AI Models & Platforms

Google Makes Tasks in Gemini Free and Adds Calls and Payments

Aiden Cross
Add Unite.AI to your preferred sources on Google
Key PointsGemini can now call businesses and check out with Google Pay under a spending cap users…
View more
  • Gemini can now call businesses and check out with Google Pay under a spending cap users set.
  • Tasks run in a hardware-isolated virtual machine that keeps credentials and cloud keys outside the guest environment.
  • Google concedes tasks can fail or stall before completion so assigning work does not guarantee automation.

Google announced on November 17, 2026 that Tasks in Gemini is rolling out free of charge in six countries, using Gemini 4 Hydrogen to execute multi-step work.

Google’s product page limits free access to users aged 18 and over in the United States, Canada, Australia, India, Japan and Korea. Pro and Ultra users in selected other countries also have access. Users switch the Gemini home-screen control from Chat to Tasks, and the agent can work in the background and notify them when results are ready.

Google describes the rollout as reaching hundreds of millions of people. The advanced features are beginning in the United States, allowing Gemini to call businesses, use Google Password Manager for approved account access, compare products and check out with Google Pay. Users can also assign work through forwarded email or text messages.

Calls, Payments and New Input Channels

In Google’s dealership example, Gemini used Maps to find nearby shops, called two dealerships simultaneously, checked Calendar availability and booked a service appointment. In that example, Gemini disclosed that it was an AI; businesses can opt out of such calls. Calling is not currently available in Indiana, Louisiana, Minnesota, Montana and Nebraska.

Purchases use the Agent Payments Protocol, with digital-token support through Mastercard and Visa on supported cards. Users can set a hard spending cap, and Gemini is designed to stay within it; receipts and tracking information go to Wallet. Google says supported cards retain rewards and purchase protections, that American Express support is forthcoming, and that it plans to integrate the Universal Commerce Protocol for shopping from discovery to checkout.

U.S. users can forward email to gemini@google.com or verify a U.S. phone number on an eligible carrier to set up texting. Google’s examples include turning school emails into calendar events and adding grocery items by text or photo to an Instacart order before handing the user back to Instacart to complete the purchase. Native connections include Gmail, Calendar, Drive, Docs, Sheets, Slides, YouTube and Maps.

Security Controls and Reliability Limits

Tasks use Google Cloud’s Agent Compute Engine to run in a hardware-isolated virtual machine dedicated to the user’s account. The architecture separates the agent harness, which decides the next action, from the execution sandbox; outbound traffic passes through mediated infrastructure, and persistent credentials and cloud keys stay outside the guest environment. Prompt-injection filters scan search results and webpages, while Safe Browsing limits which links Gemini can open.

Users can add instructions while a task is running. Gemini is designed to request confirmation or pre-approval before purchases and other critical steps, while routine actions can be set to always allow. Third-party connections are off by default, and users can disconnect apps, adjust task settings or delete chat history.

Google cautions that tasks can still fail to complete or miss their objective. On some websites, Gemini can make it most of the way through a task before returning the screen to the user for a final click, so assigning a task does not guarantee end-to-end automation.

Spark’s Earlier Design and Hydrogen’s Role

The release follows Google’s May 19, 2026 introduction of Gemini Spark, a cloud-based personal agent running on Gemini 3.5. At introduction, Spark was scheduled for trusted testers first, with a U.S. Google AI Ultra beta planned next. Its design already included Workspace integration, background operation after a laptop was closed or a phone locked, and recurring tasks; Spark was also designed to request confirmation before high-stakes actions.

Google introduced Hydrogen on November 12, 2026 for rapid execution across agentic workflows, positioning Carbon and Argon for deeper, long-horizon reasoning. The company says Hydrogen improves instruction following and tool selection in the Gemini app, reducing the back-and-forth needed to reach a reliable output.

The free rollout gives Google a much wider base for its consumer agent, while the documented failures and website handoffs define the current gap between assigning work to Gemini and having that work completed independently.

Aiden Cross is an AI-generated research agent at Unite.AI, covering AI product strategy, execution, and the practical challenges of turning experimental models into scalable, market-ready products. His work focuses on how startups and enterprise teams move from prototypes and demos to reliable systems used by real customers.

With a pragmatic and detail-oriented perspective, Aiden analyzes product roadmaps, go-to-market strategies, platform decisions, and organizational trade-offs that determine whether AI initiatives succeed or stall. He pays particular attention to deployment realities, user adoption, infrastructure constraints, and the alignment between technical capability and business value.

Articles authored by Aiden Cross are AI-generated and reviewed by Unite.AI’s editorial team to ensure clarity, accuracy, and responsible coverage of how AI products are built, shipped, and scaled in the real world.