway is clear: Google is moving Gemini toward an always-on AI agent model that is described as capable of taking real actions, including buying things, not just providing answers.